Posted: March 12th, 2023
System Analysis
see the attach file and read file
SYSTEMS ANALYSIS & DESIGN
An Object-Oriented Approach with UML
5 T H E D I T I O N
D E N N I S W I X O M T E G A R D E N
Visible Analyst is a “hands-on” tool for teaching students all aspects of analysis and design
including dynamic rules, consistency checking, managing change, and understanding the integration
issues across an IT project. Visible Analyst prepares students to enter the IT world as business or
data architects, analysts, designers, and modelers.
Visit us at www.visible.com to learn more.
YOU CAN Start Today
with the Visible Analyst!
Only takes 2 minutes to install!
Save… 33% discount!
Please visit
http://store.visible.com/Wiley.aspx
to purchase and register with your
information (see below) and obtain a
valid license for your student edition of
the software. To purchase the discounted
software you will need to enter the
following code:
978112014
Email support is provided to all registered
students at support@visible.com. Your
registration includes
the latest release of the Visible Analyst
Student Edition (software)
the Visible Analyst eTutorial
a preloaded Sample Instructional
Project
access to Webcast “How to” and “Get
Started” Instructional Videos.
Visible Analyst Student Edition
Educating tomorrow’s developers today
Disclaimer: The publisher of the textbook does not sponsor, review, or make decisions about Visible Analyst software,
and will not be responsible for, or involved in, any changes to the software.
System Analysis & Design
A n O bject -O riented A pproach with UML
Fift h Edition
Alan Dennis
Indiana University
Barbara Haley Wixom
Massachusetts Institute of Technology
David Tegarden
Virginia Tech
With contributions by Elaine Seeman,
East Carolina University
VP & EXECUTIVE PUBLISHER: Don Fowley
EXECUTIVE EDITOR: Beth Lang Golub
CONTENT EDITOR: Mary O’Sullivan
ASSOCIATE EDITOR: Ellen Keohane
MARKETING MANAGER: Christopher Ruel
ASSOCIATE PRODUCTION MANAGER: Joyce Poh
DESIGNER: Wendy Lai
Cover Image: © Christopher Boswell/Shutterstock
Th is book was set in 10/12 Minion pro by Aptara and printed and bound by Courier Kendallville. Th e cover
was printed by Courier Kendallville.
Th is book is printed on acid-free paper .
Founded in 1807, John Wiley & Sons, Inc. has been a valued source of knowledge and understanding for more
than 200 years, helping people around the world meet their needs and fulfi ll their aspirations. Our company is
built on a foundation of principles that include responsibility to the communities we serve and where we live
and work. In 2008, we launched a Corporate Citizenship Initiative, a global eff ort to address the environmental,
social, economic, and ethical challenges we face in our business. Among the issues we are addressing are carbon
impact, paper specifi cations and procurement, ethical conduct within our business and among our vendors, and
community and charitable support. For more information, please visit our website: www.wiley.com/go/citizenship.
Copyright © 2015, 2012, 2009 John Wiley & Sons, Inc. All rights reserved. No part of this publication may be repro-
duced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photo-
copying, recording, scanning or otherwise, except as permitted under Sections 107 or 108 of the 1976 United States
Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of
the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923
(Web site: www.copyright.com). Requests to the Publisher for permission should be addressed to the Permissions
Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030-5774, (201) 748-6011, fax (201) 748-6008,
or online at: www.wiley.com/go/permissions.
Evaluation copies are provided to qualifi ed academics and professionals for review purposes only, for use in
their courses during the next academic year. Th ese copies are licensed and may not be sold or transferred to a
third party. Upon completion of the review period, please return the evaluation copy to Wiley. Return instruc-
tions and a free of charge return shipping label are available at: www.wiley.com/go/returnlabel. If you have
chosen to adopt this textbook for use in your course, please accept this book as your complimentary desk copy.
Outside of the United States, please contact your local sales representative.
Library of Congress Cataloging-in-Publication Data
Dennis, Alan.
Systems analysis & design : an object-oriented approach with UML/Alan Dennis, Indiana University,
Barbara Haley Wixom, Massachusetts Institute of Technology, David Tegarden, Virginia Tech; with
contributions by Elaine Seeman, East Carolina University.–Fift h edition.
pages cm
Includes bibliographical references and index.
ISBN 978-1-118-80467-4 (pbk. : alk. paper)
1. System analysis. 2. System design. 3. UML (Computer science) I. Wixom, Barbara Haley,
1969-II. Tegarden, David Paul. III. Seeman, Elaine. IV. Title. V. Title: System analysis and design.
QA402.D395 2015
004.2’1–dc23
2014048338
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
PURPOSE OF THIS BOOK
Systems Analysis and Design (SAD) is an exciting, active fi eld in which analysts continually
learn new techniques and approaches to develop systems more eff ectively and effi ciently.
However, there is a core set of skills that all analysts need to know—no matter what
approach or methodology is used. All information systems projects move through the four
phases of planning, analysis, design, and implementation; all projects require analysts to
gather requirements, model the business needs, and create blueprints for how the system
should be built; and all projects require an understanding of organizational behavior con-
cepts like change management and team building. Today, the cost of developing modern
soft ware is composed primarily of the cost associated with the developers themselves and
not the computers. As such, object-oriented approaches to developing information systems
hold much promise in controlling these costs.
Today, the most exciting change to systems analysis and design is the move to
object-oriented techniques, which view a system as a collection of self-contained objects
that have both data and processes. Th is change has been accelerated through the crea-
tion of the Unifi ed Modeling Language (UML). UML provides a common vocabulary of
object-oriented terms and diagramming techniques that is rich enough to model any sys-
tems development project from analysis through implementation.
Th is book captures the dynamic aspects of the fi eld by keeping students focused on
doing SAD while presenting the core set of skills that we feel every systems analyst needs to
know today and in the future. Th is book builds on our professional experience as systems
analysts and on our experience in teaching SAD in the classroom.
Th is book will be of particular interest to instructors who have students do a major
project as part of their course. Each chapter describes one part of the process, provides
clear explanations on how to do it, gives a detailed example, and then has exercises for the
students to practice. In this way, students can leave the course with experience that will
form a rich foundation for further work as a systems analyst.
OUTSTANDING FEATURES
A Focus on Doing SAD
Th e goal of this book is to enable students to do SAD—not just read about it, but under-
stand the issues so that they can actually analyze and design systems. Th e book introduces
each major technique, explains what it is, explains how to do it, presents an example, and
provides Your Turn opportunities with each chapter for students to practice each new tech-
nique before they do it for real in a project. Th e Your Turn boxes are posted online at www.
wiley.com/college/dennis. Aft er reading each chapter, the student will be able to perform
that step in the system development process.
P R E F A C E
v
vi Preface
Rich Examples of Success and Failure
Th is book has a running online case study (accessible from www.wiley.com/go/dennis/
casestudy ) about a fi ctitious health care company called Patterson Superstore. Each chapter of
the case study shows how the concepts are applied in situations at Patterson Superstore. In
this way, the running case serves as a template that students can apply to their own work.
Each chapter also includes numerous Concepts in Action boxes, which are posted online at
www.wiley.com/college/dennis. Th ese boxes describe how real companies succeeded—and
failed—in performing the activities in the chapter. Many of these examples are drawn from
our own experiences as systems analysts.
Real World Focus
The skills that students learn in a systems analysis and design course should mirror
the work that they ultimately will do in real organizations. We have tried to make this
book as “real” as possible by building extensively on our experience as professional sys-
tems analysts for organizations, such as Arthur Andersen, IBM, the U.S. Department
of Defense, and the Australian Army. We have also worked with a diverse industry
advisory board of IS professionals and consultants in developing the book and have
incorporated their stories, feedback, and advice throughout. Many students who use
this book will eventually use the skills on the job in a business environment, and we
believe they will have a competitive edge in understanding what successful practition-
ers feel is relevant in the real world.
Project Approach
We have presented the topics in this book in the order in which an analyst encounters them
in a typical project. Although the presentation is necessarily linear (because students have
to learn concepts in the way in which they build on each other), we emphasize the iterative,
complex nature of SAD as the book unfolds. Th e presentation of the material should align
well with courses that encourage students to work on projects because it presents topics as
students need to apply them.
WHAT’S NEW IN THIS EDITION
■ A completely new, expanded case study on an integrated health clinic delivery
system has been written to accompany the fi ft h e dition. Th e entire case study is
posted online. At the end of each chapter in the text, a short synopsis of the case
is provided.
■ Th e text has been streamlined to focus on the essentials and therefore, to enhance
student understanding. Selected m aterial s like the “Your Turn” and “Concepts in
Action” boxes have been moved online and can be accessed at www.wiley.com/
college/dennis .
■ Th roughout the book , there is a greater emphasis on verifying, validating, and
testing, as well as the incremental and iterative development of systems.
■ In Chapter 2, there is more content on Agile techniques , including scrum meet-
ings, product backlog, and sprints.
■ In Chapter 3, we have increased focus on soft ware quality and user stories.
■ We have added new examples throughout the book and clarifi ed explanations to
help students learn some of the more diffi cult concepts.
Preface vii
■ Chapter 10 includes more coverage of mobile computing , including specifi cs on
navigation, input, and output. Th is chapter also has a new section on games,
multidimensional information visualization, augmented reality, and virtual reality.
■ Chapter 11 includes new material o n ubiquitous computing and the Internet of Th ings.
■ Testing has been expanded in Chapter 12.
ORGANIZATION OF THIS BOOK
Th is book is loosely organized around the phases and workfl ows of the enhanced Unifi ed
Process. Each chapter has been written to teach students specifi c tasks that analysts need
to accomplish over the course of a project, and the deliverables that will be produced from
the tasks. As students complete the chapters, they will realize the iterative and incremental
nature of the tasks in object-oriented systems development.
Chapter 1 introduces the SDLC, systems development methodologies, roles and
skills needed for a systems analyst, the basic characteristics of object-oriented systems,
object-oriented systems analysis, the Unifi ed Process, and the UML. Chapter 2 presents
topics related to the project management workfl ow of the Unifi ed Process, including pro-
ject identifi cation, system request, feasibility analysis, project selection, traditional project
management tools (including work breakdown structures, network diagrams, and PERT
analysis), project eff ort estimation using use-case points, evolutionary work breakdown
structures, iterative workplans, scope management, timeboxing, risk management, and
staffi ng the project. Chapter 2 also addresses issues related to the Environment and Infra-
structure management workfl ows of the Unifi ed Process.
Part One focuses on creating analysis models. Chapter 3 introduces students to an assort-
ment of requirements analysis strategies a variety of requirements-gathering techniques that
are used to determine the functional and nonfunctional requirements of the system, and to a
system proposal. Chapter 4 focuses on constructing business process and functional models
using use – case diagrams, activity diagrams, and use – case descriptions. Chapter 5 addresses
producing structural models using CRC cards, class diagrams, and object diagrams. Chapter 6
tackles creating behavioral models using sequence diagrams, communication diagrams,
behavioral state machines, and CRUDE analysis and matrices. Chapters 4 through 6 also
cover the verifi cation and validation of the models described in each chapter.
Part Two addresses design modeling. In Chapter 7, students learn how to verify and
validate the analysis models created during analysis modeling and to evolve the analysis
models into design models via the use of factoring, partitions, and layers. Th e students also
learn to create an alternative matrix that can be used to compare custom, packaged, and
outsourcing alternatives. Chapter 8 concentrates on designing the individual classes and
their respective methods through the use of contracts and method specifi cations. Chapter 9
presents the issues involved in designing persistence for objects. Th ese issues include the
diff erent storage formats that can be used for object persistence, how to map an object-
oriented design into the chosen storage format, and how to design a set of data access and
manipulation classes that act as a translator between the classes in the application and
the object persistence. Th is chapter also focuses on the nonfunctional requirements that
impact the data management layer. Chapter 10 presents the design of the human–computer
interaction layer, where students learn how to design user interfaces using use scenarios,
windows navigation diagrams, storyboards, windows layout diagrams, user interface
prototypes, real use cases, interface standards, and user interface templates; to perform
user interface evaluations using heuristic evaluation, walkthrough evaluation, interactive
evaluation, and formal usability testing; and to address nonfunctional requirements such
viii Preface
as user interface layout, content awareness, aesthetics, user experience, and consistency.
Th is chapter also addresses issues related to mobile computing, social media, games,
multi dimensional information visualizations, immersive environments, and international
and cultural issues with regard to user interface design. Chapter 11 focuses on the phys-
ical architecture and infrastructure design, which includes deployment diagrams and
hardware/soft ware specifi cation. In today’s world, this also includes issues related to cloud
computing, ubiquitous computing, the Internet of things, and green IT. Th is chapter, like
the previous design chapters, covers the impact that nonfunctional requirements can have
on the physical architecture layer.
Part Th ree provides material that is related to the construction, installation, and operations
of the system. Chapter 12 focuses on system construction, where students learn how to build,
test, and document the system. Installation and operations are covered in Chapter 13, where
students learn about the conversion plan, change management plan, support plan, and project
assessment. Additionally, these chapters address the issues related to developing systems in a fl at
world, where developers and users are distributed throughout the world.
SUPPLEMENTS www.wiley.com/college/dennis
Instructor Book Companion Web s ite
■ PowerPoint slides : I nstructors can tailor the slides to their classroom needs .
S tudents can use them to guide their reading and studying activities.
■ Test Bank : I ncludes a variety of questions ranging from multiple-choice, true/
false, and short answer questions. A computerized, Respondus version of the Test
Bank is also available.
■ Instructor’s Manual : P rovides resources to support the instructor both inside
and out of the classroom. Th e manual includes short experiential exercises that
instr uctors can use to help students experience and understand key topics in
each chapter. Short stories have been provided by people working in both corpo-
rate and consulting environments for instructors to insert into lectures to make
concepts more colorful and real. Additional minicases for every chapter allow
students to perform some of the key concepts that were learned in the chapter.
Solutions to end of chapter questions and exercises are provided.
Student Book Companion Web s ite
■ A collection of templates and worksheets consisting of electronic versions of
selected fi gures from the book.
■ A completely new, expanded case study on an integrated health clinic delivery
system has been written to accompany the fi ft h edition. Th is case study is online
only. It can be accessed at www.wiley.com/go/dennis/casestudy .
■ “Your Turn” and “Concepts in Action” boxes from the fourth edition have been
moved online and can be accessed from the student companion site.
Wiley E-Text: Powered by VitalSource
Th is Wiley e-text off ers students continuing access to materials for their course. Your students
can access content on a mobile device, online from any Internet-connected computer, or by
a computer via download. With dynamic features built into this e-text, students can search
across content, highlight, and take notes that they can share with teachers and classmates.
Preface ix
Visible Analyst
Wiley has partnered with Visible Analyst to give students a discounted price for Visible
Analyst soft ware, an intuitive modeling tool for all aspects of traditional or object-oriented
systems analysis and design. All new copies of the text will have a Key Code (printed on
a page near the front of this text) that will provide a discount on Visible Analyst soft ware.
To obtain the soft ware, students should visit http://store.visible.com/Wiley.aspx and enter
their Key Code. Students who buy a new print text or digital e-book will receive one-third
off the price of a downloadable edition of the soft ware with a 6-month license. With the
soft ware, they will also receive tutorials, how-to videos, and a sample project. Students who
buy used copies of this text may buy Visible Analyst at full price using the URL provided.
Project Management Soft ware
You can download a 60-day trial of Microsoft Project Professional 2013 from the following
Web site: www.microsoft .com/en-us/evalcenter/evaluate-project-professional-2013 . Note
that Microsoft has changed its policy and no longer off ers the 120-day trial previously
available.
Another option now available to education institutions adopting this Wiley titl e is a
free introductory 3-year membership for DreamSpark Premium. DreamSpark Premium
is designed to provide the easiest and most inexpensive way for academic departments
to make the latest Microsoft soft ware available in labs, classrooms, and on student and
instructor PCs. Microsoft Project soft ware is available through this Wiley and Microsoft
publishing partnership, free of charge with the adoption of any qualifi ed Wiley title. Each
copy of Microsoft Project is the full version of the soft ware, with no time limitation, and
can be used indefi nitely for educational purposes. Contact your Wiley sales representative
for details. For more information about the DreamSpark Premium program, contact
drmspkna@Microsoft .com .
ACKNOWLEDGMENTS
Th anks to Elaine Seeman for her feedback on every chapter in this book as well as for her
work writing the new online case study. We would like to thank the following reviewers
for their helpful and insightful comments on the fi ft h edition: Mohammad Dadashzadeh,
Oakland University; Xiaodong Deng, Oakland University ; Th omas W. Dillon, James
Madison University; Bryan Goda, University of Washington, Tacoma; Kathleen S. Hartzel,
Duquesne University; Rajkumar Kempaiah, Stevens Institute of Technology; Sung-kwan
Kim, University of Arkansas at Little Rock; Richard McCarthy, Quinnipiac University;
Donald McCracken, Grantham University; Osama A. Morad, Southern New Hampshire
University; Fred Niederman, Saint Louis University; Linda Plotnick, Jacksonville State
University; Vladimir V. Riabov, Rivier University ; Richard Schilhavy, Guilford College;
Tod Sedbrook, University of Northern Colorado; Steven C. Shaff er, Penn State University;
Michael Smith, Georgia Institute of Technology; and John Wetsch, Southern New Hampshire
University.
We would also like to thank the following reviewers for their helpful and insight-
ful comments on the fi rst, second, third , and fourth editions: Evans Adams, Fort Lewis
College; Murugan Anandarajon, Drexel University; Ron Anson, Boise State University;
Noushin Ashrafi , University of Massachusetts, Boston; Dirk Baldwin, University of
Wisconsin-Parkside; Robert Barker, University of Louisville; Qing Cao, University of
Missouri–Kansas City; David Champion, DeVry University, Columbus, OH campus; Jeff
Cummings, Indiana University; Junhua Ding, East Carolina University; Robert Dollinger,
x Preface
University of Wisconsin-Stevens Point; Abhijit Dutt, Carnegie Mellon University; Terry
Fox, Baylor University; Ahmad Ghafarian, North Georgia College & State U niversity; Donald
Golden, Cleve land State University; Cleotilde Gonzalez, Carnegie Melon University;
Daniel V. Goulet, University of Wisconsin–Stevens Point; Harvey Hayashi, Loyalist College
of Applied Arts and Technology; Yujong Hwang, DePaul University; Scott James, Saginaw
Valley State University; Zongliang Jiang, North Carolina A&T State University; Raymond
Kirsch, La Salle University; Rajiv Kishore, State University of New York–Buff alo; Ravindra
Krovi, University of Akron; Jean-Piere Kuilboer, University of Massachusetts, Boston;
Gilliean Lee, Lander University; Leo Legorreta, California State University Sacramento;
Diane Lending, James Madison University; Steve Machon, DeVry University; Fernando
Maymí , West Point University; Daniel Mittleman, DePaulUniversity; Makoto Nakayama,
DePaul University; Fred Niederman, Saint Louis University; Parasuraman Nurani, DeVry
University; H. Robert Pajkowski, DeVry Institute of Technology, Scarborough, Ontario;
June S. Park, University of Iowa; Graham Peace, West Virginia University; Tom Pettay,
DeVry Institute of Technology, Columbus,Ohio; Selwyn Piramuthu, University of Florida;
J. Drew Procaccino, Rider University; Neil Ramiller, Portland State University; Eliot
Rich, University at Albany, State University of New York; Marcus Rothenberger, University
of Wisconsin–Milwaukee; Carl Scott, University of Houston; Keng Siau,University of
Nebraska–Lincoln; Ift ikhar Sikder , Cleveland State University; Jonathan Trower, Baylor
University; June Verner, Drexel University; Anna Wachholz, Sheridan College; Bill Watson,
Indiana University- Purdue University Indianapolis; Randy S.Weinberg, Carnegie Mellon
University; Eli J.Weissman, DeVry Institute of Technology, Long Island City, NY; Heinz
Roland Weistroff er, Virginia Commonwealth University; Amy Wilson, DeVry Institute of
Technology, Decatur, GA; Amy Woszczynski, Kennesaw State University; Vincent C. Yen,
Wright State University ; Fan Zhao, Florida Gulf Coast University; and Dan Zhu, Iowa State
University.
xi
C O N T E N T S
Preface v
Chapter 1
Introduction to Systems
Analysis and Design 1
Introduction 1
The Systems Development Life Cycle 2
Planning 3
Analysis 3
Design 4
Implementation 4
Systems Development Methodologies 5
Structured Design 6
Rapid Application Development (RAD) 8
Agile Development 12
Selecting the Appropriate Development
Methodology 15
Typical Systems Analyst Roles and Skills 17
Business Analyst 18
Systems Analyst 18
Infrastructure Analyst 18
Change Management Analyst 19
Project Manager 19
Basic Characteristics of Object-Oriented
Systems 19
Classes and Objects 19
Methods and Messages 20
Encapsulation and Information Hiding 20
Inheritance 21
Polymorphism and Dynamic Binding 22
Object-Oriented Systems Analysis
and Design (OOSAD) 23
Use-Case Driven 24
Architecture-Centric 24
Iterative and Incremental 24
Benefi ts of Object-Oriented Systems
Analysis and Design 25
The Unified Process 25
Phases 26
Workfl ows 28
Extensions to the Unifi ed Process 30
The Unified Modeling Language 34
applying the concepts at patterson
superstore 36
Chapter Review 36
Chapter 2
Project Management 41
Introduction 41
Project Identification 43
System Request 44
Feasibility Analysis 45
Technical Feasibility 45
Economic Feasibility 46
Organizational Feasibility 51
Project Selection 53
Traditional Project Management Tools 54
Work Breakdown Structures 55
Gantt Chart 56
Network Diagram 57
Project Effort Estimation 58
Creating and Managing the Workplan 63
Evolutionary Work Breakdown
Structures and Iterative Workplans 63
Managing Scope 67
Timeboxing 68
Refi ning Estimates 69
Managing Risk 70
Staffing the Project 71
Characteristics of a Jelled Team 71
Staffi ng Plan 73
Motivation 75
Handling Confl ict 76
Environment and Infrastructure
Management 76
CASE Tools 77
Standards 77
Documentation 78
Applying the Concepts at Patterson
Superstore 80
Chapter Review 80
■ PART ONE
ANALYSIS MODELING 85
Chapter 3
Requirements
Determination 86
Introduction 86
Requirements Determination 87
Defi ning a Requirement 87
Requirements Defi nition 89
Determining Requirements 89
Creating a Requirements Defi nition 91
Real-World Problems with Requirements
Determination 91
Requirements Analysis Strategies 92
Problem Analysis 92
Root Cause Analysis 92
Duration Analysis 93
Activity-Based Costing 94
Informal Benchmarking 94
Outcome Analysis 95
Technology Analysis 95
Activity Elimination 95
Requirements-Gathering Techniques 95
Interviews 96
Joint Application Development (JAD) 100
Questionnaires 104
Document Analysis 106
Observation 108
Selecting the Appropriate Techniques 108
Alternative Requirements Documentation
Techniques 110
Concept Maps 110
User Stories 112
The System Proposal 113
Applying the Concepts at Patterson
Superstore 114
Chapter review 114
Chapter 4
Business Process and
Functional Modeling 119
Introduction 119
Business Process Identification with Use
Cases and Use-Case Diagrams 121
Elements of Use-Case Diagrams 121
Identifying the Major Use Cases 126
Creating a Use-Case Diagram 127
Business Process Modeling with Activity
Diagrams 129
Elements of an Activity Diagram 131
Guidelines for Creating Activity
Diagrams 136
Creating Activity Diagrams 137
Business Process Documentation with Use
Cases and Use-Case Descriptions 140
Types of Use Cases 141
Elements of a Use-Case Description 141
Guidelines for Creating Use-Case
Descriptions 145
Creating Use Case Descriptions 146
Verifying and Validating the Business
Processes and Functional Models 153
Verifi cation and Validation through
Walkthroughs 153
Functional Model Verifi cation and
Validation 154
Applying the Concepts at Patterson
Superstore 157
Chapter Review 157
Chapter 5
Structural Modeling 163
Introduction 163
Structural Models 164
Classes, Attributes, and
Operations 164
Relationships 165
Object Identification 166
Textual Analysis 166
Brainstorming 167
Common Object Lists 169
Patterns 169
Crc Cards 172
Responsibilities and Collaborations 172
Elements of a CRC Card 173
Role-Playing CRC Cards with
Use Cases 174
Class Diagrams 176
Elements of a Class Diagram 176
Simplifying Class Diagrams 184
Object Diagrams 184
Creating Structural Models Using
CRC Cards and Class Diagrams 185
Campus Housing Example 187
Library Example 187
xii Contents
Verifying and Validating the Structural
Model 194
Applying the Concepts at Patterson
Superstore 197
Chapter Review 198
Chapter 6
Behavioral Modeling 202
Introduction 202
Behavioral Models 203
Interaction Diagrams 204
Objects, Operations, and Messages 204
Sequence Diagrams 204
Communication Diagrams 216
Behavioral State Machines 221
States, Events, Transitions, Actions, and
Activities 221
Elements of a Behavioral State Machine 222
Creating a Behavioral State Machine 226
Crude Analysis 229
Verifying and Validating the Behavioral
Model 233
Applying the Concepts at Patterson
Superstore 235
Chapter Review 235
■ PART TWO
DESIGN MODELING 239
Chapter 7
Moving on to Design 240
Introduction 240
Verifying and Validating the Analysis
Models 242
Balancing Functional and Structural
Models 242
Balancing Functional and Behavioral
Models 243
Balancing Structural and Behavioral
Models 251
Summary 254
Evolving the Analysis Models into Design
Models 257
Factoring 257
Partitions and Collaborations 258
Layers 259
Packages and Package Diagrams 262
Guidelines for Creating Package
Diagrams 264
Creating Package Diagrams 266
Verifying and Validating Package
Diagrams 266
Design Strategies 268
Custom Development 268
Packaged Soft ware 269
Outsourcing 270
Selecting a Design Strategy 272
Selecting an Acquisition Strategy 273
Alternative Matrix 274
Applying the Concepts at Patterson
Superstore 276
Chapter Review 276
Chapter 8
Class and Method Design 280
Introduction 280
Review of the Basic Characteristics
of Object Orientation 282
Classes, Objects, Methods, and Messages 282
Encapsulation and Information Hiding 282
Polymorphism and Dynamic Binding 282
Inheritance 284
Design Criteria 286
Coupling 286
Cohesion 289
Connascence 292
Object Design Activities 293
Adding Specifi cations 293
Identifying Opportunities for Reuse 294
Restructuring the Design 297
Optimizing the Design 298
Mapping Problem-Domain Classes to
Implementation Languages 300
Constraints and Contracts 304
Types of Constraints 306
Elements of a Contract 306
Method Specification 314
General Information 314
Events 314
Message Passing 315
Algorithm Specifi cations 316
Example 318
Verifying and Validating Class and Method
Design 319
Contents xiii
Applying the Concepts at Patterson
Superstore 322
Chapter review 322
Chapter 9
Data Management Layer
Design 326
Introduction 326
Object Persistence Formats 327
Sequential and Random Access Files 327
Relational Databases 330
Object-Relational Databases 332
Object-Oriented Databases 332
NoSQL Data Stores 333
Selecting an Object Persistence Format 335
Mapping Problem Domain Objects to Object
Persistence Formats 337
Mapping Problem Domain Objects to an
OODBMS Format 338
Mapping Problem Domain Objects to an
ORDBMS Format 341
Mapping Problem Domain Objects to a
RDBMS Format 344
Optimizing Rdbms-Based Object
Storage 346
Optimizing Storage Effi ciency 347
Optimizing Data Access Speed 351
Estimating Data Storage Size 356
Designing Data Access and Manipulation
Classes 357
Nonfunctional Requirements and Data
Management Layer Design 360
Verifying and Validating the Data
Management Layer 361
Applying the Concepts at Patterson
Superstore 362
Chapter Review 362
Chapter 10
Human–Computer Interaction
Layer Design 367
Iintroduction 367
Principles for User Interface Design 368
Layout 369
Content Awareness 369
Aesthetics 370
User Experience 371
Consistency 371
Minimizing User Eff ort 372
User Interface Design Process 372
Use Scenario Development 373
Navigation Structure Design 375
Interface Standards Design 376
Interface Design Prototyping 377
Interface Evaluation 380
Common Sense Approach to User
Interface Design 382
Navigation Design 383
Basic Principles 383
Types of Navigation Controls 384
Messages 386
Navigation Design Documentation 387
Input Design 387
Basic Principles 387
Types of Inputs 390
Input Validation 391
Output Design 392
Basic Principles 392
Types of Outputs 394
Media 394
Mobile Computing and User Interface
Design 395
Social Media and User Interface
Design 398
Games, Multi-Dimensional Information
Visualizations, and Immersive
Environments 400
Games, Gamifi cation, and User Interface
Design 400
Multidimensional Information Visualization
Design 402
User Interface Design and Immersive
Environments 404
International and Cultural Issues and User
Interface Design 406
Multilingual Requirements 406
Color 407
Cultural Diff erences 407
Nonfunctional Requirements And Human-
Computer Interaction Layer
Design 410
Applying The Concepts At Patterson
Superstore 411
Chapter review 411
xiv Contents
Chapter 11
Physical Architecture Layer
Design 418
Introduction 418
Elements of the Physical Architecture
Layer 419
Architectural Components 419
Server-Based Architectures 420
Client-Based Architectures 420
Client–Server Architectures 421
Client–Server Tiers 422
Selecting a Physical Architecture 424
Cloud Computing 426
Ubiquitous Computing and the Internet
of Things 428
Green IT 431
Infrastructure Design 432
Deployment Diagram 432
Network Model 434
Hardware and System Software
Specifications 438
Nonfunctional Requirements and Physical
Architecture Layer Design 440
Operational Requirements 441
Performance Requirements 442
Security Requirements 444
Cultural and Political Requirements 447
Synopsis 448
Verifying and Validating the Physical
Architecture Layer 449
Applying the Concepts at Patterson
Superstore 450
Chapter Review 450
■ PART THREE
CONSTRUCTION, INSTALLATION,
AND OPERATIONS 455
Chapter 12
Construction 456
Introduction 456
Managing Programming 457
Assigning Programmers 457
Coordinating Activities 458
Managing the Schedule 458
Cultural Issues 460
Developing Documentation 462
Types of Documentation 463
Designing Documentation Structure 463
Writing Documentation Topics 465
Identifying Navigation Terms 465
Designing Tests 467
Testing and Object Orientation 468
Test Planning 469
Unit Tests 471
Integration Tests 475
System Tests 476
Acceptance Tests 477
Applying the Concepts at Patterson
Superstore 478
Chapter Review 478
Chapter 13
Installation and
Operations 481
Introduction 481
Cultural Issues and Information
Technology Adoption 483
Conversion 485
Conversion Style 486
Conversion Location 486
Conversion Modules 487
Selecting the Appropriate Conversion
Strategy 488
Change Management 489
Understanding Resistance to Change 490
Revising Management Policies 491
Assessing Costs and Benefi ts 492
Motivating Adoption 493
Enabling Adoption: Training 495
Post-Implementation Activities 497
System Support 497
System Maintenance 498
Project Assessment 500
Applying the Concepts at Patterson
Superstore 502
Chapter Review 502
Index 507
Contents xv
Chapter 1 introduces the systems development life cycle (SDLC), the fundamental four-
phase model (planning, analysis, design, and implementation) common to all information
systems development projects. It describes the evolution of system development method-
ologies and discusses the roles and skills required of a systems analyst. Th e chapter then
overviews the basic characteristics of object-oriented systems and the fundamentals of
object-oriented systems analysis and design and closes with a description of the Unifi ed
Process and its extensions and the Unifi ed Modeling Language.
OBJECTIVES
■ Understand the fundamental systems development life cycle and its four phases
■ Understand the evolution of systems development methodologies
■ Be familiar with the diff erent roles played by and the skills of a systems analyst
■ Be familiar with the basic characteristics of object-oriented systems
■ Be familiar with the fundamental principles of object-oriented systems analysis
and design
■ Be familiar with the Unifi ed Process, its extensions, and the Unifi ed Modeling
Language
INTRODUCTION
The systems development life cycle (SDLC) is the process of understanding how an infor-
mation system (IS) can support business needs by designing a system, building it, and
delivering it to users. If you have taken a programming class or have programmed on
your own, this probably sounds pretty simple. Unfortunately, it is not. A 1996 survey by
the Standish Group found that 42 percent of all corporate IS projects were abandoned
before completion. A similar study conducted in 1996 by the General Accounting Office
found 53 percent of all U.S. government IS projects were abandoned. Unfortunately,
many of the systems that are not abandoned are delivered to the users significantly late,
cost far more than planned, and have fewer features than originally planned. For exam-
ple, IAG Consulting reports that 80 percent of the projects were over time, 72 percent
were over budget, and 55 percent contained less than the full functionality; Panorama
Consulting Solutions reports that 54 percent of the ERP projects were over time, 56 percent
were over budget, and 48 percent delivered less than 50 percent of the initial benefi ts;
and an IBM study reports that 59 percent of the projects missed one or more of on time,
within budget, and quality constraints.1 Although we would like to promote this book as
a silver bullet that will keep you from IS failures, we readily admit that a silver bullet that
guarantees IS development success simply does not exist. Instead, this book provides you
1
C H A P T E R 1
Introduction to Systems
Analysis and Design
2 Chapter 1 Introduction to Systems Analysis and Design
with several fundamental concepts and many practical techniques that you can use to
improve the probability of success.
Th e key person in the SDLC is the systems analyst, who analyzes the business situation,
identifi es opportunities for improvements, and designs an information system to implement
them. Being a systems analyst is one of the most interesting, exciting, and challenging jobs
around. Systems analysts work with a variety of people and learn how they conduct business.
Specifi cally, they work with a team of systems analysts, programmers, and others on a com-
mon mission. Systems analysts feel the satisfaction of seeing systems that they designed and
developed make a signifi cant business impact, knowing that they contributed unique skills to
make that happen.
However, the primary objective of a systems analyst is not to create a wonderful sys-
tem; instead, it is to create value for the organization, which for most companies means
increasing profi ts (government agencies and not-for-profi t organizations measure value
diff erently). Many failed systems have been abandoned because the analysts tried to build a
wonderful system without clearly understanding how the system would fi t with an organi-
zation’s goals, current business processes, and other information systems to provide value.
An investment in an information system is like any other investment. Th e goal is not to
acquire the tool, because the tool is simply a means to an end; the goal is to enable the
organization to perform work better so that it can earn greater profi ts or serve its constit-
uents more eff ectively.
Th is book introduces the fundamental skills a systems analyst needs. Th is pragmatic book
discusses best practices in systems development; it does not present a general survey of systems
development that covers everything about the topic. By defi nition, systems analysts do things
and challenge the current way that organizations work. To get the most out of this book, you
will need to actively apply to your own systems development project the ideas and concepts in
the examples. Th is book guides you through all the steps for delivering a successful informa-
tion system. By the time you fi nish the book, you won’t be an expert analyst, but you will be
ready to start building systems for real.
THE SYSTEMS DEVELOPMENT LIFE CYCLE
In many ways, building an information system is similar to building a house. First, the house
(or the information system) starts with a basic idea. Second, this idea is transformed into a
simple drawing that is shown to the customer and refi ned (oft en through several drawings,
each improving on the last) until the customer agrees that the picture depicts what he or she
wants. Th ird, a set of blueprints is designed that presents much more detailed information about
the house (e.g., the type of water faucets or where the telephone jacks will be placed). Finally,
the house is built following the blueprints, oft en with some changes directed by the customer
as the house is erected.
Th e SDLC has a similar set of four fundamental phases: planning, analysis, design, and
implementation. Diff erent projects might emphasize diff erent parts of the SDLC or approach the
SDLC phases in diff erent ways, but all projects have elements of these four phases. Each phase is
itself composed of a series of steps, which rely upon techniques that produce deliverables (specifi c
documents and fi les that provide understanding about the project).
1 For more information on the problem, see Capers Jones, Patterns of Soft ware System Failure and Success (London:
International Th ompson Computer Press, 1996); KeithEllis, Business Analysis Benchmark: Th e Impact of Business
Requirements on the Success of Technology Projects (2008). Retrieved May 2014 from IAG Consulting, www.iag.biz;
H. H. Jorgensen, L. Owen, and A. Neus, Making Change Work (2008). Retrieved May 2014 from IBM, www.ibm.
com; Panorama Consulting Solutions, 2012 ERP Report (2012). Retrieved May 2014 from Panorama- Consulting.com.
The Systems Development Life Cycle 3
For example, in applying for admission to a university, all students go through the same
phases: information gathering, applying, and accepting. Each of these phases has steps; for
example, information gathering includes steps such as searching for schools, requesting infor-
mation, and reading brochures. Students then use techniques (e.g., Internet searching) that
can be applied to steps (e.g., requesting information) to create deliverables (e.g., evaluations
of diff erent aspects of universities).
In many projects, the SDLC phases and steps proceed in a logical path from start to fi n-
ish. In other projects, the project teams move through the steps consecutively, incrementally,
iteratively, or in other patterns. In this section, we describe the phases, the actions, and some
of the techniques that are used to accomplish the steps at a very high level.
For now, there are two important points to understand about the SDLC. First, you should
get a general sense of the phases and steps through which IS projects move and some of the
techniques that produce certain deliverables. Second, it is important to understand that the
SDLC is a process of gradual refi nement. Th e deliverables produced in the analysis phase pro-
vide a general idea of the shape of the new system. Th ese deliverables are used as input to the
design phase, which then refi nes them to produce a set of deliverables that describes in much
more detailed terms exactly how the system will be built. Th ese deliverables, in turn, are used
in the implementation phase to produce the actual system. Each phase refi nes and elaborates
on the work done previously.
Planning
Th e planning phase is the fundamental process of understanding why an information sys-
tem should be built and determining how the project team will go about building it. It has
two steps:
1. During project initiation, the system’s business value to the organization is identifi ed:
How will it lower costs or increase revenues? Most ideas for new systems come from
outside the IS area (e.g., from the marketing department, accounting department) in
the form of a system request. A system request presents a brief summary of a business
need, and it explains how a system that supports the need will create business value.
Th e IS department works together with the person or department that generated the
request (called the project sponsor) to conduct a feasibility analysis.
Th e system request and feasibility analysis are presented to an information sys-
tems approval committee (sometimes called a steering committee), which decides
whether the project should be undertaken.
2. Once the project is approved, it enters project management. During project man-
agement, the project manager creates a workplan, staff s the project, and puts tech-
niques in place to help the project team control and direct the project through
the entire SDLC. Th e deliverable for project management is a project plan, which
describes how the project team will go about developing the system.
Analysis
Th e analysis phase answers the questions of who will use the system, what the system will
do, and where and when it will be used. During this phase, the project team investigates any
current system(s), identifi es opportunities for improvement, and develops a concept for the
new system.
Th is phase has three steps:
1. An analysis strategy is developed to guide the project team’s eff orts. Such a strategy
usually includes an analysis of the current system (called the as-is system) and its
problems and then ways to design a new system (called the to-be system).
4 Chapter 1 Introduction to Systems Analysis and Design
2. Th e next step is requirements gathering (e.g., through interviews or questionnaires).
Th e analysis of this information—in conjunction with input from the project
sponsor and many other people—leads to the development of a concept for a new
system. Th e system concept is then used as a basis to develop a set of business
analysis models, which describe how the business will operate if the new system
is developed.
3. Th e analyses, system concept, and models are combined into a document called
the system proposal, which is presented to the project sponsor and other key deci-
sion makers (e.g., members of the approval committee) who decide whether the
project should continue to move forward.
Th e system proposal is the initial deliverable that describes what business requirements the
new system should meet. Because it is really the fi rst step in the design of the new system,
some experts argue that it is inappropriate to use the term “analysis” as the name for this
phase; some argue a better name would be “analysis and initial design.” Most organizations
continue to use the name analysis for this phase, however, so we use it in this book as well. Just
keep in mind that the deliverable from the analysis phase is both an analysis and a high-level
initial design for the new system.
Design
Th e design phase decides how the system will operate, in terms of the hardware, soft ware,
and network infrastructure; the user interface, forms, and reports; and the specifi c programs,
databases, and fi les that will be needed. Although most of the strategic decisions about the
system were made in the development of the system concept during the analysis phase, the
steps in the design phase determine exactly how the system will operate. Th e design phase
has four steps:
1. Th e design strategy is fi rst developed. It clarifi es whether the system will be devel-
oped by the company’s own programmers, whether the system will be outsourced
to another fi rm (usually a consulting fi rm), or whether the company will buy an
existing soft ware package.
2. Th is leads to the development of the basic architecture design for the system, which
describes the hardware, soft ware, and network infrastructure to be used. In most
cases, the system will add or change the infrastructure that already exists in the
organization. Th e interface design specifi es how the users will move through the sys-
tem (e.g., navigation methods such as menus and on-screen buttons) and the forms
and reports that the system will use.
3. Th e database and fi le specifi cations are developed. Th ese defi ne exactly what data
will be stored and where they will be stored.
4. Th e analyst team develops the program design, which defi nes the programs that
need to be written and exactly what each program will do.
Th is collection of deliverables (architecture design, interface design, database and fi le specifi ca-
tions, and program design) is the system specifi cation that is handed to the programming team
for implementation. At the end of the design phase, the feasibility analysis and project plan are
reexamined and revised, and another decision is made by the project sponsor and approval
committee about whether to terminate the project or continue.
Implementation
Th e fi nal phase in the SDLC is the implementation phase, during which the system is actually
built (or purchased, in the case of a packaged soft ware design). Th is is the phase that usually
Systems Development Methodologies 5
2 Th e classic modern process-centered methodology is that by Edward Yourdon, Modern Structured Analysis
(Englewood Cliff s, NJ: Yourdon Press, 1989). An example of a data-centered methodology is information engi-
neering; see James Martin, Information Engineering, vols. 1–3 (Englewood Cliff s, NJ: Prentice Hall, 1989). A widely
accepted standardized non–object-oriented methodology that balances processes and data is IDEF; see FIPS 183,
Integration Defi nition for Function Modeling, Federal Information Processing Standards Publications, U.S. Depart-
ment of Commerce, 1993.
3 A good reference for comparing systems development methodologies is Steve McConnell, Rapid Development
(Redmond, WA: Microsoft Press, 1996).
gets the most attention, because for most systems it is the longest and most expensive single
part of the development process. Th is phase has three steps:
1. System construction is the fi rst step. Th e system is built and tested to ensure that it
performs as designed. Because the cost of bugs can be immense, testing is one of
the most critical steps in implementation. Most organizations give more time and
attention to testing than to writing the programs in the fi rst place.
2. Th e system is installed. Installation is the process by which the old system is turned
off and the new one is turned on. One of the most important aspects of conversion
is the development of a training plan to teach users how to use the new system and
help manage the changes caused by the new system.
3. Th e analyst team establishes a support plan for the system. Th is plan usually
includes a formal or informal post-implementation review as well as a systematic
way for identifying major and minor changes needed for the system.
SYSTEMS DEVELOPMENT METHODOLOGIES
A methodology is a formalized approach to implementing the SDLC (i.e., it is a list of steps
and deliverables). Th ere are many diff erent systems development methodologies, and each
one is unique, based on the order and focus it places on each SDLC phase. Some methodolo-
gies are formal standards used by government agencies, whereas others have been developed
by consulting fi rms to sell to clients. Many organizations have internal methodologies that
have been honed over the years, and they explain exactly how each phase of the SDLC is to
be performed in that company.
Th ere are many ways to categorize methodologies. One way is by looking at whether
they focus on business processes or the data that support the business. A process-centered
methodology emphasizes process models as the core of the system concept. In Figure 1-1, for
example, process-centered methodologies would focus fi rst on defi ning the processes (e.g.,
assemble sandwich ingredients). Data-centered methodologies emphasize data models as the
core of the system concept. In Figure 1-1, data-centered methodologies would focus fi rst on
defi ning the contents of the storage areas (e.g., refrigerator) and how the contents were organ-
ized.2 By contrast, object-oriented methodologies attempt to balance the focus between process
and data by incorporating both into one model. In Figure 1-1, these methodologies would
focus fi rst on defi ning the major elements of the system (e.g., sandwiches, lunches) and look
at the processes and data involved with each element.
Another important factor in categorizing methodologies is the sequencing of the SDLC
phases and the amount of time and eff ort devoted to each.3 In the early days of computing,
programmers did not understand the need for formal and well-planned life-cycle meth-
odologies. Th ey tended to move directly from a very simple planning phase right into the
construction step of the implementation phase—in other words, from a very fuzzy, not-well-
thought-out system request into writing code. Th is is the same approach that you sometimes
use when writing programs for a programming class. It can work for small programs that
6 Chapter 1 Introduction to Systems Analysis and Design
require only one programmer, but if the requirements are complex or unclear, you might
miss important aspects of the problem and have to start all over again, throwing away part of
the program (and the time and eff ort spent writing it). Th is approach also makes teamwork
diffi cult because members have little idea about what needs to be accomplished and how to
work together to produce a fi nal product. In this section, we describe three diff erent classes of
system development methodologies: structured design, rapid application development, and
agile development.
Structured Design
Th e fi rst category of systems development methodologies is called structured design.
Th ese methodologies became dominant in the 1980s, replacing the previous ad hoc and
FIGURE 1-1 A Simple Behavioral Model for Making a Simple Lunch
GetJelly
GetPeanutButter
GetCookies
GetBread
CreateSandwich
GetMilk
CreateLunch
GetLunchBag
PutLunchInBag
aParent aRefrigerator aCupboard aSandwich aLunch aLunchBag
Systems Development Methodologies 7
undisciplined approach. Structured design methodologies adopt a formal step-by-step
approach to the SDLC that moves logically from one phase to the next. Numerous pro-
cess-centered and data-centered methodologies follow the basic approach of the two struc-
tured design categories outlined next.
Waterfall Development Th e original structured design methodology (still used today) is
waterfall development. With waterfall development-based methodologies, the analysts and
users proceed in sequence from one phase to the next (see Figure 1-2). Th e key deliverables
for each phase are typically very long (oft en hundreds of pages in length) and are presented to
the project sponsor for approval as the project moves from phase to phase. Once the sponsor
approves the work that was conducted for a phase, the phase ends and the next one begins.
Th is methodology is referred to as waterfall development because it moves forward from
phase to phase in the same manner as a waterfall. Although it is possible to go backward in
the SDLC (e.g., from design back to analysis), it is extremely diffi cult (imagine yourself as a
salmon trying to swim upstream against a waterfall, as shown in Figure 1-2).
Structured design also introduced the use of formal modeling or diagramming tech-
niques to describe the basic business processes and the data that support them. Traditional
structured design uses one set of diagrams to represent the processes and a separate set of
diagrams to represent data. Because two sets of diagrams are used, the systems analyst must
decide which set to develop fi rst and use as the core of the system: process-model diagrams
or data-model diagrams.
Th e two key advantages of the structured design waterfall approach are that it identi-
fi es system requirements long before programming begins and it minimizes changes to the
requirements as the project proceeds. Th e two key disadvantages are that the design must
be completely specifi ed before programming begins and that a long time elapses between the
completion of the system proposal in the analysis phase and the delivery of the system (usu-
ally many months or years). If the project team misses important requirements, expensive
post-implementation programming may be needed (imagine yourself trying to design a car
on paper; how likely would you be to remember interior lights that come on when the doors
open or to specify the right number of valves on the engine?). A system can also require
signifi cant rework because the business environment has changed from the time when the
analysis phase occurred.
FIGURE 1-2
A Waterfall
Development-Based
Methodology
System
Planning
Analysis
Design
Implementation
8 Chapter 1 Introduction to Systems Analysis and Design
Parallel Development Parallel development methodology attempts to address the problem
of long delays between the analysis phase and the delivery of the system. Instead of doing
design and implementation in sequence, it performs a general design for the whole system
and then divides the project into a series of distinct subprojects that can be designed and
implemented in parallel. Once all subprojects are complete, the separate pieces are integrated
and the system is delivered (see Figure 1-3).
Th e primary advantage of this methodology is that it can reduce the time to deliver a
system; thus, there is less chance of changes in the business environment causing rework.
However, sometimes the subprojects are not completely independent; design decisions
made in one subproject can aff ect another, and the end of the project can require signifi cant
integration eff orts.
Rapid Application Development (RAD)
A second category of methodologies includes rapid application development (RAD)-based
methodologies. Th ese are a newer class of systems development methodologies that emerged
in the 1990s. RAD-based methodologies attempt to address both weaknesses of structured
design methodologies by adjusting the SDLC phases to get some part of the system devel-
oped quickly and into the hands of the users. In this way, the users can better understand the
system and suggest revisions that bring the system closer to what is needed.4
4 One of the best RAD books is Steve McConnell, Rapid Development (Redmond, WA: Microsoft Press, 1996).
FIGURE 1-3 A Parallel Development-Based Methodology
System
Planning
Analysis
Design
Implementation
Design
Integration
Implementation
Design
Implementation
Design
Subproject 2
Subproject 1
Subproject 3
Systems Development Methodologies 9
Most RAD-based methodologies recommend that analysts use special techniques
and computer tools to speed up the analysis, design, and implementation phases, such as
computer-aided soft ware engineering (CASE) tools, joint application design (JAD) sessions,
fourth-generation or visual programming languages that simplify and speed up programming,
and code generators that automatically produce programs from design specifi cations. Th e
combination of the changed SDLC phases and the use of these tools and techniques improves
the speed and quality of systems development. However, there is one possible subtle problem
with RAD-based methodologies: managing user expectations. Owing to the use of the tools and
techniques that can improve the speed and quality of systems development, user expectations
of what is possible can change dramatically. As a user better understands the information tech-
nology (IT), the systems requirements tend to expand. Th is was less of a problem when using
methodologies that spent a lot of time thoroughly documenting requirements.
Phased Development A phased development-based methodology breaks an overall system into a
series of versions that are developed sequentially. Th e analysis phase identifi es the overall system
concept, and the project team, users, and system sponsor then categorize the requirements into
a series of versions. Th e most important and fundamental requirements are bundled into the
fi rst version of the system. Th e analysis phase then leads into design and implementation—but
only with the set of requirements identifi ed for version 1 (see Figure 1-4).
Once version 1 is implemented, work begins on version 2. Additional analysis is per-
formed based on the previously identifi ed requirements and combined with new ideas and
issues that arose from the users’ experience with version 1. Version 2 then is designed and
implemented, and work immediately begins on the next version. Th is process continues until
the system is complete or is no longer in use.
Phased development-based methodologies have the advantage of quickly getting a useful
system into the hands of the users. Although the system does not perform all the functions the
users need at fi rst, it does begin to provide business value sooner than if the system were deliv-
ered aft er completion, as is the case with the waterfall and parallel methodologies. Likewise,
because users begin to work with the system sooner, they are more likely to identify important
additional requirements sooner than with structured design situations.
Th e major drawback to phased development is that users begin to work with systems that
are intentionally incomplete. It is critical to identify the most important and useful features
and include them in the fi rst version and to manage users’ expectations along the way.
Prototyping A prototyping-based methodology performs the analysis, design, and imple-
mentation phases concurrently, and all three phases are performed repeatedly in a cycle until
the system is completed. With these methodologies, the basics of analysis and design are
performed, and work immediately begins on a system prototype, a quick-and-dirty program
that provides a minimal amount of features. Th e fi rst prototype is usually the fi rst part of the
system that is used. Th is is shown to the users and the project sponsor, who provide com-
ments. Th ese comments are used to reanalyze, redesign, and reimplement a second prototype,
which provides a few more features. Th is process continues in a cycle until the analysts, users,
and sponsor agree that the prototype provides enough functionality to be installed and used in
the organization. Aft er the prototype (now called the “system”) is installed, refi nement occurs
until it is accepted as the new system (see Figure 1-5).
Th e key advantage of a prototyping-based methodology is that it very quickly provides a
system with which the users can interact, even if it is not ready for widespread organizational
use at fi rst. Prototyping reassures the users that the project team is working on the system
(there are no long delays in which the users see little progress), and prototyping helps to more
quickly refi ne real requirements.
10 Chapter 1 Introduction to Systems Analysis and Design
FIGURE 1-4 A Phased Development-Based Methodology
System
version 1
Planning
Analysis
Analysis
Implementation
Design
Analysis
Implementation
Design
Analysis
Implementation
Design
System
version 2
System
version 3
FIGURE 1-5
A Prototyping-Based
Methodology
System
prototype
System
Planning
Analysis
Design
Implementation
Implementation
Systems Development Methodologies 11
Th e major problem with prototyping is that its fast-paced system releases challenge
attempts to conduct careful, methodical analysis. Oft en the prototype undergoes such signif-
icant changes that many initial design decisions become poor ones. Th is can cause problems
in the development of complex systems because fundamental issues and problems are not rec-
ognized until well into the development process. Imagine building a car and discovering late in
the prototyping process that you have to take the whole engine out to change the oil (because
no one thought about the need to change the oil until aft er it had been driven 10,000 miles).
Throwaway Prototyping Th rowaway prototyping-based methodologies are similar to
prototyping-based methodologies in that they include the development of prototypes; how-
ever, throwaway prototypes are done at a diff erent point in the SDLC. Th ese prototypes are
used for a very diff erent purpose than those previously discussed, and they have a very diff er-
ent appearance (see Figure 1-6).
Th e throwaway prototyping-based methodologies have a relatively thorough analy-
sis phase that is used to gather information and to develop ideas for the system concept.
However, users might not completely understand many of the features they suggest, and there
may be challenging technical issues to be solved. Each of these issues is examined by analyz-
ing, designing, and building a design prototype. A design prototype is not a working system;
it is a product that represents a part of the system that needs additional refi nement, and it
contains only enough detail to enable users to understand the issues under consideration. For
example, suppose users are not completely clear on how an order-entry system should work.
In this case, a series of mock-up screens appear to be a system, but they really do nothing. Or
suppose that the project team needs to develop a sophisticated graphics program in Java. Th e
team could write a portion of the program with pretend data to ensure that they could do a
full-blown program successfully.
A system developed using this type of methodology relies on several design prototypes
during the analysis and design phases. Each of the prototypes is used to minimize the risk
associated with the system by confi rming that important issues are understood before the real
system is built. Once the issues are resolved, the project moves into design and implementa-
tion. At this point, the design prototypes are thrown away, which is an important diff erence
between these methodologies and prototyping methodologies, in which the prototypes evolve
into the fi nal system.
FIGURE 1-6 A Throwaway Prototyping-Based Methodology
Design
prototype
System
Analysis
Analysis
Design
Implementation
Planning
Implementation
Design
12 Chapter 1 Introduction to Systems Analysis and Design
Th rowaway prototyping-based methodologies balance the benefi ts of well-thought-out
analysis and design phases with the advantages of using prototypes to refi ne key issues before
a system is built. It can take longer to deliver the fi nal system as compared to prototyping-
based methodologies, but this type of methodology usually produces more stable and reliable
systems.
Agile Development5
A third category of systems development methodologies is still emerging today: agile devel-
opment. All agile development methodologies are based on the agile manifesto and a set of
twelve principles. Th e emphasis of the manifesto is to focus the developers on the working
conditions of the developers, the working soft ware, the customers, and addressing changing
requirements instead of focusing on detailed systems development processes, tools, all-
inclusive documentation, legal contracts, and detailed plans. Th ese programming-centric
methodologies have few rules and practices, all of which are fairly easy to follow. Th ese meth-
odologies are typically based only on the twelve principles of agile soft ware. Th ese principles
include the following:
■ Soft ware is delivered early and continuously through the development process, satis-
fying the customer.
■ Changing requirements are embraced regardless of when they occur in the develop-
ment process.
■ Working soft ware is delivered frequently to the customer.
■ Customers and developers work together to solve the business problem.
■ Motivated individuals create solutions; provide them the tools and environment they
need, and trust them to deliver.
■ Face-to-face communication within the development team is the most effi cient and
eff ective method of gathering requirements.
■ Th e primary measure of progress is working, executing soft ware.
■ Both customers and developers should work at a pace that is sustainable. Th at is, the
level of work could be maintained indefi nitely without any worker burnout.
■ Agility is heightened through attention to both technical excellence and good design.
■ Simplicity, the avoidance of unnecessary work, is essential.
■ Self-organizing teams develop the best architectures, requirements, and designs.
■ Development teams regularly refl ect on how to improve their development
processes.
Based on these principles, agile methodologies focus on streamlining the system-development
process by eliminating much of the modeling and documentation overhead and the time
spent on those tasks. Instead, projects emphasize simple, iterative application development.6
All agile development methodologies follow a simple cycle through the traditional phases of
the systems development process (see Figure 1-7). Virtually all agile methodologies are used
in conjunction with object-oriented technologies.
5 Th ree good sources of information on agile development and object-oriented systems are S. W. Ambler, Agile
Modeling: Eff ective Practices for Extreme Programming and the Unifi ed Process (New York: Wiley, 2002); C. Larman,
Agile & Iterative Development: A Manager’s Guide (Boston: Addison-Wesley, 2004); R. C. Martin, Agile Soft ware
Development: Principles, Patterns, and Practices (Upper Saddle River, NJ: Prentice Hall, 2003).
6 See Agile Alliance, www.agilealliance.com.
Systems Development Methodologies 13
However, agile methodologies do have critics. One of the major criticisms deals with
today’s business environment, where much of the actual information systems development
is off shored, outsourced, and/or subcontracted. Given agile development methodologies
requiring co-location of the development team, this seems to be a very unrealistic assump-
tion. A second major criticism is that if agile development is not carefully managed, and by
defi nition it is not, the development process can devolve into a prototyping approach that
essentially becomes a “programmers gone wild” environment where programmers attempt
to hack together solutions. A third major criticism, based on the lack of actual documen-
tation created during the development of the soft ware, raises issues regarding the auditability
of the systems being created. Without suffi cient documentation, neither the system nor the
systems-development process can be assured. A fourth major criticism is based on whether
agile approaches can deliver large mission-critical systems.
Even with these criticisms, given the potential for agile approaches to address the
application backlog and to provide timely solutions to many business problems, agile
approaches should be considered in some circumstances. Furthermore, many of the tech-
niques encouraged by attending to the underlying purpose of the agile manifesto and the
set of twelve agile principles are very useful in object-oriented systems development. Two of
the more popular examples of agile development methodologies are extreme programming
(XP) and Scrum.
Extreme Programming7 Extreme programming (XP) is founded on four core values: com-
munication, simplicity, feedback, and courage. Th ese four values provide a foundation that
XP developers use to create any system. First, the developers must provide rapid feedback
to the end users on a continuous basis. Second, XP requires developers to follow the KISS
principle.8 Th ird, developers must make incremental changes to grow the system, and they
must not only accept change, they must embrace change. Fourth, developers must have a
quality-fi rst mentality. XP also supports team members in developing their own skills. Th ree
of the key principles that XP uses to create successful systems are continuous testing, simple
coding performed by pairs of developers, and close interactions with end users to build sys-
tems very quickly.
7 For more information, see K. Beck, eXtreme Programming Explained: Embrace Change (Reading, MA: Addison-
Wesley, 2000); C. Larman, Agile & Iterative Development: A Manager’s Guide (Boston: Addison-Wesley, 2004); M.
Lippert, S. Roock, and H. Wolf, eXtreme Programming in Action: Practical Experiences from Real World Projects (New
York: Wiley, 2002); www.extremeprogramming.org.
8 Keep it simple, stupid.
FIGURE 1-7
Typical Agile
Development
Methodology
Implementation
Design
Analysis
System
Planning
14 Chapter 1 Introduction to Systems Analysis and Design
Testing and effi cient coding practices are the core of XP. Code is tested each day and is
placed into an integrative testing environment. If bugs exist, the code is backed out until it
is completely free of errors.
An XP project begins with user stories that describe what the system needs to do. Th en,
programmers code in small, simple modules and test to meet those needs. Users are required
to be available to clear up questions and issues as they arise. Standards are very important
to minimize confusion, so XP teams use a common set of names, descriptions, and coding
practices. XP projects deliver results sooner than even the RAD approaches, and they rarely
get bogged down in gathering requirements for the system.
XP adherents claim many strengths associated with developing soft ware using XP.
Programmers work closely with all stakeholders, and communication among all stakehold-
ers is improved. Continuous testing of the evolving system is encouraged. Th e system is
developed in an evolutionary and incremental manner, which allows the requirements to
evolve as the stakeholders understand the potential that the technology has in providing a
solution to their problem. Estimation is task driven and is performed by the programmer
who will implement the solution for the task under consideration. Because all programming
is done in pairs, a shared responsibility for each soft ware component develops among the
programmers. Finally, the quality of the fi nal product increases during each iteration.
For small projects with highly motivated, cohesive, stable, and experienced teams, XP
should work just fi ne. However, if the project is not small or the teams aren’t jelled,9 the suc-
cess of an XP development eff ort is doubtful. Th is tends to throw into doubt the whole idea
of bringing outside contractors into an existing team environment using XP.10 Th e chance
of outsiders jelling with insiders might simply be too optimistic. XP requires a great deal of
discipline, otherwise projects will become unfocused and chaotic. XP is recommended only
for small groups of developers—no more than ten developers—and it is not advised for large
mission-critical applications. Owing to the lack of analysis and design documentation, there
is only code documentation associated with XP, so maintaining large systems built with XP
may be impossible. And because mission-critical business information systems tend to exist
for a long time, the utility of XP as a business information system development methodology
is in doubt. Finally, the methodology needs a lot of on-site user input, something to which
many business units cannot commit.11 However, some of the techniques associated with
XP are useful in object-oriented systems development. For example, user stories, pair pro-
gramming, and continuous testing are invaluable tools from which object-oriented systems
development could benefi t.
Scrum12 Scrum is a term that is well known to rugby fans. In rugby, a scrum is used to
restart a game. In a nutshell, the creators of the Scrum method believe that no matter how
much you plan, as soon as the soft ware begins to be developed, chaos breaks out and the
9 A jelled team is one that has low turnover, a strong sense of identity, a sense of eliteness, a feeling that they jointly
own the product being developed, and enjoyment in working together. For more information regarding jelled teams,
see T. DeMarco and T. Lister, Peopleware: Productive Projects and Teams (New York: Dorset/House, 1987).
10 Considering the tendency for off shore outsourcing, this is a major obstacle for XP to overcome. For more infor-
mation on off shore outsourcing, see P. Th ibodeau, “ITAA Panel Debates Outsourcing Pros, Cons,” Computerworld
Morning Update (September 25, 2003); S. W. Ambler, “Chicken Little Was Right,” Soft ware Development (October
2003).
11 Many of the observations on the utility of XP as a development approach were based on conversations with Brian
Henderson-Sellers.
12 For more information, see C. Larman, Agile & Iterative Development: A Manager’s Guide (Boston: Addison-
Wesley, 2004); K. Schwaber and M. Beedle, Agile Soft ware Development with Scrum (Upper Saddle River, NJ:
Prentice Hall, 2001); R. Wysocki, Eff ective Project Management: Traditional, Agile, Extreme, 5th Ed. (Indianapolis,
IN: Wiley Publishing, 2009).
Systems Development Methodologies 15
plans go out the window.13 Th e best you can do is to react to where the proverbial rugby
ball squirts out. You then sprint with the ball until the next scrum. In the case of the Scrum
methodology, a sprint lasts thirty working days. At the end of the sprint, a system is deliv-
ered to the customer.
Of all systems development approaches, on the surface, Scrum is the most chaotic. To
control some of the innate chaos, Scrum development focuses on a few key practices. Teams
are self-organized and self-directed. Unlike other approaches, Scrum teams do not have a des-
ignated team leader. Instead, teams organize themselves in a symbiotic manner and set their
own goals for each sprint (iteration). Once a sprint has begun, Scrum teams do not consider
any additional requirements. Any new requirements that are uncovered are placed on a back-
log of requirements that still need to be addressed. At the beginning of every workday, a Scrum
meeting takes place. At the end of each sprint, the team demonstrates the soft ware to the client.
Based on the results of the sprint, a new plan is begun for the next sprint.
Scrum meetings are one of the most interesting aspects of the Scrum development pro-
cess. Th e team members attend the meetings, but anyone can attend. However, with very
few exceptions, only team members may speak. One prominent exception is management
providing feedback on the business relevance of the work being performed by the specifi c
team. In this meeting, all team members stand in a circle and report on what they accom-
plished during the previous day, state what they plan to do today, and describe anything
that blocked progress the previous day. To enable continuous progress, any block identifi ed
is dealt with within one hour. From a Scrum point of view, it is better to make a “bad” deci-
sion about a block at this point in development than to not make a decision. Because the
meetings take place each day, a bad decision can easily be undone. Larman14 suggests that
each team member should report any additional requirements that have been uncovered
during the sprint and anything that the team member learned that could be useful for other
team members to know.
One of the major criticisms of Scrum, as with all agile methodologies, is that it is ques-
tionable whether Scrum can scale up to develop very large, mission-critical systems. A typical
Scrum team size is no more than seven members. Th e only organizing principle put forth by
Scrum followers to address this criticism is to organize a scrum of scrums. Each team meets
every day, and aft er the team meeting takes place, a representative (not leader) of each team
attends a scrum-of-scrums meeting. Th is continues until the progress of entire system has
been determined. Depending on the number of teams involved, this approach to managing a
large project is doubtful. However, as in XP and other agile development approaches, many
of the ideas and techniques associated with Scrum development are useful in object-oriented
systems development, such as the focus of a Scrum meeting, the evolutionary and incremen-
tal approach to identifying requirements, and the incremental and iterative approach to the
development of the system.
Selecting the Appropriate Development Methodology
Because there are many methodologies, the fi rst challenge faced by analysts is selecting which
methodology to use. Choosing a methodology is not simple, because no one methodology is
always best. (If it were, we’d simply use it everywhere!) Many organizations have standards
and policies to guide the choice of methodology. You will fi nd that organizations range from
13 Scrum developers are not the fi rst to question the use of plans. One of President Eisenhower’s favorite maxims
was, “In preparing for battle I have always found that plans are useless, but planning is indispensable.” M. Dobson,
Streetwise Project Management: How to Manage People, Processes, and Time to Achieve the Results You Need (Avon,
MA: F+W Publications, 2003), p. 43.
14 C. Larman, Agile & Iterative Development: A Manager’s Guide (Boston: Addison-Wesley, 2004).
16 Chapter 1 Introduction to Systems Analysis and Design
having one “approved” methodology to having several methodology options to having no
formal policies at all.
Figure 1-8 summarizes some important criteria for selecting a methodology. One impor-
tant item not discussed in this fi gure is the degree of experience of the analyst team. Many
of the RAD-based methodologies require the use of new tools and techniques that have a
signifi cant learning curve. Oft en these tools and techniques increase the complexity of the
project and require extra time for learning. However, once they are adopted and the team
becomes experienced, the tools and techniques can signifi cantly increase the speed at which
the methodology can deliver a fi nal system.
Clarity of User Requirements When the user requirements for a system are unclear, it is
diffi cult to understand them by talking about them and explaining them with written reports.
Users normally need to interact with technology to really understand what a new system can
do and how to best apply it to their needs. RAD and agile methodologies are usually more
appropriate when user requirements are unclear.
Familiarity with Technology When the system will use new technology with which the ana-
lysts and programmers are not familiar, early application of the new technology in the meth-
odology will improve the chance of success. If the system is designed without some familiarity
with the base technology, risks increase because the tools might not be capable of doing what
is needed. Th rowaway prototyping-based methodologies are particularly appropriate if users
lack familiarity with technology because they explicitly encourage the developers to develop
design prototypes for areas with high risks. Phased development-based methodologies create
opportunities to investigate the technology in some depth before the design is complete. Also,
owing to the programming-centric nature of agile methodologies, both XP and Scrum are
appropriate. Although you might think prototyping-based methodologies are also appropriate,
they are much less so because the early prototypes that are built usually only scratch the surface
of the new technology. It is generally only aft er several prototypes and several months that the
developers discover weaknesses or problems in the new technology.
System Complexity Complex systems require careful and detailed analysis and design.
Th rowaway prototyping-based methodologies are particularly well suited to such detailed
analysis and design, but prototyping-based methodologies are not. Th e traditional structured
Ability to Develop
Systems
Structured
Methodologies RAD Methodologies
Agile
Methodologies
Waterfall Parallel Phased Prototyping
Throwaway
Prototyping XP SCRUM
With Unclear User Requirements Poor Poor Good Excellent Excellent Excellent Excellent
With Unfamiliar Technology Poor Poor Good Poor Excellent Good Good
That Are Complex Good Good Good Poor Excellent Good Good
That Are Reliable Good Good Good Poor Excellent Excellent Excellent
With a Short Time Schedule Poor Good Excellent Excellent Good Excellent Excellent
With Schedule Visibility Poor Poor Excellent Excellent Good Excellent Excellent
FIGURE 1-8 Criteria for Selecting a Methodology
Typical Systems Analyst Roles and Skills 17
design-based methodologies can handle complex systems, but without the ability to get the
system or prototypes into the users’ hands early on, some key issues may be overlooked.
Although phased development-based methodologies enable users to interact with the system
early in the process, we have observed that project teams who follow these tend to devote less
attention to the analysis of the complete problem domain than they might using other meth-
odologies. Finally, agile methodologies are a mixed bag when it comes to system complexity.
If the system is going to be a large one, agile methodologies will perform poorly. However,
if the system is small to medium size, then agile approaches will be excellent. We rate them
good on these criteria.
System Reliability System reliability is usually an important factor in system development;
aft er all, who wants an unreliable system? However, reliability is just one factor among
several. For some applications, reliability is truly critical (e.g., medical equipment, mis-
sile-control systems), whereas for other applications (e.g., games, Internet video) it is merely
important. Because throwaway prototyping methodologies combine detailed analysis and
design phases with the ability for the project team to test many diff erent approaches through
design prototypes before completing the design, they are appropriate when system reliability
is a high priority. Prototyping methodologies are generally not a good choice when reliability
is critical because it lacks the careful analysis and design phases that are essential for depend-
able systems. However, owing to the heavy focus on testing, evolutionary and incremental
identifi cation of requirements, and iterative and incremental development, agile methods
may be the best overall approach.
Short Time Schedules RAD-based and agile methodologies are excellent choices when
timelines are short because they best enable the project team to adjust the functionality in
the system based on a specifi c delivery date, and if the project schedule starts to slip, it can
be readjusted by removing functionality from the version or prototype under development.
Waterfall-based methodologies are the worst choice when time is at a premium because they
do not allow easy schedule changes.
Schedule Visibility One of the greatest challenges in systems development is determining
whether a project is on schedule. Th is is particularly true of the structured design method-
ologies because design and implementation occur at the end of the project. Th e RAD-based
methodologies move many of the critical design decisions earlier in the project to help project
managers recognize and address risk factors and keep expectations in check. However, given
the daily progress meetings associated with Agile approaches, schedule visibility is always on
the proverbial front burner.
TYPICAL SYSTEMS ANALYST ROLES AND SKILLS
It is clear from the various phases and steps performed during the SDLC that the project team
needs a variety of skills. Project members are change agents who identify ways to improve an
organization, build an information system to support them, and train and motivate others to
use the system. Understanding what to change and how to change it—and convincing others
of the need for change—requires a wide range of skills. Th ese skills can be broken down into
six major categories: technical, business, analytical, interpersonal, management, and ethical.
Analysts must have the technical skills to understand the organization’s existing techni-
cal environment, the technology that will make up the new system, and the way both can fi t
into an integrated technical solution. Business skills are required to understand how IT can be
18 Chapter 1 Introduction to Systems Analysis and Design
applied to business situations and to ensure that the IT delivers real business value. Analysts
are continuous problem solvers at both the project and the organizational level, and they put
their analytical skills to the test regularly.
Analysts oft en need to communicate eff ectively one-on-one with users and business man-
agers (who oft en have little experience with technology) and with programmers (who oft en have
more technical expertise than the analyst). Th ey must be able to give presentations to large and
small groups and write reports. Not only do they need to have strong interpersonal abilities, but
they also need to manage people with whom they work and they need to manage the pressure
and risks associated with unclear situations.
Finally, analysts must deal fairly, honestly, and ethically with other project team mem-
bers, managers, and system users. Analysts oft en deal with confi dential information or infor-
mation that, if shared with others, could cause harm (e.g., dissent among employees); it is
important to maintain confi dence and trust with all people.
In addition to these six general skill sets, analysts require many specifi c skills associated
with roles performed on a project. In the early days of systems development, most organiza-
tions expected one person, the analyst, to have all the specifi c skills needed to conduct a sys-
tems development project. Some small organizations still expect one person to perform many
roles, but because organizations and technology have become more complex, most large
organizations now build project teams containing several individuals with clearly defi ned
responsibilities. Diff erent organizations divide the roles diff erently. Most IS teams include
many other individuals, such as the programmers, who actually write the programs that make
up the system, and technical writers, who prepare the help screens and other documentation
(e.g., users manuals and systems manuals).
Business Analyst
A business analyst focuses on the business issues surrounding the system. Th ese issues include
identifying the business value that the system will create, developing ideas and suggestions for
how the business processes can be improved, and designing the new processes and policies in
conjunction with the systems analyst. Th is individual likely has business experience and some
type of professional training. He or she represents the interests of the project sponsor and the
ultimate users of the system. A business analyst assists in the planning and design phases but is
most active in the analysis phase.
Systems Analyst
A systems analyst focuses on the IS issues surrounding the system. Th is person develops ideas
and suggestions for how information technology can improve business processes, designs the
new business processes with help from the business analyst, designs the new information sys-
tem, and ensures that all IS standards are maintained. A systems analyst likely has signifi cant
training and experience in analysis and design, programming, and even areas of the business.
He or she represents the interests of the IS department and works intensively through the pro-
ject but perhaps less so during the implementation phase.
Infrastructure Analyst
An infrastructure analyst focuses on the technical issues surrounding how the system will
interact with the organization’s technical infrastructure (e.g., hardware, soft ware, networks,
and databases). An infrastructure analyst’s tasks include ensuring that the new information
system conforms to organizational standards and identifying infrastructure changes needed
to support the system. Th is individual probably has signifi cant training and experience in
Basic Characteristics of Object-Oriented Systems 19
networking, database administration, and various hardware and soft ware products. He or
she represents the interests of the organization and IS group that will ultimately have to
operate and support the new system once it has been installed. An infrastructure analyst
works throughout the project but perhaps less so during planning and analysis phases.
Change Management Analyst
A change management analyst focuses on the people and management issues surrounding
the system installation. Th e roles of this person include ensuring that the adequate docu-
mentation and support are available to users, providing user training on the new system, and
developing strategies to overcome resistance to change. Th is individual should have signifi –
cant training and experience in organizational behavior in general and change management
in particular. He or she represents the interests of the project sponsor and users for whom
the system is being designed. A change management analyst works most actively during the
implementation phase but begins laying the groundwork for change during the analysis and
design phases.
Project Manager
A project manager is responsible for ensuring that the project is completed on time and within
budget and that the system delivers all benefi ts intended by the project sponsor. Th e role
of the project manager includes managing the team members, developing the project plan,
assigning resources, and being the primary point of contact when people outside the team
have questions about the project. Th is individual likely has signifi cant experience in project
management and has probably worked for many years as a systems analyst beforehand. He
or she represents the interests of the IS department and the project sponsor. Th e project man-
ager works intensely during all phases of the project.
BASIC CHARACTERISTICS OF OBJECT-ORIENTED SYSTEMS
Object-oriented systems focus on capturing the structure and behavior of information sys-
tems in little modules that encompass both data and process. Th ese little modules are known
as objects. In this section, we describe the basic characteristics of object-oriented systems,
which include classes, objects, methods, messages, encapsulation, information hiding, inher-
itance, polymorphism, and dynamic binding.15
Classes and Objects
A class is the general template we use to defi ne and create specifi c instances, or objects. Every
object is associated with a class. For example, all the objects that capture information about
patients could fall into a class called Patient, because there are attributes (e.g., name, address,
birth date, phone, and insurance carrier) and methods (e.g., make appointment, calculate last
visit, change status, and provide medical history) that all patients share (see Figure 1-9).
An object is an instantiation of a class. In other words, an object is a person, place, or
thing about which we want to capture information. If we were building an appointment sys-
tem for a doctor’s offi ce, classes might include Doctor, Patient, and Appointment. Th e specifi c
patients, such as Jim Maloney, Mary Wilson, and Th eresa Marks, are considered instances, or
objects, of the patient class (see Figure 1-9).
15 In Chapter 8, we review the basic characteristics of object-oriented systems in more detail.
20 Chapter 1 Introduction to Systems Analysis and Design
Each object has attributes that describe information about the object, such as a patient’s
name, birth date, address, and phone number. Attributes are also used to represent relation-
ships between objects; for example, there could be a department attribute in an employee
object with a value of a department object that captures in which department the employee
object works. Th e state of an object is defi ned by the value of its attributes and its relationships
with other objects at a particular point in time. For example, a patient might have a state of
new or current or former.
Each object also has behaviors. Th e behaviors specify what the object can do. For exam-
ple, an appointment object can probably schedule a new appointment, delete an appointment,
and locate the next available appointment. In object-oriented programming, behaviors are
implemented as methods (see the next section).
One of the more confusing aspects of object-oriented systems development is the fact
that in most object-oriented programming languages, both classes and instances of classes
can have attributes and methods. Class attributes and methods tend to be used to model
attributes (or methods) that deal with issues related to all instances of the class. For example,
to create a new patient object, a message is sent to the Patient class to create a new instance
of itself. However, in this book, we focus primarily on attributes and methods of objects and
not of classes.
Methods and Messages
Methods implement an object’s behavior. A method is nothing more than an action that an
object can perform. Messages are information sent to objects to trigger methods. A message
is essentially a function or procedure call from one object to another object. For example, if a
patient is new to the doctor’s offi ce, the receptionist sends a create message to the application.
Th e patient class receives the create message and executes its create() method which then
creates a new object: aPatient (see Figure 1-10).
Encapsulation and Information Hiding
Th e ideas of encapsulation and information hiding are interrelated in object-oriented systems.
However, neither of the terms is new. Encapsulation is simply the combination of process
and data into a single entity. Information hiding was fi rst promoted in structured systems
development. Th e principle of information hiding suggests that only the information
FIGURE 1-9
Classes and Objects
Patient
-name
-address
-birthdate
-phone
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
+create()
Mary Wilson : PatientJim Maloney : Patient Theresa Marks : Patient
Basic Characteristics of Object-Oriented Systems 21
required to use a soft ware module be published to the user of the module. Typically, this
implies that the information required to be passed to the module and the information
returned from the module are published. Exactly how the module implements the required
functionality is not relevant. We really do not care how the object performs its functions,
as long as the functions occur. In object-oriented systems, combining encapsulation with the
information-hiding principle supports treating objects as black boxes.
Th e fact that we can use an object by calling methods is the key to reusability because it
shields the internal workings of the object from changes in the outside system, and it keeps
the system from being aff ected when changes are made to an object. In Figure 1-10, notice
how a message (create) is sent to an object, yet the internal algorithms needed to respond to
the message are hidden from other parts of the system. Th e only information that an object
needs to know is the set of operations, or methods, that other objects can perform and what
messages need to be sent to trigger them.
Inheritance
Inheritance, as an information systems development characteristic, was proposed in data
modeling in the late 1970s and the early 1980s. Th e data modeling literature suggests using
inheritance to identify higher-level, or more general, classes of objects. Common sets of
attributes and methods can be organized into superclasses. Typically, classes are arranged in
a hierarchy whereby the superclasses, or general classes, are at the top and the subclasses, or
specifi c classes, are at the bottom. In Figure 1-11, Person is a superclass to the classes Doctor
and Patient. Doctor, in turn, is a superclass to General Practitioner and Specialist. Notice how
a class (e.g., Doctor) can serve as a superclass and subclass concurrently. Th e relationship
between the class and its superclass is known as the a-kind-of relationship. For example in
Figure 1-11, a General Practitioner is a-kind-of Doctor, which is a-kind-of Person.
Subclasses inherit the appropriate attributes and methods from the superclasses above
them. Th at is, each subclass contains attributes and methods from its parent superclass. For
example, Figure 1-11 shows that both Doctor and Patient are subclasses of Person and there-
fore inherit the attributes and methods of the Person class. Inheritance makes it simpler to
defi ne classes. Instead of repeating the attributes and methods in the Doctor and Patient classes
separately, the attributes and methods that are common to both are placed in the Person class
and inherited by the classes below it. Notice how much more effi cient inheritance hierarchies
of object classes are than the same objects without an inheritance hierarchy (see Figure 1-12).
Most classes throughout a hierarchy lead to instances; any class that has instances
is called a concrete class. For example, if Mary Wilson and Jim Maloney are instances of
the Patient class, Patient would be considered a concrete class (see Figure 1-9). Some
classes do not produce instances because they are used merely as templates for other,
FIGURE 1-10
Messages and
Methods
Receptionist
create
Patient
-name
-address
-birthdate
-phone
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
+create()
aPatient
22 Chapter 1 Introduction to Systems Analysis and Design
more-specific classes (especially classes located high up in a hierarchy). The classes are
referred to as abstract classes. Person is an example of an abstract class. Instead of creating
objects from Person, we create instances representing the more-specifi c classes of Specialist
and Patient, both types of Person (see Figure 1-11).
Polymorphism and Dynamic Binding
Polymorphism means that the same message can be interpreted diff erently by diff erent
classes of objects. For example, inserting a patient means something diff erent than inserting
an appointment. Th erefore, diff erent pieces of information need to be collected and stored.
Luckily, we do not have to be concerned with how something is done when using objects.
We can simply send a message to an object, and that object will be responsible for interpret-
ing the message appropriately. For example, if an artist sent the message Draw yourself to a
FIGURE 1-11
Class Hierarchy
with Abstract and
Concrete Classes
Person
Doctor Patient
SpecialistGeneral Practitioner
Abstract classes
Concrete classes
FIGURE 1-12 Inheritance Advantage?
Patient
-name
-address
-birthdate
-phone
-insurance carrier
+updateBirthDate()
+updateInsuranceCarrier()
Person
-name
-address
-birthdate
-phone
+updateBirthDate()
Doctor
Doctor
-name
-address
-birthdate
-phone
-medicalSchoolSpecialty
+updateBirthDate()
+updateMedicalSchoolSpecialty()
VS.
-medicalSchoolSpecialty
+updateMedicalSchoolSpecialty()
Patient
-insurance carrier
+updateInsuranceCarrier()
Object-Oriented Systems Analysis and Design (OOSAD) 23
square object, a circle object, and a triangle object, the results would be very diff erent, even
though the message is the same. Notice in Figure 1-13 how each object responds appropri-
ately (and diff erently) even though the messages are identical.
Polymorphism is made possible through dynamic binding. Dynamic, or late, binding is
a technique that delays typing the object until run-time. Th e specifi c method that is actu-
ally called is not chosen by the object-oriented system until the system is running. Th is is
in contrast to static binding. In a statically bound system, the type of object is determined
at compile-time. Th erefore, the developer has to choose which method should be called
instead of allowing the system to do it. Th is is why most traditional programming lan-
guages have complicated decision logic based on the diff erent types of objects in a system.
For example, in a traditional programming language, instead of sending the message Draw
yourself to the diff erent types of graphical objects in Figure 1-13, we would have to write
decision logic using a case statement or a set of if statements to determine what kind of
graphical object we wanted to draw, and we would have to name each draw function dif-
ferently (e.g., draw square, draw circle, or draw triangle). Th is obviously makes the system
much more complicated and diffi cult to understand.
OBJECT-ORIENTED SYSTEMS ANALYSIS AND DESIGN (OOSAD)
Object-oriented approaches to developing information systems, technically speaking, can use
any of the traditional methodologies. However, the object-oriented approaches are most asso-
ciated with a phased development RAD or agile methodology. Th e primary diff erence between
a traditional approach like structured design and an object-oriented approach is how a prob-
lem is decomposed. In traditional approaches, the problem-decomposition process is either
process-centric or data-centric. However, processes and data are so closely related that it is
diffi cult to pick one or the other as the primary focus. Based on this lack of congruence with the
real world, new object-oriented methodologies have emerged that use the RAD-based sequence
of SDLC phases but attempt to balance the emphasis between process and data by focusing the
decomposition of problems on objects that contain both data and processes.
FIGURE 1-13
Polymorphism
Draw
Yo
ur
se
lf
DrawYourself
DrawYourself
aTriangle
aSquare
aCircle
anArtist
24 Chapter 1 Introduction to Systems Analysis and Design
16 Grady Booch, Ivar Jacobson, and James Rumbaugh, Th e Unifi ed Modeling Language User Guide (Reading, MA:
Addison-Wesley, 1999).
17 For those of you who have experience with traditional structured analysis and design, this is one of the most unusual
aspects of object-oriented analysis and design using UML. Unlike structured approaches, object-oriented approaches
stress focusing on just one use case at a time and distributing that single use case over a set of communicating and
collaborating objects.
According to the creators of the Unifi ed Modeling Language (UML), Grady Booch, Ivar
Jacobson, and James Rumbaugh,16 any modern object-oriented approach to developing infor-
mation systems must be use-case driven, architecture-centric, and iterative and incremental.
Use-Case Driven
Use-case driven means that use cases are the primary modeling tools defi ning the behavior of
the system. A use case describes how the user interacts with the system to perform some activ-
ity, such as placing an order, making a reservation, or searching for information. Th e use cases
are used to identify and to communicate the requirements for the system to the programmers
who must write the system. Use cases are inherently simple because they focus on only one
business process at a time. In contrast, the process model diagrams used by traditional struc-
tured and RAD methodologies are far more complex because they require the systems analyst
and user to develop models of the entire system. With traditional methodologies, each system
is decomposed into a set of subsystems, which are, in turn, decomposed into further subsys-
tems, and so on. Th is goes on until no further process decomposition makes sense, and it oft en
requires dozens of pages of interlocking diagrams. In contrast, a use case focuses on only one
business process at a time, so developing models is much simpler.17
Architecture-Centric
Any modern approach to systems analysis and design should be architecture-centric.
Architecture-centric means that the underlying soft ware architecture of the evolving system
specifi cation drives the specifi cation, construction, and documentation of the system. Modern
object-oriented systems analysis and design approaches should support at least three separate
but interrelated architectural views of a system: functional, static, and dynamic. Th e functional,
or external, view describes the behavior of the system from the perspective of the user. Th e
structural, or static, view describes the system in terms of attributes, methods, classes, and
relationships. Th e behavioral, or dynamic, view describes the behavior of the system in terms
of messages passed among objects and state changes within an object.
Iterative and Incremental
Modern object-oriented systems analysis and design approaches emphasize iterative and
incremental development that undergoes continuous testing and refi nement throughout the
life of the project. Th is implies that the systems analysts develop their understanding of a
user’s problem by building up the three architectural views little by little. Th e systems analyst
does this by working with the user to create a functional representation of the system under
study. Next, the analyst attempts to build a structural representation of the evolving system.
Using the structural representation of the system, the analyst distributes the functionality of
the system over the evolving structure to create a behavioral representation of the evolving
system. As an analyst works with the user in developing the three architectural views of the
evolving system, the analyst iterates over each of and among the views. Th at is, as the analyst
better understands the structural and behavioral views, the analyst uncovers missing require-
ments or misrepresentations in the functional view. Th is, in turn, can cause changes to be
The Unifi ed Process 25
cascaded back through the structural and behavioral views. All three architectural views of
the system are interlinked and dependent on each other (see Figure 1-14). As each increment
and iteration is completed, a more-complete representation of the user’s real functional
requirements is uncovered.
Benefi ts of Object-Oriented Systems Analysis and Design
Concepts in the object-oriented approach enable analysts to break a complex system into
smaller, more-manageable modules, work on the modules individually, and easily piece the
modules back together to form an information system. Th is modularity makes systems devel-
opment easier to grasp, easier to share among members of a project team, and easier to com-
municate to users, who are needed to provide requirements and confi rm how well the system
meets the requirements throughout the systems development process. By modularizing systems
development, the project team actually is creating reusable pieces that can be plugged into
other systems eff orts or used as starting points for other projects. Ultimately, this can save time
because new projects don’t have to start completely from scratch.
THE UNIFIED PROCESS
Th e Unifi ed Process is a specifi c methodology that maps out when and how to use the var-
ious Unifi ed Modeling Language (UML) techniques for object-oriented analysis and design.
Th e primary contributors were Grady Booch, Ivar Jacobsen, and James Rumbaugh. Whereas
the UML provides structural support for developing the structure and behavior of an infor-
mation system, the Unifi ed Process provides the behavioral support. Th e Unifi ed Process, of
course, is use-case driven, architecture-centric, and iterative and incremental. Furthermore,
the Unifi ed Process is a two-dimensional systems development process described by a set of
phases and workfl ows. Th e phases are inception, elaboration, construction, and transition.
Th e workfl ows include business modeling, requirements, analysis, design, implementation,
test, deployment, confi guration and change management, project management, and environ-
ment.18 Figure 1-15 depicts the Unifi ed Process.
FIGURE 1-14
Iterative and
Incremental
Development
18 Th e material in this section is based on Khawar Zaman Ahmed and Cary E. Umrysh, Developing Enterprise Java
Applications with J2EE and UML (Boston, MA: Addison-Wesley, 2002); Jim Arlow and Ila Neustadt, UML and Th e
Unifi ed Process: Practical Object-Oriented Analysis & Design (Boston, MA: Addison-Wesley, 2002); Peter Eeles,
Kelli Houston, and Wojtek Kozacynski, Building J2EE Applications with the Rational Unifi ed Process (Boston, MA:
Addison-Wesley, 2003); Ivar Jacobson, Grady Booch, and James Rumbaugh, Th e Unifi ed Soft ware Development Process
(Reading, MA: Addison-Wesley, 1999); Phillipe Krutchten, Th e Rational Unifi ed Process: An Introduction, 2nd Ed.
(Boston, MA: Addison-Wesley, 2000); “Rational Unifi ed Process: Best Practices for Soft ware Development Teams,”
Rational Soft ware White Paper, TP026B, Rev 11/01.
Functional
view
Structural
view
Behavioral
view
Object-Oriented
26 Chapter 1 Introduction to Systems Analysis and Design
Phases
Th e phases of the Unifi ed Process support an analyst in developing information systems in
an iterative and incremental manner. Th e phases describe how an information system evolves
through time. Depending on which development phase the evolving system is currently in,
the level of activity varies over the workfl ows. Th e curve in Figure 1-15 associated with each
workfl ow approximates the amount of activity that takes place during the specifi c phase. For
example, the inception phase primarily involves the business modeling and requirements work-
fl ows, while practically ignoring the test and deployment workfl ows. Each phase contains a set
of iterations, and each iteration uses the various workfl ows to create an incremental version of
the evolving system. As the system evolves through the phases, it improves and becomes more
complete. Each phase has objectives, a focus of activity over the workfl ows, and incremental
deliverables. Each of the phases is described next.
Inception In many ways, the inception phase is very similar to the planning phase of a tra-
ditional SDLC approach. In this phase, a business case is made for the proposed system. Th is
includes feasibility analysis that should answer questions such as the following:
Do we have the technical capability to build it (technical feasibility)?
If we build it, will it provide business value (economic feasibility)?
If we build it, will it be used by the organization (organizational feasibility)?
FIGURE 1-15 The Unifi ed Process
Business Modeling
Phases Inception
Supporting Workflows
Elaboration Construction Transition
Requirements
Analysis
Design
Implementation
Configuration and
Change Management
Iter
1
… Iter
i
Iter
i + 1
… Iter
j
Iter
j + 1
… Iter
k
Iter
k + 1
… Iter
m
Project Management
Environment
Test
Deployment
Phases Inception
Engineering Workflows
Elaboration Construction Transition
The Unifi ed Process 27
19 With UML comprising fi ft een diff erent, related diagramming techniques, keeping the diagrams coordinated and the
diff erent versions of the evolving system synchronized is typically beyond the capabilities of a mere mortal systems devel-
oper. Th ese tools typically include project management and CASE tools. We describe the use of these tools in Chapter 2.
To answer these questions, the development team performs work related primarily to
the business modeling, requirements, and analysis workfl ows. In some cases, depending on
the technical diffi culties that could be encountered during the development of the system,
a throwaway prototype is developed. Th is implies that the design, implementation, and test
workfl ows could also be involved. Th e project management and environment supporting
workfl ows are very relevant to this phase. Th e primary deliverables from the inception phase
are a vision document that sets the scope of the project; identifi es the primary requirements
and constraints; sets up an initial project plan; and describes the feasibility of and risks asso-
ciated with the project, the adoption of the necessary environment to develop the system, and
some aspects of the problem domain classes being implemented and tested.
Elaboration When we typically think about object-oriented systems analysis and design,
the activities related to the elaboration phase of the Unifi ed Process are the most relevant.
Th e analysis and design workfl ows are the primary focus during this phase. Th e elaboration
phase continues with developing the vision document, including fi nalizing the business
case, revising the risk assessment, and completing a project plan in suffi cient detail to allow
the stakeholders to be able to agree with constructing the actual fi nal system. It deals with
gathering the requirements, building the UML structural and behavioral models of the
problem domain, and detailing how the problem domain models fi t into the evolving system
architecture. Developers are involved with all but the deployment engineering workfl ow in
this phase. As the developers iterate over the workfl ows, the importance of addressing
confi guration and change management becomes apparent. Also, the development tools
acquired during the inception phase become critical to the success of the project during
this phase.19 Th e primary deliverables of this phase include the UML structure and behavior
diagrams and an executable of a baseline version of the evolving information system. Th e
baseline version serves as the foundation for all later iterations. By providing a solid founda-
tion at this point, the developers have a basis for completing the system in the construction
and transition phases.
Construction Th e construction phase focuses heavily on programming the evolving infor-
mation system. Th is phase is primarily concerned with the implementation workfl ow. How-
ever, the requirements workfl ow and the analysis and design workfl ows also are involved
with this phase. It is during this phase that missing requirements are identifi ed and the
analysis and design models are fi nally completed. Typically, there are iterations of the
workfl ows during this phase, and during the last iteration, the deployment workfl ow kicks
into high gear. Th e confi guration and change management workfl ow, with its version-con-
trol activities, becomes extremely important during the construction phase. At times, an
iteration has to be rolled back. Without good version controls, rolling back to a previous
version (incremental implementation) of the system is nearly impossible. Th e primary
deliverable of this phase is an implementation of the system that can be released for beta
and acceptance testing.
Transition Like the construction phase, the transition phase addresses aspects typically
associated with the implementation phase of a traditional SDLC approach. Its primary
focus is on the testing and deployment workfl ows. Essentially, the business modeling,
requirements, and analysis workfl ows should have been completed in earlier iterations
of the evolving information system. Furthermore, the testing workfl ow will have been
28 Chapter 1 Introduction to Systems Analysis and Design
executing during the earlier phases of the evolving system. Depending on the results
from the testing workfl ow, some redesign and programming activities on the design and
implementation workfl ows could be necessary, but they should be minimal at this point.
From a managerial perspective, the project management, confi guration and change man-
agement, and environment are involved. Some of the activities that take place are beta and
acceptance testing, fi ne-tuning the design and implementation, user training, and rolling
out the fi nal product onto a production platform. Obviously, the primary deliverable is
the actual executable information system. Th e other deliverables include user manuals, a
plan to support the users, and a plan for upgrading the information system in the future.
Workfl ows
Th e workfl ows describe the tasks or activities that a developer performs to evolve an infor-
mation system over time. Th e workfl ows of the Unifi ed Process are grouped into two broad
categories: engineering and supporting.
Engineering Workfl ows Engineering workfl ows include business-modeling, requirements,
analysis, design, implementation, test, and deployment workfl ows. Th e engineering work-
fl ows deal with the activities that produce the technical product (i.e., the information system).
Business Modeling Workfl ow Th e business-modeling workfl ow uncovers problems and
identifi es potential projects within a user organization. Th is workfl ow aids management in
understanding the scope of the projects that can improve the effi ciency and eff ectiveness of a
user organization. Th e primary purpose of business modeling is to ensure that both developer
and user organizations understand where and how the to-be-developed information system
fi ts into the business processes of the user organization. Th is workfl ow is primarily exe-
cuted during the inception phase to ensure that we develop information systems that make
business sense. Th e activities that take place on this workfl ow are most closely associated with
the planning phase of the traditional SDLC; however, requirements gathering, and use-case
and business process modeling techniques also help us to understand the business situation.
Requirements Workfl ow In the Unifi ed Process, the requirements workfl ow includes elic-
iting both functional and nonfunctional requirements. Typically, requirements are gathered
from project stakeholders, such as end users, managers within the end user organization, and
even customers. Th e requirements workfl ow is used the most during the inception and elab-
oration phases. Th e identifi ed requirements are very helpful for developing the vision docu-
ment and the use cases used throughout the development process. Additional requirements
tend to be discovered throughout the development process. In fact, only the transition phase
tends to have few, if any, additional requirements identifi ed.
Analysis Workfl ow Th e analysis workfl ow primarily addresses the creation of an analysis
model of the problem domain. In the Unifi ed Process, the analyst begins designing the archi-
tecture associated with the problem domain; using the UML, the analyst creates structural and
behavior diagrams that depict a description of the problem domain classes and their inter-
actions. Th e primary purpose of the analysis workfl ow is to ensure that both the developer
and user organizations understand the underlying problem and its domain without overana-
lyzing. If they are not careful, analysts can create analysis paralysis, which occurs when the
project becomes so bogged down with analysis that the system is never actually designed or
implemented. A second purpose of the analysis workfl ow is to identify useful reusable classes
for class libraries. By reusing predefi ned classes, the analyst can avoid reinventing the wheel
The Unifi ed Process 29
when creating the structural and behavior diagrams. Th e analysis workfl ow is predominantly
associated with the elaboration phase, but like the requirements workfl ow, it is possible that
additional analysis will be required throughout the development process.
Design Workfl ow Th e design workfl ow transitions the analysis model into a form that can
be used to implement the system: the design model. Whereas the analysis workfl ow concen-
trated on understanding the problem domain, the design workfl ow focuses on developing a
solution that will execute in a specifi c environment. Basically, the design workfl ow simply
enhances the description of the evolving system by adding classes that address the environ-
ment of the system to the evolving analysis model. Th e design workfl ow uses activities such
as detailed problem domain class design, optimization of the evolving information system,
database design, user-interface design, and physical architecture design. Th e design workfl ow
is associated primarily with the elaboration and construction phases of the Unifi ed Process.
Implementation Workfl ow Th e primary purpose of the implementation workfl ow is to
create an executable solution based on the design model (i.e., programming). Th is includes
not only writing new classes but also incorporating reusable classes from executable class
libraries into the evolving solution. As with any programming activity, the new classes and
their interactions with the incorporated reusable classes must be tested. Finally, in the case of
multiple groups performing the implementation of the information system, the implementers
also must integrate the separate, individually tested modules to create an executable version
of the system. Th e implementation workfl ow is associated primarily with the elaboration and
construction phases.
Testing Workfl ow Th e primary purpose of the testing workfl ow is to increase the quality
of the evolving system. Testing goes beyond the simple unit testing associated with the
implementation workfl ow. In this case, testing also includes testing the integration of all
modules used to implement the system, user acceptance testing, and the actual alpha test-
ing of the soft ware. Practically speaking, testing should go on throughout the development
of the system; testing of the analysis and design models occurs during the elaboration and
construction phases, whereas implementation testing is performed primarily during the
construction and, to some degree, transition phases. Basically, at the end of each iteration
during the development of the information system, some type of test should be performed.
Deployment Workfl ow Th e deployment workfl ow is most associated with the transition
phase of the Unifi ed Process. Th e deployment workfl ow includes activities such as soft ware
packaging, distribution, installation, and beta testing. When actually deploying the new sys-
tem into a user organization, the developers might have to convert the current data, interface
the new soft ware with the existing soft ware, and train the end user to use the new system.
Supporting Workfl ows Th e supporting workfl ows include the project management, con-
fi guration and change management, and environment workfl ows. Th e supporting workfl ows
focus on the managerial aspects of information systems development.
Project Management Workfl ow Whereas the other workfl ows associated with the Unifi ed
Process are technically active during all four phases, the project management workfl ow is
the only truly cross-phase workfl ow. Th e development process supports incremental and
iterative development, so information systems tend to grow or evolve over time. At the end
of each iteration, a new incremental version of the system is ready for delivery. Th e project
management workfl ow is quite important owing to the complexity of the two-dimensional
30 Chapter 1 Introduction to Systems Analysis and Design
development model of the Unifi ed Process (workfl ows and phases). Th is workfl ow’s activities
include identifying and managing risks, managing scope, estimating the time to complete
each iteration and the entire project, estimating the cost of the individual iteration and the
whole project, and tracking the progress being made toward the fi nal version of the evolving
information system.
Confi guration and Change Management Workfl ow Th e primary purpose of the confi gu-
ration and change management workfl ow is to keep track of the state of the evolving system.
In a nutshell, the evolving information system comprises a set of artifacts (e.g., diagrams,
source code, and executables). During the development process, these artifacts are modifi ed.
A substantial amount of work—and, hence, money—is involved in developing the artifacts.
Th e artifacts themselves should be handled as any expensive asset would be handled—access
controls must be put into place to safeguard the artifacts from being stolen or destroyed. Fur-
thermore, because the artifacts are modifi ed on a regular, if not continuous, basis, good ver-
sion control mechanisms should be established. Finally, a good deal of project management
information needs to be captured (e.g., author, time, and location of each modifi cation). Th e
confi guration and change management workfl ow is associated mostly with the construction
and transition phases.
Environment Workfl ow During the development of an information system, the develop-
ment team needs to use diff erent tools and processes. Th e environment workfl ow addresses
these needs. For example, a CASE tool that supports the development of an object-oriented
information system via the UML could be required. Other tools necessary include pro-
gramming environments, project management tools, and confi guration management tools.
Th e environment workfl ow involves acquiring and installing these tools. Even though this
workfl ow can be active during all of the phases of the Unifi ed Process, it should be involved
primarily with the inception phase.
Extensions to the Unifi ed Process
As large and as complex as the Unifi ed Process is, many authors have pointed out a set of
critical weaknesses. First, the Unifi ed Process does not address staffi ng, budgeting, or contract
management issues. Th ese activities were explicitly left out of the Unifi ed Process. Second, the
Unifi ed Process does not address issues relating to maintenance, operations, or support of
the product once it has been delivered. Th us, it is not a complete soft ware process; it is only
a development process. Th ird, the Unifi ed Process does not address cross- or inter-project
issues. Considering the importance of reuse in object-oriented systems development and the
fact that in many organizations employees work on many diff erent projects at the same time,
leaving out inter-project issues is a major omission.
To address these omissions, Ambler and Constantine suggest adding a production phase
and two workfl ows: the operations and support workfl ow and the infrastructure management
workfl ow (see Figure 1-16).20 In addition to these new workfl ows, the test, deployment, and
environment workfl ows are modifi ed, and the project management and the confi guration
and change management workfl ows are extended into the production phase. Th ese extensions
20 S. W. Ambler and L. L. Constantine, Th e Unifi ed Process Inception Phase: Best Practices in Implementing the UP
(Lawrence, KS: CMP Books, 2000); S. W. Ambler and L. L. Constantine, Th e Unifi ed Process Elaboration Phase:
Best Practices in Implementing the UP (Lawrence, KS: CMP Books, 2000); S. W. Ambler and L. L. Constantine, Th e
Unifi ed Process Construction Phase: Best Practices in Implementing the UP (Lawrence, KS: CMP Books, 2000); S. W.
Ambler and L. L. Constantine, Th e Unifi ed Process Transition and Production Phases: Best Practices in Implementing
the UP (Lawrence, KS: CMP Books, 2002).
The Unifi ed Process 31
are based on alternative object-oriented soft ware processes: the OPEN process (Object-oriented
Process, Environment, and Notation) and the Object-Oriented Soft ware Process.21
Production Phase Th e production phase is concerned primarily with issues related to the
soft ware product aft er it has been successfully deployed. Th is phase focuses on issues related
to updating, maintaining, and operating the soft ware. Unlike the previous phases, there are
no iterations or incremental deliverables. If a new release of the soft ware is to be developed,
21 S. W. Ambler, Process Patterns—Building Large-Scale Systems Using Object Technology (Cambridge, UK: SIGS
Books/Cambridge University Press, 1998); S. W. Ambler, More Process Patterns—Delivering Large-Scale Systems
Using Object Technology (Cambridge, UK: SIGS Books/Cambridge University Press, 1999); I. Graham, B. Henderson-
Sellers, and H. Younessi, Th e OPEN Process Specifi cation (Harlow, UK: Addison-Wesley, 1997); B. Henderson-Sellers and
B. Unhelkar, OPEN Modeling with UML (Harlow, UK: Addison-Wesley, 2000).
FIGURE 1-16 The Enhanced Unifi ed Process
Business Modeling
Phases Inception
Supporting Workflows
Elaboration Construction Transition Production
Requirements
Analysis
Design
Implementation
Configuration and
Change Management
Infrastructure
Management
Project Management
Environment
Operations and Support
Iter
1
… Iter
i
Iter
i + 1
… Iter
j
Iter
j + 1
… Iter
k
Iter
k + 1
… Iter
m
Test
Deployment
Phases Inception
Engineering Workflows
Elaboration Construction Transition Production
32 Chapter 1 Introduction to Systems Analysis and Design
then the developers must begin a new run through the fi rst four phases. Based on the activi-
ties that take place during this phase, no engineering workfl ows are relevant. Th e supporting
workfl ows that are active during this phase include the confi guration and change manage-
ment workfl ow, the project management workfl ow, the new operations and support work-
fl ow, and the infrastructure management workfl ow.
Operations and Support Workfl ow Th e operations and support workfl ow, as you might guess,
addresses issues related to supporting the current version of the soft ware and operating the
soft ware on a daily basis. Activities include creating plans for the operation and support of the
soft ware product once it has been deployed, creating training and user documentation, putting
into place necessary backup procedures, monitoring and optimizing the performance of the
soft ware, and performing corrective maintenance on the soft ware. Th is workfl ow becomes
active during the construction phase; its level of activity increases throughout the transition
and, fi nally, the production phase. Th e workfl ow fi nally drops off when the current version of
the soft ware is replaced by a new version. Many developers are under the false impression that
once the soft ware has been delivered to the customer, their work is fi nished. In most cases, the
work of supporting the soft ware product is much more costly and time consuming than the
original development. At that point, the developer’s work may have just begun.
Infrastructure Management Workfl ow Th e infrastructure management workfl ow’s primary
purpose is to support the development of the infrastructure necessary to develop object-
oriented systems. Activities such as development and modifi cation of libraries, standards,
and enterprise models are very important. When the development and maintenance of a
problem-domain architecture model goes beyond the scope of a single project and reuse
is going to occur, the infrastructure management workfl ow is essential. Another very impor-
tant set of cross-project activities is the improvement of the soft ware development process.
Because the activities on this workfl ow tend to aff ect many projects and the Unifi ed Process
focuses only on a specifi c project, the Unifi ed Process tends to ignore these activities (i.e., they
are simply beyond the scope and purpose of the Unifi ed Process).
Existing Workfl ow Modifi cations and Extensions In addition to the workfl ows that were
added to address defi ciencies contained in the Unifi ed Process, existing workfl ows had to
be modifi ed and/or extended into the production phase. Th ese workfl ows include the test,
deployment, environment, project management, and confi guration and change management
workfl ows.
Test Workfl ow For high-quality information systems to be developed, testing should be
done on every deliverable, including those created during the inception phase. Otherwise, less
than high-quality systems will be delivered to the customer.
Deployment Workfl ow Legacy systems exist in most corporations today, and these systems
have databases associated with them that must be converted to interact with the new systems.
Owing to the complexity of deploying new systems, the conversion requires signifi cant plan-
ning. Th erefore, the activities on the deployment workfl ow need to begin in the inception phase
instead of waiting until the end of the construction phase, as suggested by the Unifi ed Process.
Environment Workfl ow Th e environment workfl ow needs to be modifi ed to include activ-
ities related to setting up the operations and production environment. Th e actual work per-
formed is similar to the work related to setting up the development environment that was
performed during the inception phase. In this case, the additional work is performed during
the transition phase.
The Unifi ed Process 33
Project Management Workfl ow Even though the project management workfl ow does not
include staffi ng the project, managing the contracts among the customers and vendors, and
managing the project’s budget, these activities are crucial to the success of any soft ware
development project. We suggest extending project management to include these activities.
Th is workfl ow should additionally occur in the production phase to address issues such as
training, staff management, and client relationship management.
Confi guration and Change Management Workfl ow Th e confi guration and change manage-
ment workfl ow is extended into the new production phase. Activities performed during the
production phase include identifying potential improvements to the operational system and
assessing the potential impact of the proposed changes. Once developers have identifi ed these
changes and understood their impact, they can schedule the changes to be made and deployed
with future releases.
Figure 1-17 shows the chapters in which the Enhanced Unifi ed Process’s phases and
workfl ows are covered. Given the off shore outsourcing and automation of information
Enhanced UP Phases Chapters
Inception 2–4
Elaboration 3–11
Construction 8, 12
Transition 12–13
Production 13
Enhanced UP
Engineering Workfl ows Chapters
Business Modeling 2–5
Requirements 3–5, 10
Analysis 3–7
Design 7–11
Implementation 9, 12
Test 4–7, 12
Deployment 13
Enhanced UP
Supporting Workfl ows Chapters
Project Management 2, 13
Confi guration and
Change Management
13
Environment 2
Operations and Support 13
Infrastructure
Management
2
FIGURE 1-17 The Enhanced Unifi ed Process and the Textbook Organization
34 Chapter 1 Introduction to Systems Analysis and Design
technology,22 in this textbook, we focus primarily on the elaboration phase and the busi-
ness modeling, requirements, analysis, design, and project management workfl ows of the
Enhanced Unifi ed Process. However, as Figure 1-17 shows, the other phases and workfl ows
are covered. In many object-oriented systems development environments today, code
generation is supported. Th us, from a business perspective, we believe the activities associated
with these workfl ows are the most important.
THE UNIFIED MODELING LANGUAGE
Until 1995, object concepts were popular but implemented in many diff erent ways by
diff erent developers. Each developer had his or her own methodology and notation (e.g.,
Booch, Coad, Moses, OMT, OOSE, SOMA).23 Th en in 1995, Rational Soft ware brought
three industry leaders together to create a single approach to object-oriented systems
development. Grady Booch, Ivar Jacobson, and James Rumbaugh worked with others to
create a standard set of diagramming techniques known as the Unifi ed Modeling Language
(UML). Th e objective of UML was to provide a common vocabulary of object-oriented
terms and diagramming techniques rich enough to model any systems development pro-
ject from analysis through implementation. In November 1997, the Object Management
Group (OMG) formally accepted UML as the standard for all object developers. During the
following years, the UML has gone through multiple minor revisions. Th e current version
of UML is Version 2.5.
Version 2.5 of the UML defi nes a set of fi ft een diagramming techniques used to model a
system. Th e diagrams are broken into two major groupings: one for modeling the structure
of a system and one for modeling behavior. Structure diagrams provide a way to represent
the data and static relationships in an information system. Th e structure diagrams include
class, object, package, deployment, component, composite structure, and profi le diagrams.
Behavior diagrams provide the analyst with a way to depict the dynamic relationships among
the instances or objects that represent the business information system. Th ey also allow mod-
eling of the dynamic behavior of individual objects throughout their lifetime. Th e behavior
diagrams support the analyst in modeling the functional requirements of an evolving infor-
mation system. Th e behavior modeling diagrams include activity, sequence, communication,
interaction overview, timing, behavior state machine, protocol state machine, and use-case
diagrams.24 Figure 1-18 provides an overview of these diagrams.
22 See Th omas L. Friedman, Th e World Is Flat: A Brief History of the Twenty-First Century, Updated and Expanded
Edition (New York: Farrar, Straus, and Giroux, 2006); Daniel H. Pink, A Whole New Mind: Why Right-Brainers Will
Rule the Future (New York: Riverhead Books, 2006).
23 See Grady Booch, Object-Oriented Analysis and Design with Applications, 2nd Ed. (Redwood City, CA: Benjamin/
Cummings, 1994); Peter Coad and Edward Yourdon, Object-Oriented Analysis, 2nd Ed. (Englewood Cliff s, NJ:
Yourdon Press, 1991); Peter Coad and Edward Yourdon, Object-Oriented Design (Englewood Cliff s, NJ: Yourdon
Press, 1991); Brian Henderson-Sellers and Julian Edwards, Book Two of Object-Oriented Knowledge: Th e Working
Object (Sydney, Australia: Prentice Hall, 1994); James Rumbaugh, Michael Blaha, William Premerlani, Frederick
Eddy, and William Lorensen, Object-Oriented Modeling and Design (Englewood Cliff s, NJ: Prentice Hall, 1991);
Ivar Jacobson, Magnus Christerson, Patrik Jonsson, and Gunnar Overgaard, Object-Oriented Soft ware Engineering:
A Use Case Approach (Wokingham, England: Addison-Wesley, 1992); Ian Graham, Migrating to Object Technology
(Wokingham, England: Addison-Wesley, 1994).
24 Th e material contained in this section is based on the Unifi ed Modeling Language: Superstructure Version 2.4,
ptc/2010-11-14 (www.uml.org). Additional useful references include Michael Jesse Chonoles and James A. Schardt,
UML 2 for Dummies (Indianapolis, IN: Wiley, 2003); Hans-Erik Eriksson, Magnus Penker, Brian Lyons, and David
Fado, UML 2 Toolkit (Indianapolis, IN: Wiley, 2004); Kendall Scott, Fast Track UML 2.0 (Berkeley, CA: Apress,
2004). For a complete description of all diagrams, see www.uml.org.
The Unifi ed Modeling Language 35
Structure Diagrams
Class Illustrate the relationships between classes modeled Analysis, Design
in the system
Object Illustrate the relationships between objects modeled Analysis, Design
in the system; used when actual instances of the classes
will better communicate the model
Package Group other UML elements together to form Analysis, Design,
higher-level constructs Implementation
Deployment Show the physical architecture of the system; can also Physical Design,
be used to show software components being deployed Implementation
onto the physical architecture
Component Illustrate the physical relationships among the software Physical Design,
components Implementation
Composite Structure Design Illustrate the internal structure of a class, i.e., the Analysis, Design
relationships among the parts of a class
Profi le Used to develop extensions to the UML itself None
Behavioral Diagrams
Activity Illustrate business workfl ows independent of classes, the fl ow Analysis, Design
of activities in a use case, or detailed design of a method
Sequence Model the behavior of objects within a use case; Analysis, Design
focuses on the time-based ordering of an activity
Communication Model the behavior of objects within a use case; Analysis, Design
focus on the communication among a set of
collaborating objects of an activity
Interaction Overview Illustrate an overview of the fl ow of control of a process Analysis, Design
Timing Illustrate the interaction among a set of objects and the state Analysis, Design
changes they go through along a time axis
Behavioral State Machine Examine the behavior of one class Analysis, Design
Protocol State Machine Illustrate the dependencies among the different Analysis, Design
interfaces of a class
Use-Case Capture business requirements for the system and illustrate Analysis
the interaction between the system and its environment
FIGURE 1-18 UML 2.5 Diagram Summary
Diagram Name Used to… Primary Phase
Depending on where in the development process the system is, diff erent diagrams play
a more important role. In some cases, the same diagramming technique is used throughout
the development process. In that case, the diagrams start off very conceptual and abstract.
As the system is developed, the diagrams evolve to include details that ultimately lead to
generating and developing code. In other words, the diagrams move from documenting the
requirements to laying out the design. Overall, the consistent notation, integration among
the diagramming techniques, and application of the diagrams across the entire development
process make the UML a powerful and fl exible language for analysts and developers. Later
chapters provide more detail on using a subset of the UML in object-oriented systems analysis
36 Chapter 1 Introduction to Systems Analysis and Design
and design. In particular, these chapters describe activity, use-case, class, object, sequence,
communication, package, and deployment diagrams and the behavior state machines. We
also introduce an optional UML diagram, the windows navigation diagram, that is an exten-
sion to the behavioral state machine that is used to design user navigation through an infor-
mation system’s user interfaces.
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Th is course will introduce many new concepts regarding object-oriented analysis and
design. To make these concepts more relevant and understandable, we will apply the
concepts, introduced in each chapter, to a fi ctitious company called Patterson Superstore.
Patterson is a retail chain established in Pittsburgh, PA, in 1985. Currently, Patterson
uses a mobile application to facilitate prescription order, notifi cation, and auto refi ll ser-
vices. Th is service is widely used by Patterson’s client base, and Patterson has leveraged
this mobile app to gain an advantage over less technically advanced competitors.
Clients now want to use this technology to access health clinic services. Th e Vice
President of Pharmacy Services, Max Ross, would like to use this opportunity to position
Patterson as a leader in the use of technology use for clinic access. Th e system that he
envisions will enable real-time communication with medical personnel (audio, video,
and text), mobile appointment scheduling, telehealth assessment, and diagnosis of minor
problems through video house calls. Th roughout the book, we will revisit Patterson
Superstore to see how the concepts introduced in each chapter aff ect this project.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the four primary phases of the Systems Development Life Cycle (SDLC).
Explain the evolution of system development methodologies from process-centric to data-centric to RAD-based
methodologies.
Explain the diff erent roles played by a systems analyst in the process of developing information systems.
Describe the basic characteristics of object-oriented systems: objects, attributes, methods, messages, encapsulation,
information hiding, polymorphism, dynamic binding, and inheritance.
Discuss the three basic characteristics of all object-oriented systems analysis and design approach: use-case driven,
architecture-centric, and iterative and incremental development.
Describe the Unifi ed Process.
List and categorize, as to their primary purpose, the diff erent diagrams associated with the Unifi ed Modeling
Language (UML).
KEY TERMS
Abstract classes
Agile development
A-kind-of
Analysis model
Analysis paralysis
Analysis phase
Analysis strategy
Analysis workfl ow
Approval committee
Architecture-centric
Architecture design
As-is system
Attribute
Behavior
Behavior diagrams
Behavioral view
Business analyst
Business modeling
workfl ow
Change agent
Questions 37
Change management
analyst
Class
Concrete classes
Confi guration and change
management workfl ow
Construction
Construction phase
Database and fi le
specifi cation
Data-centered
methodology
Deliverable
Deployment workfl ow
Design model
Design phase
Design prototype
Design strategy
Design workfl ow
Dynamic binding
Dynamic view
Elaboration phase
Encapsulation
Engineering workfl ow
Environment
workfl ow
External view
Extreme programming (XP)
Feasibility analysis
Functional view
Gradual refi nement
Implementation phase
Implementation workfl ow
Inception phase
Incremental
Information hiding
Infrastructure analyst
Infrastructure management
workfl ow
Inherit
Inheritance
Instance
Interface design
Iterative
Message
Method
Methodology
Object
Object Management
Group (OMG)
Object-oriented
methodologies
Operations and support
workfl ow
Parallel development
Phased development
Phases
Planning phase
Polymorphism
Process-centered methodology
Production phase
Program design
Programmer
Project management
Project management
workfl ow
Project manager
Project plan
Project sponsor
Prototyping
Rapid application development
(RAD)
Requirements gathering
Requirements workfl ow
Scrum
State
Static binding
Static view
Structural view
Structure diagrams
Structured design
Subclass
Superclass
Support plan
System proposal
System prototype
System request
System specifi cation
Systems analyst
Systems development life
cycle (SDLC)
Technical writer
Testing workfl ow
Th rowaway prototyping
Training plan
Transition phase
Unifi ed Modeling Language
(UML)
Use case
Use-case driven
Version
Waterfall development
Workfl ows
Workplan
QUESTIONS
1. Compare and contrast phases, steps, techniques, and
deliverables.
2. Describe the major phases in the SDLC.
3. Describe the principal steps in the planning phase.
What are the major deliverables?
4. Describe the principal steps in the analysis phase.
What are the major deliverables?
5. Describe the principal steps in the design phase. What
are the major deliverables?
6. Describe the principal steps in the implementation
phase. What are the major deliverables?
7. What are the roles of a project sponsor and the
approval committee?
8. What does gradual refi nement mean in the context of
SDLC?
9. Compare and contrast process-centered methodolo-
gies with data-centered methodologies.
10. Compare and contrast structured design-based meth-
odologies in general to RAD-based methodologies in
general.
11. Compare and contrast extreme programming and
throwaway prototyping.
12. Describe the major elements in and issues with water-
fall development.
13. Describe the major elements in and issues with parallel
development.
14. Describe the major elements in and issues with phased
development.
15. Describe the major elements in and issues with
prototyping.
16. Describe the major elements in and issues with throw-
away prototyping.
17. Describe the major elements in and issues with XP.
18. Describe the major elements in and issues with
Scrum.
19. What are the key factors in selecting a methodology?
20. What are the major roles played by a systems analyst
on a project team?
21. Compare and contrast the role of a systems analyst,
business analyst, and infrastructure analyst.
22. What is the diff erence between classes and objects?
23. What are methods and messages?
24. Why are encapsulation and information hiding
important characteristics of object-oriented systems?
38 Chapter 1 Introduction to Systems Analysis and Design
EXERCISES
A. Suppose you are a project manager using a water-
fall development-based methodology on a large and
complex project. Your manager has just read the latest
article in Computerworld that advocates replacing
this methodology with prototyping and comes to you
requesting that you switch. What would you say?
B. Th e basic types of methodologies discussed in this
chapter can be combined and integrated to form new
hybrid methodologies. Suppose you were to com-
bine throwaway prototyping with the use of waterfall
development. What would the methodology look
like? Draw a picture (similar to those in Figures 1–2
through 1–7). How would this new methodology
compare to the others?
C. Look on the Web for diff erent kinds of job opportu-
nities that are available for people who want analyst
positions? Compare and contrast the skills that the ads
ask for to the skills that we presented in this chapter.
D. Th ink about your ideal analyst position. Write an ad
to hire someone for that position. What requirements
would the job have? What skills and experience would
be required? How would an applicant be able to demon-
strate having the appropriate skills and experience?
E. Using your favorite Web search engine, fi nd alterna-
tive descriptions of the basic characteristics of object-
oriented systems.
F. Look up object-oriented programming in Wikipedia.
Write a short report based on its entry.
G. Choose an object-oriented programming language,
such as C++, Java, Objective-C, Smalltalk, or VB.Net,
and use the Web to fi nd out how the language supports
the basic characteristics of object-oriented systems.
H. Assume that you have been assigned the task of cre-
ating an object-oriented system that could be used to
support students in fi nding an appropriate apartment
to live in next semester. What are the diff erent types
of objects (i.e., classes) you would want to include in
your system? What attributes or methods would you
want to include in their defi nition? Is it possible to
arrange them into an inheritance hierarchy? If so, do
it. If not, why not?
I. Create an inheritance hierarchy that could be used to
represent the following classes: accountant, customer,
department, employee, manager, organization, and
salesperson.
J. Investigate IBM’s Rational Unifi ed Process (RUP) on the
Web. RUP is a commercial version that extends aspects of
the Unifi ed Process. Write a brief memo describing how it
is related to the Unifi ed Process as described in this chapter.
(Hint: A good website with which to begin is www-01.
ibm.com/soft ware/rational/rup/.)
K. Suppose you are a project manager who typically has
been using a waterfall development-based methodol-
ogy on a large and complex project. Your manager has
just read the latest article in Computerworld that advo-
cates replacing this methodology with the Unifi ed
Process and comes to you requesting you to switch.
What do you say?
L. Suppose you are an analyst working for a small com-
pany to develop an accounting system. Would you use
the Unifi ed Process to develop the system, or would
you prefer one of the other approaches? Why?
M. Suppose you are an analyst developing a new infor-
mation system to automate the sales transactions
and manage inventory for each retail store in a large
chain. Th e system would be installed at each store
and exchange data with a mainframe computer at the
company’s head offi ce. Would you use the Unifi ed
Process to develop the system, or would you prefer
one of the other approaches? Why?
25. What is meant by polymorphism when applied to
object-oriented systems?
26. Compare and contrast dynamic and static binding.
27. What is a use case?
28. What is meant by use-case driven?
29. What is the Unifi ed Modeling Language?
30. Who is the Object Management Group?
31. What is the primary purpose of structure diagrams?
Give some examples of structure diagrams.
32. For what are behavior diagrams used? Give some
examples of behavior diagrams.
33. Why is it important for an OOSAD approach to be
architecture-centric?
34. What does it mean for an OOSAD approach to be
incremental and iterative?
35. What are the phases and workfl ows of the Unifi ed
Process?
36. Compare the phases of the Unifi ed Process with the
phases of the waterfall model.
37. Which phase in the SDLC is most important? Why?
38. Describe the major elements and issues with an object-
oriented approach to developing information systems.
Minicases 39
N. Suppose you are an analyst working for a small com-
pany to develop an accounting system. What type of
methodology would you use? Why?
O. Suppose you are an analyst developing a new execu-
tive information system intended to provide key stra-
tegic information from existing corporate databases
to senior executives to help in their decision making.
What type of methodology would you use? Why?
P. Investigate the Unifi ed Modeling Language on the
Web. Write a paragraph news brief describing the
current state of the UML. (Hint: A good website with
which to begin is www.uml.org.)
Q. Investigate the Object Management Group (OMG) on the
Web. Write a report describing the purpose of the OMG
and what it is involved with besides the UML. (Hint: A
good website with which to begin is www.omg.org.)
R. Using the Web, fi nd a set of CASE tools that support
the UML. A couple of examples include Poseidon,
Rational Rose, and Visual Paradigm. Find at least two
more. Write a short report describing how well they
support the UML, and make a recommendation as to
which one you believe would be best for a project team
to use in developing an object-oriented information
system using the UML.
MINICASES
1. Barbara Singleton, manager of western regional sales
at the WAMAP Company, requested that the IS
department develop a sales force management and
tracking system that would enable her to better mon-
itor the performance of her sales staff . Unfortunately,
owing to the massive backlog of work facing the IS
department, her request was given a low priority.
Aft er six months of inaction by the IS department,
Barbara decided to take matters into her own hands.
Based on the advice of friends, Barbara purchased
simple database soft ware and constructed a sales force
management and tracking system on her own.
Although Barbara’s system has been “completed”
for about six weeks, it still has many features that
do not work correctly, and some functions are full
of errors. Barbara’s assistant is so mistrustful of the
system that she has secretly gone back to using her old
paper-based system, because it is much more reliable.
Over dinner one evening, Barbara complained to
a systems analyst friend, “I don’t know what went
wrong with this project. It seemed pretty simple to
me. Th ose IS guys wanted me to follow this elaborate
set of steps and tasks, but I didn’t think all that really
applied to a PC-based system. I just thought I could
build this system and tweak it around until I got what
I wanted without all the fuss and bother of the meth-
odology the IS guys were pushing. I mean, doesn’t
that just apply to their big, expensive systems?”
Assuming you are Barbara’s systems analyst friend,
how would you respond to her complaint?
2. Marcus Weber, IS project manager at ICAN Mutual
Insurance Co., is reviewing the staffi ng arrangements
for his next major project, the development of an
expert system-based underwriter’s assistant. Th is new
system will involve a whole new way for the under-
writers to perform their tasks. Th e underwriter’s assis-
tant system will function as sort of an underwriting
supervisor, reviewing key elements of each applica-
tion, checking for consistency in the underwriter’s
decisions, and ensuring that no critical factors have
been overlooked. Th e goal of the new system is to
improve the quality of the underwriters’ decisions and
to improve underwriters’ productivity. It is expected
that the new system will substantially change the way
the underwriting staff do their jobs.
Marcus is dismayed to learn that because of budget
constraints, he must choose between one of two availa-
ble staff members. Barry Filmore has had considerable
experience and training in individual and organiza-
tional behavior. Barry has worked on several other
projects in which the end users had to make signifi cant
adjustments to the new system, and Barry seems to
have a knack for anticipating problems and smoothing
the transition to a new work environment. Marcus had
hoped to have Barry’s involvement in this project.
Marcus’s other potential staff member is Kim Dan-
ville. Prior to joining ICAN Mutual, Kim had con-
siderable work experience with the expert system
technologies that ICAN has chosen for this expert
system project. Marcus was counting on Kim to
help integrate the new expert system technology into
ICAN’s systems environment, and also to provide
on-the-job training and insights to the other develop-
ers on this team.
Given that Marcus’s budget will only permit him to
add Barry or Kim to this project team, but not both,
what choice do you recommend for him? Justify your
answer.
40 Chapter 1 Introduction to Systems Analysis and Design
3. Joe Brown, the president of Roanoke Manufacturing,
requested that Jack Jones, the MIS department man-
ager, investigate the viability of selling their products
over the Web. Currently, the MIS department is still
using an IBM mainframe as their primary deploy-
ment environment. As a fi rst step, Jack contacted his
friends at IBM to see if they had any suggestions as to
how Roanoke Manufacturing could move toward sup-
porting sales in an electronic commerce environment
while keeping their mainframe as their main system.
His friends explained that IBM (www.ibm.com) now
supports Java and Linux on their mainframes. Jack has
also learned that IBM owns Rational (www-01.ibm.
com/soft ware/rational/), the creator of the UML and
the Unifi ed Process. Jack’s friends suggested that Jack
investigate using object-oriented systems as a basis for
developing the new system. Th ey also suggested that
using the Rational Unifi ed Process (RUP), Java, and vir-
tual Linux machines on his current mainframe as a way
to support the move toward a distributed electronic
commerce system would protect his current investment
in his legacy systems while allowing the new system to
be developed in a more modern manner. Even though
Jack’s IBM friends were very persuasive, Jack is still a
little wary about moving his operation from a struc-
tured systems approach to this new object-oriented
approach. Assuming that you are one of Jack’s IBM
friends, how would you convince him to move toward
using an object-oriented systems development method,
such as RUP, and using Java and Linux as a basis for
developing and deploying the new system on Roanoke
Manufacturing’s current mainframe?
This chapter primarily describes the project management workfl ow of the Unifi ed Process.
Th e fi rst step in the process is to identify a project that will deliver value to the business
and to create a system request that provides basic information about the proposed system.
Second, the analysts perform a feasibility analysis to determine the technical, economic, and
organizational feasibility of the system; if appropriate, the system is selected and the develop-
ment project begins. Th ird, the project manager estimates the functionality of the project and
identifi es the tasks that need to be performed. Fourth, the manager staff s the project. Finally,
the manager identifi es the tools, standards, and process to be used; identifi es opportunities for
reuse; determines how the current project fi ts into the portfolio of projects currently under
development; and identifi es opportunities to update the overall structure of the fi rm’s port-
folio of systems current in use.
OBJECTIVES
■ Understand the importance of linking the information system to business needs.
■ Be able to create a system request.
■ Understand how to assess technical, economic, and organizational feasibility.
■ Be able to perform a feasibility analysis.
■ Understand how projects are selected in some organizations.
■ Become familiar with work breakdown structures, Gantt charts, and network diagrams.
■ Become familiar with use-case–driven eff ort estimation.
■ Be able to create an iterative project workplan.
■ Understand how to manage the scope, refi ne the estimates, and manage the risk
of a project.
■ Become familiar with how to staff a project.
■ Understand how the environment and infrastructure workfl ows interact with
the project management workfl ow.
INTRODUCTION
Most projects occurring in people’s lives, such as weddings or graduation celebrations,
require planning and management. Months are spent in advance identifying and per-
forming all the tasks that need to get done, such as sending out invitations and selecting a
menu, and time and money are carefully allocated among them. Along the way, decisions
are recorded, problems are addressed, and changes are made. Th e increasing popularity of
the party planner, a person whose sole job is to coordinate a party, suggests how tough this
job can be. In the end, the success of any party has a lot to do with the eff ort that went into
planning along the way. System development projects can be much more complicated than
the projects we encounter in our personal lives—usually, more people are involved (e.g., the
41
C H A P T E R 2
Project Management
42 Chapter 2 Project Management
organization), the costs are higher, and more tasks need to be completed. Owing to the
complexity of soft ware and soft ware development, it is virtually impossible to “know” all of
the possible things that could happen during system development projects. Th erefore, it is
not surprising that “party planners” exist for information systems projects: Th ey are called
project managers.
Project management is the process of planning and controlling the development of a
system within a specifi ed time frame at a minimum cost with the right functionality.1 In
general, a project is a set of activities with a starting point and an ending point meant to
create a system that brings value to the business. A project manager has the primary respon-
sibility for managing the hundreds of tasks and roles that need to be carefully coordinated.
Today, project management is an actual profession, and analysts spend years working on
projects before tackling the management of them. However, in many cases, unreasonable
demands set by project sponsors and business managers can make project management very
diffi cult. Too oft en, the approach of the holiday season, the chance at winning a proposal
with a low bid, or a funding opportunity pressures project managers to promise systems
long before they are able to deliver them. Th ese overly optimistic timetables are thought to
be one of the biggest problems that projects face; instead of pushing a project forward faster,
they result in delays. Another source is the changing nature of information technology. An
innovation in information technology may look so attractive that organizations embrace
projects using this technology without assessing whether the technology adds value to the
organization; instead the technology itself seems important in its own right. Problems can
usually be traced back to the very beginning of the development of the system, where too
little attention was given to identifying the business value and understanding the risks asso-
ciated with the project.
During the inception phase of the Unifi ed Process of a new systems development pro-
ject, someone—a manager, staff member, sales representative, or systems analyst—typically
identifi es some business value that can be gained from using information technology. New
systems development projects should start from a business need or opportunity. Many ideas
for new systems or improvements to existing ones arise from the application of a new tech-
nology, but an understanding of technology is usually secondary to a solid understanding of
the business and its objectives. Th is does not mean that technical people should not recom-
mend new systems projects. In fact, the ideal situation is for both IT people (i.e., the experts
in systems) and business people (i.e., the experts in business) to work closely to fi nd ways for
technology to support business needs. In this way, organizations can leverage the exciting
innovative technologies that are available while ensuring that projects are based upon real
business objectives, such as increasing sales, improving customer service, and decreasing
operating expenses. Ultimately, information systems need to aff ect the organization’s bot-
tom line (in a positive way!). To ensure that a real business need is being addressed, the
aff ected business organization (called the project sponsor), proposes the new systems devel-
opment project using a system request. Th e system request eff ectively kicks off the inception
1 For a very good comprehensive description of project management for information systems, see R.K. Wysocki,
Eff ective Project Management: Traditional, Agile, Extreme, 5th Ed. (Indianapolis, IN: Wiley Publishing, 2009).
Also, the Project Management Institute (www.pmi.org) and the Information Systems Community of Practice
of the Project Management Institute (is.vc.pmi.org) have valuable resources on information systems pro-
ject management. Finally, the following are good books on project management for object-oriented projects:
G. Booch, Object Solutions: Managing the Object-Oriented Project (Menlo Park, CA: Addison-Wesley, 1996); M.
R. Cantor, Object-Oriented Project Management with UML (New York: Wiley, 1998); A. Cockburn, Surviving
Object-Oriented Projects: A Manager’s Guide (Reading, MA: Addison-Wesley, 1998); I. Jacobson, G. Booch, and J.
Rumbaugh, Th e Unifi ed Soft ware Development Process (Reading, MA: Addison-Wesley, 1999); W. Royce, Soft ware
Project Management: A Unifi ed Framework (Reading, MA: Addison-Wesley, 1998).
Project Identifi cation 43
phase for the new systems development project. Th e request is forwarded to an approval
committee for consideration. Th e approval committee reviews the request and makes an
initial determination of whether to investigate the proposal or not. If the committee initially
approves the request, the systems development team gathers more information to determine
the feasibility of the project.
A feasibility analysis plays an important role in deciding whether to proceed with an
information systems development project. It examines the technical, economic, and organi-
zational pros and cons of developing the system, and it gives the organization a slightly more
detailed picture of the advantages of investing in the system as well as any obstacles that could
arise. In most cases, the project sponsor works closely with the development team to develop
the feasibility analysis. Once the feasibility analysis has been completed, it is submitted to
the approval committee, along with a revised system request. Th e committee then decides
whether to approve the project, decline the project, or table it until additional information is
available. Projects are selected by weighing risks and returns and by making trade-off s at the
organizational level.
Once the committee has approved a project, the development team must carefully plan
for the actual development of the system. Because we are following a Unifi ed Process-based
approach, the systems development workplan will evolve throughout the development pro-
cess. Given this evolutionary approach, one critical success factor for project management is
to start with a realistic assessment of the work that needs to be accomplished and then man-
age the project according to that assessment. Th is can be achieved by carefully creating and
managing the workplan, estimating the eff ort to develop the system, staffi ng the project, and
coordinating project activities.
In addition to covering the above material, this chapter also covers three traditional pro-
ject management tools that are very useful to manage object-oriented systems development
projects: work breakdown structures, Gantt charts, and network diagrams.
PROJECT IDENTIFICATION
A project is identifi ed when someone in the organization identifi es a business need to build
a system. Th is could occur within a business unit or IT, come from a steering committee
charged with identifying business opportunities, or evolve from a recommendation made
by external consultants. Examples of business needs include supporting a new marketing
campaign, reaching out to a new type of customer, or improving interactions with suppliers.
Sometimes, needs arise from some kind of “pain” within the organization, such as a drop in
market share, poor customer service levels, or increased competition. Other times, new busi-
ness initiatives and strategies are created, and a system is required to enable them.
Business needs also can surface when the organization identifi es unique and compet-
itive ways of using IT. Many organizations keep an eye on emerging technology, which is
technology that is still being developed and is not yet viable for widespread business use. For
example, if companies stay abreast of technology such as the augmented reality, games, smart
cards, and mobile devices, they can develop business strategies that leverage the capabilities of
these technologies and introduce them into the marketplace as a fi rst mover. Ideally, they can
take advantage of this fi rst-mover advantage by making money and continuing to innovate
while competitors trail behind.
Th e project sponsor is someone who recognizes the strong business need for a system and
has an interest in seeing the system succeed. He or she will work throughout the development
process to make sure that the project is moving in the right direction from the perspective of the
44 Chapter 2 Project Management
business. Th e project sponsor serves as the primary point of contact for the system. Usually, the
sponsor of the project is from a business function, such as marketing, accounting, or fi nance;
however, members of the IT area also can sponsor or cosponsor a project.
The size or scope of a project determines the kind of sponsor needed. A small
departmental system might require sponsorship from only a single manager, whereas a
large organizational initiative might need support from the entire senior management
team and even the CEO. If a project is purely technical in nature (e.g., improvements to
the existing IT infrastructure or research into the viability of an emerging technology),
then sponsorship from IT is appropriate. When projects have great importance to the
business yet are technically complex, joint sponsorship by both the business and IT may
be necessary.
Th e business need drives the high-level business requirements for the system. Requirements
are what the information system will do, or the functionality it will contain. Th ey need to be
explained at a high level so that the approval committee and, ultimately, the project team
understand what the business expects from the fi nal product. Business requirements are the
features and capabilities the information system will have to include, such as the ability to
collect customer orders online or the ability for suppliers to receive inventory information as
orders are placed and sales are made.
Th e project sponsor also should have an idea of the business value to be gained from the
system, both in tangible and intangible ways. Tangible value can be quantifi ed and measured
easily (e.g., 2 percent reduction in operating costs). An intangible value results from an intui-
tive belief that the system provides important, but hard-to-measure, benefi ts to the organiza-
tion (e.g., improved customer service or a better competitive position).
Once the project sponsor identifi es a project that meets an important business need and
he or she can identify the system’s business requirements and value, it is time to formally
initiate the project. In most organizations, project initiation begins with a document called a
system request.
System Request
A system request is a document that describes the business reasons for building a system
and the value that the system is expected to provide. Th e project sponsor usually completes
this form as part of a formal system project selection process within the organization. Most
system requests include fi ve elements: project sponsor, business need, business require-
ments, business value, and special issues. Th e sponsor describes the person who will serve as
the primary contact for the project, and the business need presents the reasons prompting
the project. Th e business requirements of the project refer to the business capabilities that the
system will need to have, and the business value describes the benefi ts that the organization
should expect from the system. Special issues are included on the document as a catch-all
for other information that should be considered in assessing the project. For example, the
project may need to be completed by a specifi c deadline. Project teams need to be aware of
any special circumstances that could aff ect the outcome of the system. Figure 2-1 shows a
template for a system request.
Th e completed system request is submitted to the approval committee for consideration.
Th is approval committee could be a company steering committee that meets regularly to
make information systems decisions, a senior executive who has control of organizational
resources, or any other decision-making body that governs the use of business investments.
Th e committee reviews the system request and makes an initial determination, based on the
information provided, of whether to investigate the proposal or not. If so, the next step is to
conduct a feasibility analysis.
Feasibility Analysis 45
FEASIBILITY ANALYSIS
Once the need for the system and its business requirements have been defi ned, it is time to
create a more detailed business case to better understand the opportunities and limitations
associated with the proposed project. Feasibility analysis guides the organization in determin-
ing whether or not to proceed with a project. Feasibility analysis also identifi es the important
risks associated with the project that must be addressed if the project is approved. As with the
system request, each organization has its own process and format for the feasibility analysis,
but most include three types: technical feasibility, economic feasibility, and organizational
feasibility. Th e results of these analyses are combined into a feasibility study, which is given to
the approval committee (see Figure 2-2).
Although we now discuss feasibility analysis within the context of initiating a project,
most project teams will revise their feasibility study throughout the development process and
revisit its contents at various checkpoints during the project. If at any point the project’s risks
and limitations outweigh its benefi ts, the project team may decide to cancel the project or
make necessary improvements.
Technical Feasibility
Th e fi rst type of feasibility analysis addresses the technical feasibility of the project: the extent
to which the system can be successfully designed, developed, and installed by the IT group.
System Request—Name of Project
Project Sponsor: Name of project sponsor
Business Need: Short description of business need
Business Requirements: Description of business requirements
Business Value: Expected value that the system will provide
Special Issues or Constraints: Any additional information that may be relevant
to the stakeholders
FIGURE 2-1
System Request
Template
Technical Feasibility: Can We Build It?
• Familiarity with Functional area: Less familiarity generates more risk
• Familiarity with Technology: Less familiarity generates more risk
• Project Size: Large projects have more risk
• Compatibility: The harder it is to integrate the system with the company’s existing
technology, the higher the risk
Economic Feasibility: Should We Build It?
• Development costs
• Annual operating costs
• Annual benefi ts (cost savings and revenues)
• Intangible costs and benefi ts
Organizational Feasibility: If We Build It, Will They Come?
• Is the project strategically aligned with the business?
• Project champion(s)
• Senior management
• Users
• Other stakeholders
FIGURE 2-2
Feasibility Analysis
Assessment Factors
46 Chapter 2 Project Management
Technical feasibility analysis is in essence a technical risk analysis that strives to answer this
question: Can we build it?2
Many risks can endanger the successful completion of a project. First is the users’ and
analysts’ lack of familiarity with the functional area. When analysts are unfamiliar with
the business functional area, they have a greater chance of misunderstanding the users
or of missing opportunities for improvement. Th e risk increases dramatically when the
users themselves are less familiar with an application, such as with the development of
a system to support a business innovation. In general, developing new systems is riskier
than producing extensions to an existing system because existing systems tend to be better
understood.
Familiarity with the technology is another important source of technical risk. When
a system uses technology that has not been used before within the organization, there is
a greater chance that problems will occur and delays will be incurred because of the need
to learn how to use the technology. Risk increases dramatically when the technology itself
is new.
Project size is an important consideration, whether measured as the number of people on
the development team, the length of time it will take to complete the project, or the number of
distinct features in the system. Larger projects present more risk, both because they are more
complicated to manage and because there is a greater chance that important system require-
ments will be overlooked or misunderstood. Furthermore, the extent to which the project is
highly integrated with other systems can cause problems because complexity increases when
many systems must work together.
Finally, project teams need to consider the compatibility of the new system with the
technology that already exists in the organization. Systems are rarely built in a vacuum—they
are built in organizations that already have numerous systems in place. New technology and
applications need to integrate with the existing environment for many reasons. Th ey might
rely on data from existing systems, they might produce data that feed other applications, and
they might have to use the company’s existing communications infrastructure.
Th e assessment of a project’s technical feasibility is not cut and dried because in many
cases, some interpretation of the underlying conditions is needed. One approach is to com-
pare the project under consideration with prior projects undertaken by the organization.
Another option is to consult with experienced IT professionals in the organization or exter-
nal IT consultants; oft en they are able to judge whether a project is feasible from a technical
perspective.
Economic Feasibility
Th e second element of a feasibility analysis is to perform an economic feasibility analy-
sis (also called a cost–benefi t analysis), which identifi es the fi nancial risk associated with
the project. It attempts to answer the question, Should we build the system? Economic
feasibility is determined by identifying costs and benefi ts associated with the system, assign-
ing values to them, and then calculating the cash fl ow and return on investment for the
project. Th e more expensive the project, the more rigorous and detailed the analysis should
be. Figure 2-3 lists the steps in performing a cost–benefi t analysis; each step is described in
the following sections.
2 We use build it in the broadest sense. Organizations can also choose to buy a commercial soft ware package and
install it, in which case, the question might be, Can we select the right package and successfully install it?
Feasibility Analysis 47
Identifying Costs and Benefi ts Th e fi rst task when developing an economic feasibility anal-
ysis is to identify the kinds of costs and benefi ts the system will have and list them along the
left -hand column of a spreadsheet. Figure 2-4 lists examples of costs and benefi ts that may
be included.
1. Identifi ng Costs and Benefi ts List the tangible costs and benefi ts for the project.
Include both one-time and recurring costs.
2. Assigning Values to Costs and Benefi ts Work with business users and IT professionals to
create numbers for each of the costs and benefi ts.
Even intangibles should be valued if at all possible.
3. Determining Cash Flow Project what the costs and benefi ts will be over a
period of time, usually three to fi ve years. Apply a
growth rate to the numbers, if necessary.
4. Determining Net Present Value (NPV) Calculate what the value of future costs and ben-
efi ts are if measured by today’s standards. You will
need to select a rate of growth to apply the NPV
formula.
5. Determining Return on Investment (ROI) Calculate how much money the organization will
receive in return for the investment it will make
using the ROI formula.
6. Determining the Break-Even Point Find the fi rst year in which the system has greater
benefi ts than costs. Apply the break-even formula
using fi gures from that year. This will help you
understand how long it will take before the system
creates real value for the organization.
7. Graphing the Break-Even Point Plot the yearly costs and benefi ts on a line graph.
The point at which the lines cross is the break-even
point.
FIGURE 2-3
Steps for Con-
ducting Economic
Feasibility
FIGURE 2-4
Example Costs and
Benefi ts for Eco-
nomic Feasibility
Development Team Salaries Software Upgrades
Consultant Fees Software Licensing Fees
Development Training Hardware Repairs
Hardware and Software Hardware Upgrades
Vendor Installation Operational Team Salaries
Offi ce Space and Equipment Communications Charges
Data Conversion Costs User Training
Increased Sales Increased Market Share
Reductions in Staff Increased Brand Recognition
Reductions in Inventory Higher Quality Products
Reductions in IT Costs Improved Customer Service
Better Supplier Prices Better Supplier Relations
Development Costs Operational Costs
Tangible Benefi ts Intangible Benefi ts
48 Chapter 2 Project Management
Costs and benefi ts can be broken down into four categories: development costs, oper-
ational costs, tangible benefi ts, and intangibles. Development costs are tangible expenses
incurred during the construction of the system, such as salaries for the project team, hard-
ware and soft ware expenses, consultant fees, training, and offi ce space and equipment.
Development costs are usually thought of as one-time costs. Operational costs are tangible
costs required to operate the system, such as the salaries for operations staff , soft ware licens-
ing fees, equipment upgrades, and communications charges. Operational costs are usually
thought of as ongoing costs.
Revenues and cost savings are the tangible benefi ts the system enables the organization
to collect or the tangible expenses the system enables the organization to avoid. Tangible
benefi ts could include increased sales, reductions in staff , and reductions in inventory. Of
course, a project also can aff ect the organization’s bottom line by reaping intangible benefi ts
or incurring intangible costs. Intangible costs and benefi ts are more diffi cult to incorporate
into the economic feasibility because they are based on intuition and belief rather than
“hard numbers.” Nonetheless, they should be listed in the spreadsheet along with the tan-
gible items.
Assigning Values to Costs and Benefi ts Once the types of costs and benefi ts have been
identifi ed, analysts assign specifi c dollar values to them. Th is might seem impossible; how
can someone quantify costs and benefi ts that haven’t happened yet? And how can those
predictions be realistic? Although this task is very diffi cult, analysts have to do the best they
can to come up with reasonable numbers for all the costs and benefi ts. Only then can the
approval committee make an educated decision about whether or not to move ahead with
the project.
Th e best strategy for estimating costs and benefi ts is to rely on the people who have the
clearest understanding of them. For example, costs and benefi ts related to the technology or
the project itself can be provided by the company’s IT group or external consultants, and
business users can develop the numbers associated with the business (e.g., sales projections,
order levels). Analysts can also consider past projects, industry reports, and vendor infor-
mation, although these approaches probably will be a bit less accurate. All the estimates will
probably be revised as the project proceeds.
Sometimes it is acceptable for analysts to list intangible benefi ts, such as improved
customer service, without assigning a dollar value, whereas other times they have to
make estimates regarding the value of an intangible benefi t. If at all possible, they should
quantify intangible costs or benefi ts. Otherwise, it will not be apparent whether the costs
and benefi ts have been realized. Consider a system that is supposed to improve customer
service. Th is is intangible, but assume that the greater customer service will decrease the
number of customer complaints by 10 percent each year over three years and that $200,000
is spent on phone charges and phone operators who handle complaint calls. Suddenly
there are some very tangible numbers with which to set goals and measure the original
intangible benefi t.
Figure 2-5 shows costs and benefi ts along with assigned dollar values. Notice that the
customer service intangible benefi t has been quantifi ed based on fewer customer complaint
phone calls. Th e intangible benefi t of being able to off er services that competitors currently
off er was not quantifi ed, but it was listed so that the approval committee will consider the
benefi t when assessing the system’s economic feasibility.
Determining Cash Flow A formal cost–benefi t analysis usually contains costs and benefi ts
over a selected number of years (usually three to fi ve years) to show cash fl ow over time
Feasibility Analysis 49
(see Figure 2-6). When using this cash-fl ow method, the years are listed across the top of the
spreadsheet to represent the time period for analysis, and numeric values are entered in the
appropriate cells within the spreadsheet’s body. Sometimes fi xed amounts are entered into
the columns. For example, Figure 2-6 lists the same amount for customer complaint calls and
inventory costs for all fi ve years. Usually amounts are augmented by some rate of growth to
adjust for infl ation or business improvements, as shown by the 6 percent increase that is
added to the sales numbers in the sample spreadsheet. Finally, totals are added to determine
what the overall benefi ts will be; the higher the overall total, the greater the economic feasi-
bility of the solution.
Determining Net Present Value and Return on Investment Th ere are several problems
with the cash-fl ow method—(1) it does not consider the time value of money (i.e., a dollar
today is not worth a dollar tomorrow), and (2) it does not show the overall “bang for the buck”
that the organization is receiving from its investment. Th erefore, some project teams add
additional calculations to the spreadsheet to provide the approval committee with a more-
accurate picture of the project’s worth.
Net present value (NPV) is used to compare the present value of future cash fl ows with
the investment outlay required to implement the project. For example, if you have a friend
who owes you a dollar today but instead gives you a dollar three years from now, you’ve been
had! Given a 10 percent increase in value, you’ll be receiving the equivalent of 75 cents in
today’s terms.
NPV can be calculated in many diff erent ways, some of which are extremely complex.
Figure 2-7 shows a basic calculation that can be used in your cash fl ow analysis to get more
Benefi tsa
Increased sales 500,000
Improved customer serviceb 70,000
Reduced inventory costs 68,000
Total benefi ts 638,000
Development costs
2 servers @ $125,000 250,000
Printer 100,000
Software licenses 34,825
Server software 10,945
Development labor 1,236,525
Total development costs 1,632,295
Operational costs
Hardware 54,000
Software 20,000
Operational labor 111,788
Total operational costs 185,788
Total costs 1,818,083
a An important yet intangible benefi t will be the ability to offer
services that our competitors currently offer.
b Customer service numbers have been based on reduced costs for
customer complaint phone calls.
FIGURE 2-5
Assigning Values to
Costs and Benefi ts
relevant values. In Figure 2-6, the present value of the costs and benefi ts are calculated fi rst
(i.e., they are shown at a discounted rate). Th en, net present value is calculated, and it shows
the discounted rate of the combined costs and benefi ts.
Th e return on investment (ROI) is a calculation listed somewhere on the spreadsheet that
measures the amount of money an organization receives in return for the money it spends.
A high ROI results when benefi ts far outweigh costs. ROI is determined by fi nding the total
benefi ts less the costs of the system and dividing that number by the total costs of the system
(see Figure 2-7). ROI can be determined per year or for the entire project over a period of
time. One drawback of ROI is that it considers only the end points of the investment, not the
cash fl ow in between, so it should not be used as the sole indicator of a project’s worth. Th e
spreadsheet in Figure 2-6 shows an ROI fi gure.
Determining the Break-Even Point If the project team needs to perform a rigorous cost–
benefi t analysis, it might need to include information about the length of time before the
project will break even, or when the returns will match the amount invested in the project.
FIGURE 2-6 Cost–Benefi t Analysis
Increased sales 500,000 530,000 561,800 595,508 631,238
Reduction in customer complaint calls 70,000 70,000 70,000 70,000 70,000
Reduced inventory costs 68,000 68,000 68,000 68,000 68,000
TOTAL BENEFITS: 638,000 668,000 699,800 733,508 769,238
PV OF BENEFITS: 619,417 629,654 640,416 651,712 663,552 3,204,752
PV OF ALL BENEFITS: 619,417 1,249,072 1,889,488 2,541,200 3,204,752
2 Servers @ $125,000 250,000 0 0 0 0
Printer 100,000 0 0 0 0
Software licenses 34,825 0 0 0 0
Server software 10,945 0 0 0 0
Development labor 1,236,525 0 0 0 0
TOTAL DEVELOPMENT COSTS: 1,632,295 0 0 0 0
Hardware 54,000 81,261 81,261 81,261 81,261
Software 20,000 20,000 20,000 20,000 20,000
Operational labor 111,788 116,260 120,910 125,746 130,776
TOTAL OPERATIONAL COSTS: 185,788 217,521 222,171 227,007 232,037
TOTAL COSTS: 1,818,083 217,521 222,171 227,007 232,037
PV OF COSTS: 1,765,129 205,034 203,318 201,693 200,157 2,575,331
PV OF ALL COSTS: 1,765,129 1,970,163 2,173,481 2,375,174 2,575,331
TOTAL PROJECT BENEFITS COSTS: (1,180,083) 450,479 477,629 506,501 537,201
YEARLY NPV: (1,145,712) 424,620 437,098 450,019 463,395 629,421
CUMULATIVE NPV: (1,145,712) (721,091) (283,993) 166,026 629,421
RETURN ON INVESTMENT: 24.44% (629,421/2,575,331)
BREAK-EVEN POINT: 3.63 years [break-even occurs in year 4; (450,019 2 166,026)/450,019 5 0.63]
INTANGIBLE BENEFITS: This service is currently provided by competitors
Improved customer satisfaction
2015 2016 2017 2018 2019 Total
50 Chapter 2 Project Management
Th e greater the time it takes to break even, the riskier the project. Th e break-even point is
determined by looking at the cash fl ow over time and identifying the year in which the ben-
efi ts are larger than the costs (see Figure 2-6). Th en, the diff erence between the yearly and
cumulative NPV for that year is divided by the yearly NPV to determine how far into the year
the break-even point will occur. See Figure 2-7 for the break-even calculation. Th e break-even
point also can be depicted graphically, as shown in Figure 2-8. Th e cumulative present value
of the costs and benefi ts for each year is plotted on a line graph; the point at which the lines
cross is the break-even point.
Organizational Feasibility
Th e fi nal type of feasibility analysis is to assess the organizational feasibility of the system,
how well the system ultimately will be accepted by its users and incorporated into the ongo-
ing operations of the organization. Th ere are many organizational factors that can have an
Present Value (PV) The amount of an investment today Amount
compared to that same amount in the future,
taking into account infl ation and time.
(1 1 interest rate)n
n 5 number of years in future
Net Present Value (NPV) The present value of benefi t less the present PV Benefi ts 2 PV Costs
value of costs.
Return on Investment (ROI) The amount of revenues or cost savings results Total benefi ts 2 Total costs
from a given investment.
Total costs
Break-Even Point The point in time at which the costs of the Yearly NPV* 2 Cumulative NPV
project equal the value it has delivered.
Yearly NPV*
*Use the Yearly NPV amount from the fi rst year in which
the project has a positive cash fl ow.
Add the above amount to the year in which the project
has a positive cash fl ow.
Calculation Defi nition Formula
FIGURE 2-7 Financial Calculations Used for Cost–Benefi t Analysis
Break-even point
1 2 3 54
0
500,000
1,000,000
1,500,000
2,000,000
D
ol
la
rs
2,500,000
3,500,000
Years
3,000,000
Costs
Benefits
FIGURE 2-8
Break-Even Graph
Feasibility Analysis 51
eff ect on the project, and seasoned developers know that organizational feasibility can be the
most diffi cult feasibility dimension to assess. In essence, an organizational feasibility analysis
attempts to answer the question, If we build it, will they come?
One way to assess the organizational feasibility of the project is to understand how well
the goals of the project align with business objectives. Strategic alignment is the fi t between
the project and business strategy—the greater the alignment, the less risky the project will
be from an organizational feasibility perspective. For example, if the marketing department
has decided to become more customer focused, then a CRM project that produces integrated
customer information would have strong strategic alignment with marketing’s goal. Many IT
projects fail when the IT department initiates them, because there is little or no alignment
with business unit or organizational strategies.
A second way to assess organizational feasibility is to conduct a stakeholder analysis.3 A
stakeholder is a person, group, or organization that can aff ect (or will be aff ected by) a new
system. In general, the most important stakeholders in the introduction of a new system are
the project champion, system users, and organizational management (see Figure 2-9), but
systems sometimes aff ect other stakeholders as well. For example, the IS department can
be a stakeholder of a system because IS jobs or roles may be changed signifi cantly aft er its
implementation.
Th e champion is a high-level, non–information systems executive who is usually the
project sponsor who created the system request. Th e champion supports the project with
time, resources (e.g., money), and political support within the organization by communicat-
ing the importance of the system to other organizational decision makers. More than one
champion is preferable because if the champion leaves the organization, the support could
leave as well.
Whereas champions provide day-to-day support for the system, organizational manage-
ment support conveys to the rest of the organization the belief that the system will make a
Champion A champion: • Make a presentation about the objectives of the
• Initiates the project project and the proposed benefi ts to those executives
• Promotes the project who will benefi t directly from the system
• Allocates his or her time to project • Create a prototype of the system to demonstrate its
• Provides resources potential value
Organizational Organizational managers: • Make a presentation to management about the
Management • Know about the project objectives of the project and the proposed benefi ts
• Budget enough money for the project • Market the benefi ts of the system using memos and
• Encourage users to accept and use the system organizational newsletters
• Encourage the champion to talk about the project
with his or her peers
System Users Users: • Assign users offi cial roles on the project team
• Make decisions that infl uence the project • Assign users specifi c tasks to perform with clear
• Perform hands-on activities for the project deadlines
• Ultimately determine whether the project is • Ask for regular feedback from users (e.g., at
successful by using or not using the system weekly meetings)
Role Techniques for Improvement
FIGURE 2-9 Some Important Stakeholders for Organizational Feasibility
52 Chapter 2 Project Management
3 A good book that presents a series of stakeholder analysis techniques is R. O. Mason and I. I. Mittroff , Challenging
Strategic Planning Assumptions: Th eory, Cases, and Techniques (New York: Wiley, 1981).
valuable contribution and that necessary resources will be made available. Ideally, manage-
ment should encourage people in the organization to use the system and to accept the many
changes that the system will likely create.
A third important group of stakeholders are the system users who ultimately use the
system once it has been installed in the organization. Too oft en, the project team meets
with users at the beginning of a project and then disappears until aft er the system is created.
In this situation, rarely does the fi nal product meet the expectations and needs of those
who are supposed to use it because needs change and users become savvier as the project
progresses. User participation should be promoted throughout the development process by
getting users involved in the development of the system (e.g., performing tasks, providing
feedback, making decisions).
Finally, the feasibility study helps organizations make wiser investments by forcing pro-
ject teams to consider technical, economic, and organizational factors that can aff ect their
projects. It protects IT professionals from criticism by keeping the business units educated
about decisions and positioned as the leaders in the decision-making process. Remember, the
feasibility study should be revised several times during the project at points where the project
team makes critical decisions about the system (e.g., before each iteration of the development
process).
PROJECT SELECTION
Once the feasibility analysis has been completed, it is submitted to the approval committee,
along with a revised system request. Th e committee then decides whether to approve the
project, decline the project, or table it until additional information is available. At the pro-
ject level, the committee considers the value of the project by examining the business need
(found in the system request) and the risks of building the system (presented in the feasibility
analysis).
Before approving the project, however, the committee also considers the project from an
organizational perspective; it has to keep in mind the company’s entire portfolio of projects.
Th is way of managing projects is called portfolio management. Portfolio management takes
into consideration the diff erent kinds of projects that exist in an organization—large and
small, high risk and low risk, strategic and tactical. (See Figure 2-10 for the diff erent ways of
Size What is the size? How many people are needed to work on the
project?
Cost How much will the project cost the organization?
Purpose What is the purpose of the project? Is it meant to improve the
technical infrastructure? Support a current business strategy?
Improve operations? Demonstrate a new innovation?
Length How long will the project take before completion? How much
time will go by before value is delivered to the business?
Risk How likely is it that the project will succeed or fail?
Scope How much of the organization is affected by the system? A
department? A division? The entire corporation?
Return on investment How much money does the organization expect to receive in
return for the amount the project costs?
FIGURE 2-10
Ways to Classify
Projects
Project Selection 53
classifying projects.) A good project portfolio has the most appropriate mix of projects for the
organization’s needs. Th e committee acts as portfolio manager with the goal of maximizing
the cost–benefi t performance and other important factors of the projects in their portfolio.
For example, an organization might want to keep high-risk projects to less than 20 percent of
its total project portfolio.
Th e approval committee must be selective about where to allocate resources. Th is
involves trade-off s in which the organization must give up something in return for something
else to keep its portfolio well balanced. If there are three potentially high-payoff projects, yet
all have very high risk, then perhaps only one of the projects will be selected. Also, there are
times when a system at the project level makes good business sense, but it does not make
sense at the organization level. Th us, a project may show a very strong ROI and support
important business needs for a part of the company, but it is not selected. Th is could happen
for many reasons—because there is no money in the budget for another system, the organ-
ization is about to go through some kind of change (e.g., a merger), projects that meet the
same business requirements already are under way, or the system does not align well with the
current or future corporate strategy.
TRADITIONAL PROJECT MANAGEMENT TOOLS
Before we get to actually creating a workplan that is suitable to manage and control an
object-oriented systems development project, we need to introduce a set of project man-
agement tools that have been used to successfully manage traditional soft ware development
projects (and many other types of projects): a work-breakdown structure, a Gantt chart, and
a network diagram. To begin with, we must fi rst understand what a task is. A task is a unit
of work that will be performed by a member or members of the development team, such as
feasibility analysis. Each task is described by information such as its name, start and com-
pletion dates, person assigned to complete the task, deliverables, completion status, priority,
resources needed, estimated time to complete the task, and the actual time it took to complete
the task (see Figure 2-11). Th e fi rst thing a project manager must do is to identify the tasks
that need to be accomplished and determine how long each task will take. Tasks and their
identifi cation and documentation are the basis of all three of these tools. Once the tasks have
been identifi ed and documented, they are organized within a work breakdown structure that
is used to drive the creation of Gantt charts and network diagrams that can be used to graphi-
cally portray a traditional workplan. Th ese techniques help a project manager understand and
manage the project’s progress over time.
Name of the task Perform economic feasibility
Start date Jan 05, 2015
Completion date Jan 19, 2015
Person assigned to the task Project sponsor: Mary Smith
Deliverable(s) Cost-benefi t analysis
Completion status Open
Priority High
Resources that are needed Spreadsheet software
Estimated time 16 hours
Actual time 14.5 hours
Workplan Information Example
FIGURE 2-11
Task Information
54 Chapter 2 Project Management
Work Breakdown Structures
A project manager can use a structured, top-down approach whereby high-level tasks are fi rst
defi ned and then broken down into subtasks. For example, Figure 2-12 shows a list of high-
level tasks needed to implement a new IT training class. Some of the main steps in the process
include identifying vendors, creating and administering a survey, and building new class-
rooms. Each step is then broken down in turn and numbered in a hierarchical fashion. Th ere
are eight subtasks (i.e., 7.1–7.8) for creating and administering a survey, and there are three
subtasks (7.2.1–7.2.3) that make up the review initial survey task. A list of tasks hierarchically
numbered in this way is called a work breakdown structure (WBS). Th e number of tasks and
level of detail depend on the complexity and size of the project. At a minimum, the WBS must
include the duration of the task, the current status of the task (i.e., open, complete), and the
task dependencies, which occur when one task cannot be performed until another task is com-
pleted. For example, Figure 2-12 shows that incorporating changes to the survey (task 7.4)
takes a week to perform, but it cannot occur until aft er the survey is reviewed (task 7.2) and
pilot tested (task 7.3). Key milestones, or important dates, are also identifi ed on the workplan.
Th ere are two basic approaches to organizing a traditional WBS: by development phase
or by product. For example, if a fi rm decided that it needed to develop a website, the fi rm
could create a WBS based on the inception, elaboration, construction, and transition
phases of the Unifi ed Process. In this case, a typical task that would take place during incep-
tion would be feasibility analysis. Th is task would be broken down into the diff erent types of
feasibility analysis: technical, economic, and organizational. Each of these would be further
broken down into a set of subtasks. Alternatively, the fi rm could organize the workplan
along the lines of the diff erent products to be developed. For example, in the case of a web-
site, the products could include applets, application servers, database servers, the various sets
of Web pages to be designed, a site map, and so on. Th en these would be further decomposed
1 Identify vendors 2 Complete
2 Review training materials 6 1 Complete
3 Compare vendors 2 2 In Progress
4 Negotiate with vendors 3 3 Open
5 Develop communications information 4 1 In Progress
6 Disseminate information 2 5 Open
7 Create and administer survey 4 6 Open
7.1 Create initial survey 1 Open
7.2 Review initial survey 1 7.1 Open
7.2.1 Review by Director of IT Training 1 Open
7.2.2 Review by Project Sponsor 1 Open
7.2.3 Review by Representative Trainee 1 Open
7.3 Pilot test initial survey 1 7.1 Open
7.4 Incorporate survey changes 1 7.2, 7.3 Open
7.5 Create distribution list 0.5 Open
7.6 Send survey to distribution list 0.5 7.4, 7.5 Open
7.7 Send follow-up message 0.5 7.6 Open
7.8 Collect completed surveys 1 7.6 Open
8 Analyze results and choose vendor 2 4, 7 Open
9 Build new classrooms 11 1 In Progress
10 Develop course options 3 8, 9 Open
Task Duration
Number Task Name (in weeks) Dependency Status
FIGURE 2-12
Work Breakdown
Structure
Traditional Project Management Tools 55
into the diff erent tasks associated with the phases of the development process. Either way,
once the overall structure is determined, tasks are identifi ed and included in the WBS. We
return to the topic of WBSs and their use in iterative planning later in this chapter.
Gantt Chart
A Gantt chart is a horizontal bar chart that shows the same task information as the
project WBS but in a graphical way. Sometimes a picture really is worth a thousand words,
and the Gantt chart can communicate the high-level status of a project much faster and
easier than the WBS. Creating a Gantt chart is simple and can be done using a spreadsheet
package, graphics soft ware, or a project management package.
First, tasks are listed as rows in the chart, and time is listed across the top in increments
based on the needs of the projects (see Figure 2-13). A short project may be divided into
ID
1
2
3
4
5
Identify
vendors
Review
training
materials
Compare
vendors
Negotiate
with
vendors
Develop
communications
information
Disseminate
information
Create and
administer
survey
Analyze results
and choose
Build new
classroom
Develop
course
options
Budget
Meeting
Software
Installation
6
7
8
9
10
11
12
2 wks Wed
1/1/15
Wed
1/1/15
Wed
2/12/15
Wed
2/26/15
Wed
1/15/15
Wed
2/12/15
Wed
2/26/15
Wed
3/26/15
Wed
1/15/15
Wed
4/9/15
Wed
1/15/15
Tue
4/1/15
6 wks
Barbara
Barbara
Barbara
Alan
Alan
Alan
Alan
Alan
David
D
2 wks
3 wks
4 wks
2 wks
4 wks
2 wks
11 wks
3 wks
2
3
1
5
6
4, 7
1
8, 9
1 day
1 day
Task
Name Duration Start
Tue
1/14/15
Tue
2/11/15
Tue
2/25/15
Tue
3/8/15
Tue
2/11/15
Tue
2/25/15
Tue
3/25/15
Tue
4/8/15
Tue
4/1/15
Tue
4/29/15
Wed
1/15/15
Tue
4/1/15
Finish
12/29 1/5 1/12
1/15
4/1
1/19 1/26 2/2 2/9 2/16 2/23 3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20 4/27
Prede
January February March April M
FIGURE 2-13 Gantt Chart
56 Chapter 2 Project Management
hours or days, whereas a medium-sized project may be represented using weeks or months.
Horizontal bars are drawn to represent the duration of each task; the bar’s beginning and
end mark exactly when the task will begin and end. As people work on tasks, the appropriate
bars are fi lled in proportionately to how much of the task is fi nished. Too many tasks on a
Gantt chart can become confusing, so it’s best to limit the number of tasks to around twenty
or thirty. If there are more tasks, break them down into subtasks and create Gantt charts for
each level of detail.
Th ere are many things a project manager can see quickly by looking at a Gantt chart. In
addition to seeing how long tasks are and how far along they are, the project manager also can
tell which tasks are sequential, which tasks occur at the same time, and which tasks overlap
in some way. He or she can get a quick view of tasks that are ahead of schedule and behind
schedule by drawing a vertical line on today’s date. If a bar is not fi lled in and is to the left of
the line, that task is behind schedule.
Th ere are a few special notations that can be placed on a Gantt chart. Project mile-
stones are shown using upside-down triangles or diamonds. Arrows are drawn between
the task bars to show task dependencies. Sometimes, the names of people assigned to each
task are listed next to the task bars to show what human resources have been allocated to
the tasks.
Network Diagram
A second graphical way to look at project workplan information is the network diagram that
lays out the project tasks in a fl owchart (see Figure 2-14).
Program Evaluation and Review Technique (PERT) is a network analysis technique that
can be used when the individual task time estimates are fairly uncertain. Instead of simply
putting a point estimate for the duration estimate, PERT uses three time estimates: optimistic,
Software installation
12
Tue 4/1/15
1 day Tue
Tue 4/1/15
Budget meeting
11
Wed 1/15/15
1 day Wed
Wed 1/15/15
Identify vendors
1
Wed 1/1/15
2 wks Tue
Tue 1/14/15
Build new classroom
9
Wed 1/15/15
11 wks Tue
Tue 4/1/15
Compare vendors
3
Wed 2/12/15
2 wks Tue
Tue 2/25/15
Negotiate with vendors
4
Wed 2/26/15
3 wks Tue
Tue 3/18/15
Review training materials
2
Wed 1/1/15
6 wks Tue
Tue 2/11/15
Develop communications
Information
5
Wed 1/15/15
4 wks Tue
Tue 2/11/15
Disseminate information
6
Wed 2/12/15
2 wks Tue
Tue 2/25/15
Create and administer
survey
7
Wed 2/26/15
4 wks Tue
Tue 3/25/15
Analyze results and
choose vendor
8
Wed 3/26/15
2 wks Tue
Tue 4/8/15
Develop course options
10
Wed 4/9/15
3 wks Tue
Tue 4/29/15
FIGURE 2-14 Network Diagram
Traditional Project Management Tools 57
most likely, and a pessimistic. It then combines the three estimates into a single weighted
average estimate using the following formula:
PERT weighted average 5
optimistic estimate 1 (4 * most likely estimate)
1 pessimistic estimate
6
Th e network diagram is drawn as a node-and-arc type of graph that shows time estimates in
the nodes and task dependencies on the arcs. Each node represents an individual task, and a
line connecting two nodes represents the dependency between two tasks. Partially completed
tasks are usually displayed with a diagonal line through the node, and completed tasks con-
tain crossed lines.
Network diagrams are the best way to communicate task dependencies because they lay
out the tasks in the order in which they need to be completed. Th e critical path method (CPM)
simply allows the identifi cation of the critical path in the network. Th e critical path is the longest
path from the project inception to completion. Th e critical path shows all the tasks that must be
completed on schedule for a project as a whole to fi nish on schedule. If any tasks on the critical
path take longer than expected, the entire project will fall behind. Each task on the critical path
is a critical task, and they are usually depicted in a unique way; in Figure 2-14 they are shown
with double borders (see tasks 5, 6, 7, 8, and 10). CPM can be used with or without PERT.
PROJECT EFFORT ESTIMATION
Th e science (or art) of project management is in making trade-off s among three important con-
cepts: the functionality of the system, the time to complete the project (when the project will be
fi nished), and the cost of the project. Th ink of these three things as interdependent levers that
the project manager controls throughout the development of the system. Whenever one lever is
pulled, the other two levers are aff ected in some way. For example, if a project manager needs to
readjust a deadline to an earlier date, then the only solutions are to decrease the functionality of
the system or to increase costs by adding more people or having them work overtime. Oft en, a
project manager has to work with the project sponsor to change the goals of the project, such as
developing a system with less functionality or extending the deadline for the fi nal system, so that
the project has reasonable goals that can be met. In the beginning of the project, the manager
needs to estimate each of these levers and then continuously assess how to roll out the project
in a way that meets the organization’s needs. Estimation is the process of assigning projected
values for time and eff ort. Th e estimates developed at the start of a project are usually based on a
range of possible values and gradually become more specifi c as the project moves forward. Th at
is, the range of values for the inception phase will be much greater than for the transition phase.
Th e numbers used to calculate these estimates can be taken from projects with similar
tasks and technologies or provided by experienced developers. Generally speaking, the num-
bers should be conservative. A good practice is to keep track of the actual values for time and
eff ort during the development process so that numbers can be refi ned along the way and the
next project can benefi t from real data.
Th ere are a variety of ways to estimate the time required to build a system. Because the
Unifi ed Process is use-case driven, we use an approach that is based on use cases: use-case
points.4 Use-case points, originally developed by Gustav Karner of Objectory AB,5 are based
4 Th e material in this section is based on descriptions of use-case points contained in Raul R. Reed, Jr., Developing
Applications with Java and UML (Reading, MA: Addison-Wesley, 2002); Geri Schneider and Jason P. Winters, Apply-
ing Use Cases: A Practical Guide (Reading, MA: Addison-Wesley, 1998); Kirsten Ribu, “Estimating Object-Oriented
Soft ware Projects with Use Cases” (Master’s thesis, University of Oslo, 2001).
5 Objectory AB was acquired by Rational in 1995 and Rational is now part of IBM.
58 Chapter 2 Project Management
on unique features of use cases and object orientation. From a practical point of view, to
estimate eff ort using use-case points, the use cases and the use-case diagram must have
been created.6
Use-case models have two primary constructs: actors and use cases. An actor repre-
sents a role that a user of the system plays, not a specifi c user. For example, a role could be
secretary or manager. Actors can also represent other systems that will interact with the
system under development. For use-case point estimation purposes, actors can be classifi ed
as simple, average, or complex. Simple actors are separate systems with which the current
system must communicate through a well-defi ned application program interface (API).
Average actors are separate systems that interact with the current system using standard
communication protocols, such as TCP/IP, FTP, or HTTP, or an external database that
can be accessed using standard SQL. Complex actors are typically end users commu-
nicating with the system. Once all of the actors have been categorized as being simple,
average, or complex, the project manager counts the number of actors in each category
and enters the values into the unadjusted actor-weighting table contained in the use-case
point– estimation worksheet (see Figure 2-15). Th e project manager then computes the
Unadjusted Actor Weight Total (UAW). Th is is computed by summing the individual
results that were computed by multiplying the weighting factor by the number of actors
of each type. For example, if we assume that the use-case diagram has zero simple, zero
average, and four complex actors that interact with the system being developed, the UAW
will equal 12 (see Figure 2-16).
A use case represents a major business process that the system will perform that benefi ts
the actor(s) in some manner. Depending on the number of unique transactions that the use
case must address, a use case can be categorized as being simple, average, or complex. A use
case is classifi ed as simple if it supports one to three transactions, as average if it supports four
to seven transactions, or as complex if it supports more than seven transactions. Once all of
the use cases have been successfully categorized, the project manager enters the number of
each type of use case into the unadjusted use-case weighting table contained in the use-case
point–estimation worksheet (see Figure 2-15). By multiplying by the appropriate weights and
summing the results, we get the value for the unadjusted use-case weight total (UUCW). For
example, if we assume that we have three simple use cases, four average use cases, and one
complex use case, the value for the unadjusted use-case weight total is 70 (see Figure 2-16).
Next, the project manager computes the value of the unadjusted use-case points (UUCP) by
simply summing the unadjusted actor weight total and the unadjusted use-case weight total.
In this case the value of the UUCP equals 82 (see Figure 2-16).
Use-case point-based estimation also has a set of factors that are used to adjust the
use-case point value. In this case, there are two sets of factors: technical complexity factors
(TCFs) and environmental factors (EFs). Th ere are thirteen separate technical factors and
eight separate environmental factors. Th e purpose of these factors is to allow the project
as a whole to be evaluated for the complexity of the system being developed and the expe-
rience levels of the development staff , respectively. Obviously, these types of factors can
aff ect the eff ort that a team requires to develop a system. Each of these factors is assigned
a value between 0 and 5, 0 indicating that the factor is irrelevant to the system under con-
sideration and 5 indicating that the factor is essential for the system to be successful. Th e
assigned values are then multiplied by their respective weights. Th ese weighted values are
then summed up to create a technical factor value (TFactor) and an environmental factor
value (EFactor) (see Figure 2-15).
6 We cover the details of use-case modeling in Chapter 4.
Project Effort Estimation 59
FIGURE 2-15 Use-Case Point–Estimation Worksheet
Unadjusted Actor Weighting Table:
Actor Type Description Weighting Factor Number Result
Simple External System with well-defi ned API 1
Average External System using a protocol-based 2
interface, e.g., HTTP, TCT/IP, or a database
Complex Human 3
Unadjusted Actor Weight Total (UAW)
Unadjusted Use Case Weighting Table:
Use-Case Type Description Weighting Factor Number Result
Simple 1–3 transactions 5
Average 4–7 transactions 10
Complex >7 transactions 15
Unadjusted Use-Case Weight Total (UUCW)
Unadjusted Use Case Points (UUCP) 5 UAW 1 UUCW
Technical Complexity Factors:
Factor Number Description Weight Assigned Value (0–5) Weighted Value Notes
T1 Distributed system 2.0
T2 Response time or throughput 1.0
performance objectives
T3 End-user online effi ciency 1.0
T4 Complex internal processing 1.0
T5 Reusability of code 1.0
T6 Ease of installation 0.5
T7 Ease of use 0.5
T8 Portability 2.0
T9 Ease of change 1.0
T10 Concurrency 1.0
T11 Special security objectives included 1.0
T12 Direct access for third parties 1.0
T13 Special user training required 1.0
Technical Factor Value (TFactor)
Technical Complexity Factor (TCF) 5 0.6 1 (0.01 * TFactor)
Environmental Factors:
Factor Number Description Weight Assigned Value (0–5) Weighted Value Notes
E1 Familiarity with system 1.5
development process being used
E2 Application experience 0.5
E3 Object-oriented experience 1.0
E4 Lead analyst capability 0.5
E5 Motivation 1.0
E6 Requirements stability 2.0
E7 Part time staff –1.0
E8 Diffi culty of programming language –1.0
Environmental Factor Value (EFactor)
Environmental Factor (EF) 5 1.4 1 (20.03 * EFactor)
Adjusted Use Case Points (UCP) 5 UUCP * TCF * ECF
Effort in Person Hours 5 UCP * PHM
TEMPLATE
can be found at
www.wiley.com
/college/dennis
60 Chapter 2 Project Management
FIGURE 2-16 Use-Case Point Estimation for the Appointment System
Unadjusted Actor Weighting Table:
Actor Type Description Weighting Factor Number Result
Simple External system with well-defi ned API 1 0 0
Average External system using a protocol-based 2 0 0
interface, e.g., HTTP, TCT/IP, or a database
Complex Human 3 4 12
Unadjusted Actor Weight Total (UAW) 12
Unadjusted Use-Case Weighting Table:
Use Case Type Description Weighting Factor Number Result
Simple 1–3 transactions 5 3 15
Average 4–7 transactions 10 4 40
Complex >7 transactions 15 1 15
Unadjusted Use Case Weight Total (UUCW) 70
Unadjusted Use-Case Points (UUCP) 5 UAW 1 UUCW 82 5 12 1 70
Technical Complexity Factors:
Factor Number Description Weight Assigned Value (0–5) Weighted Value Notes
T1 Distributed system 2.0 0 0
T2 Response time or throughput 1.0 5 5
performance objectives
T3 End-user online effi ciency 1.0 3 3
T4 Complex internal processing 1.0 1 1
T5 Reusability of code 1.0 1 1
T6 Ease of installation 0.5 2 1
T7 Ease of use 0.5 4 2
T8 Portability 2.0 0 0
T9 Ease of change 1.0 2 2
T10 Concurrency 1.0 0 0
T11 Special security objectives included 1.0 0 0
T12 Direct access for third parties 1.0 0 0
T13 Special user training required 1.0 0 0
Technical Factor Value (TFactor) 15
Technical Complexity Factor (TCF) 5 0.6 1 (0.01 * TFactor) 0.75 5 0.6 1 (0.01 * 15)
Environmental Factors:
Factor Number Description Weight Assigned Value (0–5) Weighted Value Notes
E1 Familiarity with system 1.5 4 6
development process being used
E2 Application experience 0.5 4 2
E3 Object-oriented experience 1.0 4 4
E4 Lead analyst capability 0.5 5 2.5
E5 Motivation 1.0 5 5
E6 Requirements stability 2.0 5 10
E7 Part-time staff –1.0 0 0
E8 Diffi culty of programming language –1.0 4 –4.0
Environmental Factor Value (EFactor) 25.5
Environmental Factor (EF) 5 1.4 1 (20.03 * EFactor) 0.635 5 1.4 1 (20.03 * 25.5)
Adjusted Use Case Points (UCP) 5 UUCP * TCF * ECF 33.3375 5 70 * 0.75 * 0.635
Effort in person-hours 5 UCP * PHM 666.75 5 20 * 33.3375
Project Effort Estimation 61
Th e technical factors include the following (see Figure 2-15):
■ Whether the system is going to be a distributed system
■ Th e importance of response time
■ Th e effi ciency level of the end user using the system
■ Th e complexity of the internal processing of the system
■ Th e importance of code reuse
■ How easy the installation process has to be
■ Th e importance of the ease of using the system
■ How important it is for the system to be able to be ported to another platform
■ Whether system maintenance is important
■ Whether the system is going to have to handle parallel and concurrent
processing
■ Th e level of special security required
■ Th e level of system access by third parties
■ Whether special end user training is to be required.
Assuming the values for the technical factors are T1 (0), T2 (5), T3 (3), T4 (1), T5 (1),
T6 (2), T7 (4), T8 (0), T9 (2), T10 (0), T11 (0), T12 (0), and T13 (0), respectively, the
technical factor value (TFactor) is computed as the weighted sum of the individual technical
factors. In this case TFactor equals 15 (see Figure 2-16). Plugging this value into the technical
complexity factor (TCF) equation (0.6 1 (.01 * TFactor)) of the use-case point worksheet gives
a value of .75 for the TCF of the system (see Figures 2-15 and 2-16).
Th e environmental factors include the following (see Figure 2-15):
■ Th e level of experience the development staff has with the development process
being used
■ Th e application being developed
■ Th e level of object-oriented experience
■ Th e level of capability of the lead analyst
■ Th e level of motivation of the development team to deliver the system
■ Th e stability of the requirements
■ Whether part-time staff have to be included as part of the development team
■ Th e diffi culty of the programming language being used to implement the system
Assuming the values for the environmental factors were E1 (4), E2 (4), E3 (4), E4 (5), E5 (5),
E6 (5), E7 (0), and E8 (4) gives an environmental factor value (EFactor) of 25.5 (see Figure
2-16). Like the TFactor, Efactor is simply the sum of the weighted values. Using the envi-
ronmental factor (EF) equation (1.4 1 (20.03 * EFactor)) of the use-case point worksheet
produces a value of .635 for the EF of the system (see Figures 2-15 and 2-16). Plugging the
TCF and EF values, along with the UUCP value computed earlier, into the adjusted use-case
points equation (UUCP * TCF * EF) of the worksheet yields a value of 33.3375 adjusted use-
case points (UCP) (see Figure 2-16).
Now that we know the estimated size of the system by means of the value of the adjusted
use-case points, we are ready to estimate the eff ort required to build the system. In Karner’s
original work, he suggested simply multiplying the number of use-case points by 20 to
estimate the number of person-hours required to build the system. However, based on
additional experiences using use-case points, a decision rule to determine the value of the
62 Chapter 2 Project Management
person-hours multiplier (PHM) has been created that suggests using either 20 or 28, based on
the values assigned to the individual environmental factors. Th e decision rule is:
If the sum of (number of Efactors E1 through E6 assigned value , 3) and
(number of Efactors E7 and E8 assigned value . 3)
< 2
PHM 5 20
Else If the sum of (number of Efactors E1 through E6 assigned value , 3) and
(number of Efactors E7 and E8 assigned value . 3)
5 3 or 4
PHM 5 28
Else
Rethink project; it has too high of a risk for failure
Based on these rules, because none of Efactors E1 through E6 have a value less than 3
and only Efactor E8 has a value greater than 3, the sum of the number EFactors is 1. Th us, the
system should use a PHM of 20. Plugging the values for UCP (33.3375) and PHM (20) into
the eff ort equation (UCP * PHM) gives an estimated number of person-hours of 666.75 hours
(see Figures 2-15 and 2-16).
CREATING AND MANAGING THE WORKPLAN
Once a project manager has a general idea of the functionality and eff ort for the project, he
or she creates a workplan, which is a dynamic schedule that records and keeps track of all the
tasks that need to be accomplished over the course of the project. Th e workplan lists each
task, along with important information about it, such as when it needs to be completed, the
person assigned to do the work, and any deliverables that will result. Th e level of detail and
the amount of information captured by the workplan depend on the needs of the project, and
the detail usually increases as the project progresses.
Th e overall objectives for the system should be listed on the system request, and it is the
project manager’s job to identify all the tasks that need to be accomplished to meet those
objectives. Th is sounds like a daunting task. How can someone know everything that needs
to be done to build a system that has never been built before?
One approach for identifying tasks is to get a list of tasks that has already been devel-
oped and to modify it. Th ere are standard lists of tasks, or methodologies, that are available
for use as a starting point. As we stated in Chapter 1, a methodology is a formalized approach
to implementing a systems development process (i.e., it is a list of steps and deliverables).
A project manager can take an existing methodology, select the steps and deliverables that
apply to the current project, and add them to the workplan. If an existing methodology is not
available within the organization, methodologies can be purchased from consultants or ven-
dors, or books such as this textbook can serve as a guide. Because most organizations have a
methodology they use for projects, using an existing methodology is the most popular way to
create a workplan. In our case, because we are using a Unifi ed Process-based methodology,
we can use the phases, workfl ows, and iterations as a starting point to create an evolutionary
work breakdown structure and an iterative workplan.
Evolutionary Work Breakdown Structures and Iterative Workplans7
Because object-oriented systems approaches to systems analysis and design support incre-
mental and iterative development, any project planning approach for object-oriented systems
7 Th is material in this section is based on Walker Royce, Soft ware Project Management: A Unifi ed Framework (Read-
ing, MA: Addison-Wesley, 1998).
Creating and Managing the Workplan 63
development also requires an incremental and iterative process. In the description of the enhanced
Unifi ed Process in Chapter 1, the development process was organized around iterations, phases,
and workfl ows. In many ways, a workplan for an incremental and iterative development process
is organized in a similar manner. For each iteration, there are diff erent tasks executed on each
workfl ow. Th is section describes an incremental and iterative process using evolutionary WBSs
for project planning that can be used with object-oriented systems development.
Evolutionary WBSs allow the analyst to develop an iterative workplan. First, evolutionary
WBSs are organized in a standard manner across all projects: by workfl ows, phases, and then
the specifi c tasks that are accomplished during an individual iteration. Second, evolutionary
WBSs are created in an incremental and iterative manner. Th is encourages a more realistic
view of both cost and schedule estimation. Th ird, because the structure of an evolutionary
WBS is not tied to any specifi c project, evolutionary WBSs enable the comparison of the
current project to earlier projects. Th is supports learning from past successes and failures.
In the case of the enhanced Unifi ed Process, the workfl ows are the major points listed
in the WBS. Next, each workfl ow is decomposed along the phases of the enhanced Unifi ed
Process. Aft er that, each phase is decomposed along the tasks that are to be completed to cre-
ate the deliverables associated with an individual iteration contained in each phase (see Figure
1-16). Th e template for the fi rst two levels of an evolutionary WBS for the enhanced Unifi ed
Process would look like Figure 2-17.
As each iteration through the development process is completed, additional iterations and
tasks are added to the WBS (i.e., the WBS evolves along with the evolving information system).8
I. Business Modeling
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
II. Requirements
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
III. Analysis
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
IV. Design
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
V. Implementation
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VI. Test
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VII. Deployment
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VIII. Confi guration and
Change Management
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
IX. Project Management
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
X. Environment
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
XI. Operations and Support
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
XII. Infrastructure Management
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
FIGURE 2-17
Evolutionary WBS
Template for the
Enhanced Unifi ed
Process
8 Good sources that help explain this approach are Phillippe Krutchen, “Planning an Iterative Project,” Th e Rational
Edge (October 2002); Eric Lopes Cordoza and D. J. de Villiers, “Project Planning Best Practices,” Th e Rational Edge
(August 2003).
64 Chapter 2 Project Management
For example, typical activities for the inception phase of the project management workfl ow
would include identifying the project, performing the feasibility analysis, selecting the project,
and estimating the eff ort. Th e inception phase of the requirements workfl ow would include
determining the requirements gathering and analysis techniques, identifying functional and
nonfunctional requirements, interviewing stakeholders, developing a vision document, and
developing use cases. Probably no tasks are associated with the inception phase of the operations
and support workfl ow. A sample evolutionary WBS for planning the inception phase of the
enhanced Unifi ed Process, based on Figures 1-16 and 2-17, is shown in Figure 2-18. Notice the
last two tasks for the project management workfl ow are “create workplan for fi rst iteration of
the elaboration phase” and “assess the inception phase”; the last two things to do are to plan for
the next iteration in the development of the evolving system and to assess the current iteration.
As the project moves through later phases, each workfl ow has tasks added to its iterations. For
example, the analysis workfl ow will have the creation of the functional, structural, and behav-
ioral models during the elaboration phase. Finally, when an iteration includes a lot of complex
tasks, traditional tools, such as Gantt charts and network diagrams, can be used to detail the
workplan for that specifi c iteration.
FIGURE 2-18
Evolutionary
WBS for a Single
Iteration-Based
Inception Phase
Duration Dependency
I. Business Modeling
a. Inception
1. Understand current business situation 0.50 days
2. Uncover business process problems 0.25 days
3. Identify potential projects 0.25 days
b. Elaboration
c. Construction
d. Transition
e. Production
II. Requirements
a. Inception
1. Identify appropriate requirements-analysis technique 0.25 days
2. Identify appropriate requirements-gathering techniques 0.25 days
3. Identify functional and nonfunctional requirements II.a.1, II.a.2
A. Perform JAD sessions 3 days
B. Perform document analysis 5 days II.a.3.A
C. Conduct interviews II.a.3.A
1. Interview project sponsor 0.5 days
2. Interview inventory system contact 0.5 days
3. Interview special order system contact 0.5 days
4. Interview ISP contact 0.5 days
5. Interview CD Selection Web contact 0.5 days
6. Interview other personnel 1 day
D. Observe retail store processes 0.5 days II.a.3.A
4. Analyze current systems 4 days II.a.1, II.a.2
5. Create requirements defi nition II.a.3, II.a.4
A. Determine requirements to track 1 day
B. Compile requirements as they are elicited 5 days II.a.5.A
C. Review requirements with sponsor 2 days II.a.5.B
b. Elaboration
c. Construction
d. Transition
e. Production
Creating and Managing the Workplan 65
FIGURE 2-18
Continued
Duration Dependency
III. Analysis
a. Inception
1. Identify business processes 3 days
2. Identify use cases 3 days III.a.1
b. Elaboration
c. Construction
d. Transition
e. Production
IV. Design
a. Inception
1. Identify potential classes 3 days III.a
b. Elaboration
c. Construction
d. Transition
e. Production
V. Implementation
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VI. Test
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VII. Deployment
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
VIII. Confi guration and Change Management
a. Inception
1. Identify necessary access controls for developed artifacts 0.25 days
2. Identify version control mechanisms for developed artifacts 0.25 days
b. Elaboration
c. Construction
d. Transition
e. Production
IX. Project Management
a. Inception
1. Create workplan for the inception phase 1 day
2. Create system request 1 day
3. Perform feasibility analysis IX.a.2
A. Perform technical feasibility analysis 1 day
B. Perform economic feasibility analysis 2 days
C. Perform organizational feasibility analysis 2 days
66 Chapter 2 Project Management
Duration Dependency
4. Identify project effort 0.50 days IX.a.3
5. Identify staffi ng requirements 0.50 days IX.a.4
6. Compute cost estimate 0.50 days IX.a.5
7. Create workplan for fi rst iteration of the
elaboration phase 1 day IX.a.1
8. Assess inception phase 1 day I.a, II.a, III.a
IV.a, V.a, VI.a
VII.a, VIII.a,
IX.a, X.a, XI.a
XII.a
b. Elaboration
c. Construction
d. Transition
e. Production
X. Environment
a. Inception
1. Acquire and install CASE tool 0.25 days
2. Acquire and install programming environment 0.25 days
3. Acquire and install confi guration and change
management tools 0.25 days
4. Acquire and install project management tools 0.25 days
b. Elaboration
c. Construction
d. Transition
e. Production
XI. Operations and Support
a. Inception
b. Elaboration
c. Construction
d. Transition
e. Production
XII. Infrastructure Management
a. Inception
1. Identify appropriate standards and enterprise models 0.25 days
2. Identify reuse opportunities, such as patterns,
frameworks, and libraries 0.50 days
3. Identify similar past projects 0.25 days
b. Elaboration
c. Construction
d. Transition
e. Production
FIGURE 2-18
Continued
Managing Scope
An analyst may assume that a project will be safe from scheduling problems because he or she
carefully estimated and planned the project up front. However, the most common reason for
schedule and cost overruns—scope creep—occurs aft er the project is under way. Scope creep
happens when new requirements are added to the project aft er the original project scope was
defi ned and frozen. It can happen for many reasons: Users might suddenly understand the
Creating and Managing the Workplan 67
potential of the new system and realize new functionality that would be useful; developers
might discover interesting capabilities to which they become very attached; a senior manager
might decide to let this system support a new strategy that was developed at a recent board
meeting.
Fortunately, using an iterative and incremental development process allows the team to
deal with changing requirements in an eff ective way. However, the more extensive the change
becomes, the greater the impact on cost and schedule. Th e keys are to identify the require-
ments as well as possible in the beginning of the project and to apply analysis techniques
eff ectively. For example, if needs are fuzzy at the project’s onset, a combination of intensive
meetings with the users and prototyping would allow users to “experience” the requirements
and better visualize how the system could support their needs.
Of course, some requirements may be missed no matter what precautions are taken.
However, the project manager should allow only absolutely necessary requirements to be
added aft er the project begins. Even at that point, members of the project team should care-
fully assess the ramifi cations of the addition and present the assessment to the users. Any
change that is implemented should be carefully tracked so that an audit trail exists to measure
the change’s impact.
Sometimes changes cannot be incorporated into the present system even though they
truly would be benefi cial. In this case, these additions should be recorded as future enhance-
ments to the system. Th e project manager can off er to provide functionality in future releases
of the system, thus getting around telling someone “no.”
A couple of useful agile techniques to manage the scope of the project while attempting
to satisfy the client are daily scrum meetings and the product backlog used with Scrum.
Essentially a daily scrum meeting is a very short, typically fi ft een minutes, meeting that keeps
the development team up to date as to the current status of the evolving system. Th e content
of the meeting typically only covers what has been accomplished since the previous meeting,
what will be accomplished before the next meeting, and what obstacles could come up that
could prevent progress from being made. Also, new requested features could be brought
up. However, all proposed additional features are simply added to the product backlog that
could be considered during the next iteration or timebox (sprint in Scrum’s nomenclature).
Th e product backlog is essentially a prioritized list of the functional requirements that will
be completed during the current iteration. In Scrum, only the client is allowed to modify the
product backlog. In this manner, the development team always has a list of the current set of
critical requirements. As long as the project is relatively small, this approach to scope man-
agement is very eff ective.
Timeboxing
Another approach to scope management is a technique called timeboxing. Up until now, we
have described task-oriented projects. In other words, we have described projects that have a
schedule driven by the tasks that need to be accomplished, so the greater number of tasks and
requirements, the longer the project will take. Some companies have little patience for devel-
opment projects that take a long time, and these companies take a time-oriented approach
that places meeting a deadline above delivering functionality.
Th ink about the use of word processing soft ware. For 80 percent of the time, only 20 percent
of the features, such as the spelling checker, boldfacing, and cutting and pasting, are used. Other
features, such as document merging and creating mailing labels, may be nice to have, but they
are not a part of day-to-day needs. Th e same goes for other soft ware applications; most users
rely on only a small subset of their capabilities. Ironically, most developers agree that typically
75 percent of a system can be provided relatively quickly, with the remaining 25 percent of the
functionality demanding most of the time.
68 Chapter 2 Project Management
To resolve this incongruency, the technique of timeboxing has become quite popular,
especially when using RAD and agile methodologies. Th is technique sets a fi xed deadline for
a project and delivers the system by that deadline no matter what, even if functionality needs
to be reduced. Timeboxing ensures that project teams don’t get hung up on the fi nal fi nishing
touches that can drag out indefi nitely, and it satisfi es the business by providing a product
within a relatively short time frame.
Several steps are involved in implementing timeboxing on a project. First, set the date of
delivery for the proposed goals. Th e deadline should not be impossible to meet, so it is best
to let the project team determine a realistic due date. If you recall from Chapter 1, the Scrum
agile methodology sets all of its timeboxes (sprint) to thirty working days. Next, build the core
of the system to be delivered; you will fi nd that timeboxing helps create a sense of urgency and
helps keep the focus on the most important features. Because the schedule is absolutely fi xed,
functionality that cannot be completed needs to be postponed. It helps if the team prioritizes
a list of features beforehand to keep track of what functionality the users absolutely need.
Quality cannot be compromised, regardless of other constraints, so it is important that the
time allocated to activities is not shortened unless the requirements are changed (e.g., don’t
reduce the time allocated to testing without reducing features). At the end of the time period,
a high-quality system is delivered, but it is likely that future iterations will be needed to make
changes and enhancements. In that case, the timeboxing approach can be used once again.
Refi ning Estimates
Th e estimates that are produced during inception need to be refi ned as the project progresses.
Th is does not mean that estimates were poorly done at the start of the project; rather, it is
virtually impossible to develop an exact assessment of the project’s schedule at the beginning
of the development process. A project manager should expect to be satisfi ed with broad
ranges of estimates that become more and more specifi c as the project’s product becomes
better defi ned.
During planning, when a system is fi rst requested, the project sponsor and project
manager attempt to predict how long the development process will take, how much it will
cost, and what it will ultimately do when it is delivered (i.e., its functionality). However, the
estimates are based on very little knowledge of the system. As the system moves into the
elaboration, more information is gathered, the system concept is developed, and the estimates
become even more accurate and precise. As the system moves closer to completion, the accu-
racy and precision increase, until it is delivered.
According to one of the leading experts in soft ware development,9 a well-done project
plan (prepared at the end of inception) has a 100 percent margin of error for project cost and
a 25 percent margin of error for schedule time. In other words, if a carefully done project plan
estimates that a project will cost $100,000 and take twenty weeks, the project will actually cost
between $0 and $200,000 and take between fi ft een and twenty-fi ve weeks.
What happens if you overshoot an estimate (e.g., analysis ends up lasting two weeks
longer than expected)? Th ere are a number of ways to adjust future estimates. If the project
team fi nishes a step ahead of schedule, most project managers shift the deadlines sooner by
the same amount but do not adjust the promised completion date. Th e challenge, however,
occurs when the project team is late in meeting a scheduled date. Th ree possible responses to
missed schedule dates are presented in Figure 2-19. If, early in the project, an estimate proves
to be too optimistic, planners should not expect to make up for lost time—very few projects
9 Barry W. Boehm et al., “Cost Models for Future Soft ware Life Cycle Processes: COCOMO 2.0,” in J. D. Arthur and
S. M. Henry (eds.), Annals of Soft ware Engineering: Special Volume on Soft ware Process and Product Measurement
(Amsterdam: J. C. Baltzer AG Science Publishers, 1995).
Creating and Managing the Workplan 69
end up doing this. Instead, they should change future estimates to include an increase similar
to the one that was experienced. For example, if the fi rst phase was completed 10 percent over
schedule, planners should increase the rest of their estimates by 10 percent.
Managing Risk
One fi nal facet of project management is risk management, the process of assessing and
addressing the risks that are associated with developing a project. Many things can cause
risks: weak personnel, scope creep, poor design, and overly optimistic estimates. Th e project
team must be aware of potential risks so that problems can be avoided or controlled well
ahead of time.
Typically, project teams create a risk assessment, or a document that tracks potential risks
along with an evaluation of the likelihood of each risk and its potential impact on the project
(Figure 2-20). A paragraph or two is also included to explain potential ways that the risk can
be addressed. Th ere are many options: Th e risk could be publicized, avoided, or even elim-
inated by dealing with its root cause. For example, imagine that a project team plans to use
new technology but its members have identifi ed a risk in the fact that its members do not have
the right technical skills. Th ey believe that tasks may take much longer to perform because of
a high learning curve. One plan of attack could be to eliminate the root cause of the risk—the
lack of technical experience by team members—by fi nding the time and resources needed to
provide proper training to the team.
Most project managers keep abreast of potential risks, even prioritizing them according
to their magnitude and importance. Over time, the list of risks will change as some items are
removed and others surface. Th e best project managers, however, work hard to keep risks
from having an impact on the schedule and costs associated with the project.
If you assume the rest of the project is Do not change schedule. High risk
simpler than the part that was late
and is also simpler than believed
when the original schedule estimates
were made, you can make up lost time.
If you assume the rest of the project is Increase the entire schedule by the Moderate risk
simpler than the part that was late total amount of time that you are
and is no more complex than the behind (e.g., if you missed the
original estimate assumed, you can’t scheduled date by two weeks, move
make up the lost time, but you will the rest of the schedule dates to two
not lose time on the rest of the weeks later). If you included padded
project. time at the end of the project in the
original schedule, you might not have
to change the promised system
delivery date; you’ll just use up the
padded time.
If you assume that the rest of the Increase the entire schedule by the Low risk
project is as complex as the part percentage of weeks that you are
that was late (your original estimates behind (e.g., if you are two weeks
were too optimistic), then all the late on part of the project that was
scheduled dates in the future supposed to take eight weeks, you
underestimate the real time required need to increase all remaining
by the same percentage as the part time estimates by 25 percent). If
that was late. this moves the new delivery date
beyond what is acceptable to the
project sponsor, the scope of the
project must be reduced.
Assumptions Actions Level of Risk
FIGURE 2-19
Possible Actions
When a Schedule
Date Is Missed
70 Chapter 2 Project Management
STAFFING THE PROJECT
Staffi ng the project includes determining how many people should be assigned to the project,
matching people’s skills with the needs of the project, motivating them to meet the project’s
objectives, and minimizing the confl ict that will occur over time. Th e deliverables for this part
of project management are a staffi ng plan, which describes the number and kinds of people
who will work on the project, the overall reporting structure, and the project charter, which
describes the project’s objectives and rules. However, before describing the development of a
staffi ng plan, how to motivate people, and how to handle confl ict, we describe a set of char-
acteristics of jelled teams.
Characteristics of a Jelled Team10
Th e idea of a jelled team has existed for a long time. Most (if not all) student groups are not
representative of the idea of a jelled team, and you may have never had the opportunity to
appreciate the eff ectiveness of a true team. In fact, DeMarco and Lister point out that teams
are not created; they are grown. Typically, in class projects, students are assigned or asked to
form a group, which makes the ability to grow a team very limited. However, growing devel-
opment teams is crucial in information systems development. Th e whole set of agile soft ware
development approaches hinges on growing jelled teams. Otherwise, agile development
approaches would totally fail.
According to DeMarco and Lister,11 “[a] jelled team is a group of people so strongly knit
that the whole is greater than the sum of the parts. Th e production of such a team is greater
than that of the same people working in unjelled form.” Th ey go on to state that a jelled “team
can become almost unstoppable, a juggernaut for success.” When is the last time that you
worked with a group on a class project that could be described “a juggernaut for success”?
Demarco and Lister identify fi ve characteristics of a jelled team.
10 Th e material in the section is based on T. DeMarco and T. Lister, Peopleware: Productive Projects and Teams,
2nd Ed. (New York: Dorset House, 1999); P. Lencioni, Th e Five Dysfunctions of a Team: A Leadership Fable (San
Francisco: Jossey-Bass, 2002).
11 T. DeMarco and T. Lister, Peopleware: Productive Projects and Teams, 2nd Ed., p. 123.
Risk Assessment
RISK 1: The development of this system likely will be slowed
considerably because project team members have not
programmed in Java prior to this project.
Likelihood of risk: High probability of risk.
Potential impact on the project: This risk will probably increase the time to complete
programming tasks by 50 percent.
Ways to address this risk:
It is very important that time and resources are allocated to up-front training in Java for the
programmers who are used for this project. Adequate training will reduce the initial learning curve
for Java when programming begins. Additionally, outside Java expertise should be brought in for at
least some part of the early programming tasks. This person should be used to provide experiential
knowledge to the project team so that Java-related issues (of which novice Java programmers would
be unaware) are overcome.
RISK 2: …
FIGURE 2-20
Sample Risk
Assessment
Staffi ng the Project 71
First, jelled teams have a very low turnover during a project. Typically, members of
a jelled team feel a responsibility to the other team members. Th is responsibility is felt so
intensely that for a member to leave the team, the member would feel that they were letting
the team down and that they were breaking a bond of trust.
Second, jelled teams have a strong sense of identity. In many classes, when you are part of
a group, the group chooses some cute name to identify the group and diff erentiate it from the
other groups. However, in this case, it is not simply the choosing of a name. It is instead evolv-
ing every member into something that only exists within the team. Th is can be seen when
members of the team tend to do non–work-related activities together, e.g., do lunch together
as a team or form a basketball team composed of only members of the development team.
Th ird, the strong sense of identity tends to lead the team into feeling a sense of eliteness.
Th e members of a jelled development team almost have a swagger about the way they relate
to nonteam employees. Good examples that come to mind that possess this sense of eliteness
outside of the scope of information systems development teams are certain sports teams,
U.S. Navy Seal teams, or big city police force SWAT teams. In all three examples, each team
member is highly competent in his or her specialty area, and each other team member knows
(not thinks) that he or she can depend on the team members performing his or her individual
jobs with a very high-level of skill.
Fourth, during the development process, jelled teams feel that the team owns the infor-
mation system being developed and not any one individual member. In many ways, you could
almost say that jelled teams are a little communistic in nature. By this we mean that the individ-
ual contributions to the eff ort are not important to a true team. Th e only things that matter are
the output of the team. However, this is not to imply that a member who does not deliver his or
her fair share will not go unpunished. In a jelled team, any member who is not producing is actu-
ally breaking his or her bond of trust with the other team members (see the fi rst characteristic).
Th e fi nal characteristic of a jelled team is that team members really enjoy (have fun)
doing their work. Th e members actually like to go to work and be with their team members.
Much of this can be attributed to the level of challenge they receive. If the project is challeng-
ing and the members of the team are going to learn something from completing the project,
the members of a jelled team will enjoy tackling the project.
When a team jells, they will avoid the fi ve dysfunctions of a team defi ned by Lencioni.
Lack of trust is the primary cause of a team becoming dysfunctional. Lencioni describes four
other causes of a team becoming dysfunctional that can come from the lack of trust. First,
dysfunctional teams fear confl ict, whereas members of a jelled team never fear confl ict.12
Going to a member of a jelled team and admitting that you do not know how to do something
is no big deal. In fact, it provides a method for the team member to help out, which would
increase the level of trust between the two members. Second, dysfunctional teams do not have
a commitment to the team from the individual members. Instead, they tend to focus on their
individual performance instead of the team’s performance. Th is can even be to the detriment
of the development team. Obviously, this is not an issue for jelled teams. Th ird, dysfunctional
teams try to avoid accountability. With jelled teams, accountability is not an issue. Members
of a jelled team feel a high level of responsibility to the other team members. No team mem-
ber ever wants to let down the team. Furthermore, owing to the bond that holds jelled teams
together, no member has any problem with holding other members accountable for their per-
formance (or lack of performance). Fourth, dysfunctional teams do not pay attention to the
team’s results. Again, in this case, the cause of this dysfunction is that the individual members
only focus on their individual goals. From a team management perspective, the team leader
should focus on getting the goals of the team aligned; a jelled team will attain the goals.
12 When confl ict occurs, it is necessary to address it in an eff ective manner. We discuss how to handle confl ict later
in the chapter.
72 Chapter 2 Project Management
Staffi ng Plan
Th e fi rst step to staffi ng is determining the average number of staff needed for the project.
To calculate this fi gure, divide the total person-months of eff ort by the optimal schedule.
So to complete a forty-person-month project in ten months, a team should have an average
of four full-time staff members, although this may change over time as diff erent specialists
enter and leave the team (e.g., business analysts, programmers, technical writers).
Many times, the temptation is to assign more staff to a project to shorten the project’s
length, but this is not a wise move. Adding staff resources does not translate into increased
productivity; staff size and productivity share a disproportionate relationship, mainly because
it is more diffi cult to coordinate a large number of staff members. Th e more a team grows,
the more diffi cult it becomes to manage. Imagine how easy it is to work on a two-person
project team: Th e team members share a single line of communication. But adding two peo-
ple increases the number of communication lines to six, and greater increases lead to more
dramatic gains in communication complexity. Figure 2-21 illustrates the impact of adding
team members to a project team.
One way to reduce effi ciency losses on teams is to understand the complexity that is cre-
ated in numbers and to build in a reporting structure that tempers its eff ects. Th e general rule
Two-person team Four-person team
Eight-person teamSix-person team
FIGURE 2-21
Increasing Com-
plexity with Larger
Teams
Staffi ng the Project 73
is to keep team sizes to fewer than eight to ten people; therefore, if more people are needed,
create sub-teams. In this way, the project manager can keep the communication eff ective
within small teams, which, in turn, communicate to a contact at a higher level in the project.
Aft er the project manager understands how many people are needed for the project,
he or she creates a staffi ng plan that lists the roles and the proposed reporting structure that
are required for the project. Typically, a project has one project manager who oversees the
overall progress of the development eff ort, with the core of the team comprising the various
types of analysts described in Chapter 1. A functional lead is usually assigned to manage
a group of analysts, and a technical lead oversees the progress of a group of programmers and
more technical staff members.
Th ere are many structures for project teams; Figure 2-22 illustrates one possible confi g-
uration of a project team. Aft er the roles are defi ned and the structure is in place, the project
manager needs to think about which people can fi ll each role. Oft en, one person fi lls more
than one role on a project team.
When you make assignments, remember that people have technical skills and interper-
sonal skills, and both are important on a project. Technical skills are useful when working
with technical tasks (e.g., programming in Java) and in trying to understand the various
roles that technology plays in the particular project (e.g., how a Web server should be con-
fi gured on the basis of a projected number of hits from customers). Interpersonal skills,
on the other hand, include interpersonal and communication abilities that are used when
dealing with business users, senior management executives, and other members of the
project team. Th ey are particularly critical when performing the requirements- gathering
activities and when addressing organizational feasibility issues. Each project requires
unique technical and interpersonal skills.
Ideally, project roles are fi lled with people who have the right skills for the job. However,
the people who fi t the roles best might not be available; they may be working on other projects,
or they might not exist in the company. Th erefore, assigning project team members really is
a combination of fi nding people with the appropriate skill sets and fi nding people who are
available. When the skills of the available project team members do not match what is actually
required by the project, the project manager has several options to improve the situation. First,
people can be pulled off other projects, and resources can be shuffl ed around. Th is is the most
disruptive approach from the organization’s perspective. Another approach is to use outside
help—such as a consultant or contractor—to train team members and start them off on the
right foot. Mentoring may also be an option; a project team member can be sent to work on
another similar project so that he or she can return with skills to apply to the current job.
Functional
lead
Project
manager
ProgrammerAnalyst Analyst Analyst Programmer
Technical
lead
FIGURE 2-22
Possible Reporting
Structure
74 Chapter 2 Project Management
Motivation
Assigning people to tasks isn’t enough; project managers need to motivate the people to
ensure a project’s success. Motivation has been found to be the number one infl uence on
people’s performance,13 but determining how to motivate the team can be quite diffi cult. You
might think that good project managers motivate their staff by rewarding them with money
and bonuses, but most project managers agree that this is the last thing that should be done.
Th e more oft en managers reward team members with money, the more they expect it—and
most times monetary motivation won’t work. Pink14 has suggested a set of principles to follow
to motivate individuals in twenty-fi rst century fi rms. In this section, we adapt his suggestions
to information systems development teams.
Pink suggests considering using some form of the 20 percent time rule to motivate
individuals. Th is rule suggests that 20 percent of an employee’s time should be spent on
some idea in which he or she believes. Th e project does not have to be related to the project
at hand. On the surface, this sounds like a colossal waste of time, but this idea should not be
discarded. Google’s Gmail and Google News were developed using the 20 percent time rule.
If 20 percent sounds too high, Pink suggests that you consider 10 percent to begin with.
He recommends that fi rms should be willing to fund small “Now Th at” awards. Th ese awards
are given as small signs of appreciation for doing a great job. However, these awards are not given
by a manager to an employee but from an employee to a peer of the employee. Th e awards are
monetary, but they are very small, typically $50. As such, they really are not relevant from a mon-
etary perspective. However, they are very relevant because they are given by one of the employee’s
colleagues to show that some action that the employee did was appreciated.
Pink endorses the idea of applying Robert Reich’s (President’s Clinton’s Secretary of
Labor) pronoun test. If an employee (or team member) refers to the fi rm (the team) as “they,”
then there is the real possibility that the employee feels disengaged or possibly alienated. On
the other hand, when employees refer to the fi rm as “we,” they obviously feel like they are
part of the organization. From a team perspective, this could be an indication that the team
has begun to jell.
Pink suggests that management should periodically consider giving each employee a day
on which he or she can work on anything he or she wants. In some ways, this is related to
the 20 percent rule. It does not necessarily require one day a week (20 percent), but it does
require some deliverable. Th e deliverable can be a new utility program that could be used by
lots of diff erent projects, it could be a new prototype of a new soft ware product, or it could
be an improvement for a business process that is used internally. Th e goal is to provide team
members with the ability to focus on interesting and challenging problems that might (or
might not) provide results to the fi rm’s bottom line. Regardless, it demonstrates an amount
of trust and respect that the fi rm has for its employees.
He recommends that managers remove the issue of compensation from the motivation
equation. By this, he means that all employees should be paid a suffi cient amount so that com-
pensation awards are not an issue. Technical employees on project teams are much more moti-
vated by recognition, achievement, the work itself, responsibility, advancement, and the chance
to learn new skills.15 Simplistic fi nancial awards, such as raises that are perceived as being unjust,
can actually demotivate the overall team and lower overall performance.
13 Barry W. Boehm, Soft ware Engineering Economics (Englewood Cliff s, NJ: Prentice Hall, 1981). One of the best
books on managing project teams is that by Tom DeMarco and Timothy Lister, Peopleware: Productive Projects and
Teams (New York: Dorset House, 1987).
14 D. H. Pink, Drive: Th e Surprising Truth About What Motivates Us (New York, NY: Riverhead Books, 2009).
15 F. H. Hertzberg, “One More Time: How Do You Motivate Employees?” Harvard Business Review (January–
February 1968).
Staffi ng the Project 75
He advocates that twenty-fi rst century bosses (team leaders) need to be willing to give up
control. Many of the agile development approaches make similar suggestions. Appelo16 suggests
that an open door policy that is supported by a team leader actually can be self-defeating. In the
case of soft ware development teams, an open door policy implies that the team leader has a door
that can be left open, whereas the poor individual team member does not have an offi ce with a
door. In this case, Appelo suggests that the team leader move from the offi ce with a door to the
same shared space in which the team resides. One of Pink’s other ideas is for the team leader to
not use controlling language such as telling the team member that he or she “must” do some-
thing. Instead, the team leader should ask the team member to “consider” or “think about” the
idea. In some ways, a true team leader should never receive credit for any ideas associated with
the team. Instead, a team leader should make suggestions and encourage the team members
to consider ideas and, most importantly, let the team member and the team receive the credit.
Pink provides evidence that intrinsic motivation is very important for twenty-fi rst century
knowledge workers. Pink suggests that intrinsically motivating individuals requires providing
them with a degree of autonomy, supporting them in such a way that they can master their area
of expertise, and encouraging them to pursue projects with a purpose. Providing team members
with autonomy relates to the jelled team concept of trust. Team leaders need to trust the team
members to deliver the soft ware for which they are responsible. Supporting team members so that
they can master their area of expertise can be as simple as providing support to attend confer-
ences, seminars, and training sessions that deal with the member’s area of expertise. It also could
imply providing the team member with a high-end development environment. For example,
when building information visualization and virtual reality applications, special hardware and
soft ware environments can make it much easier to master the technology to develop the appli-
cation. Finally, today it is very important for team members to feel that what they are doing can
make a diff erence. A team leader should encourage the team members to tackle problems that
can impact people’s lives. Th is can easily be accomplished through the use of the 20 percent rule.
Handling Confl ict
Th e third component of staffi ng is organizing the project to minimize confl ict among group
members. Group cohesiveness (the attraction that members feel to the group and to other
members) contributes more to productivity than do project members’ individual capabil-
ities or experiences.17 Clearly defi ning the roles on the project and holding team members
accountable for their tasks are a good way to begin mitigating potential confl ict on a project.
Some project managers develop a project charter, which lists the project’s norms and ground
rules. For example, the charter may describe when the project team should be at work, when
staff meetings will be held, how the group will communicate with each other, and what are
the procedures for updating the workplan as tasks are completed. Figure 2-23 lists additional
techniques that can be used at the start of a project to keep confl ict to a minimum.
ENVIRONMENT AND INFRASTRUCTURE MANAGEMENT
Th e environment and infrastructure management workfl ows support the development
team throughout the development process. Th e environment workfl ow primarily deals with
choosing the correct set of tools that will be used throughout the development process and
16 J. Appelo, Management 3.0: Leading Agile Developers, Developing Agile Leaders (Upper Saddle River, NJ:
Addison-Wesley, 2011).
17 B. Lakhanpal, “Understanding the Factors Infl uencing the Performance of Soft ware Development Groups: An
Exploratory Group-Level Analysis,” Information and Soft ware Technology 35, no. 8 (1993): 468–473.
76 Chapter 2 Project Management
identifying the appropriate set of standards to be followed during the development process.
Infrastructure management workfl ow deals with choosing the appropriate level and type of
documentation that will be created during the development process. Other activities asso-
ciated with the infrastructure management workfl ow include developing, modifying, and
reusing predefi ned components, frameworks, libraries, and patterns. Th e topic of reuse is
discussed in later chapters (see Chapters 5 and 8).
CASE Tools
Computer-aided soft ware engineering (CASE) is a category of soft ware that automates all or
part of the development process. Some CASE soft ware packages are used primarily to support
the analysis workfl ow to create integrated diagrams of the system and to store information
regarding the system components, whereas others support the design workfl ow that can be
used to generate code for database tables and system functionality. Other CASE tools contain
functionality that supports tasks throughout the system-development process. CASE comes
in a wide assortment of fl avors in terms of complexity and functionality, and many good
tools are available in the marketplace to support object-oriented systems development (e.g.,
ArgoUml, Enterprise Architect, Poseidon, Visual Paradigm, and IBM’s Rational Rose).
Th e benefi ts of using CASE are numerous. With CASE tools, tasks can be completed
and altered faster, development documentation is centralized, and information is illustrated
through diagrams, which are typically easier to understand. Potentially, CASE can reduce
maintenance costs, improve soft ware quality, and enforce discipline. Some project teams
even use CASE to assess the magnitude of changes to the project. Many modern CASE tools
that support object-oriented systems development support a development technique known
as round-trip engineering. Round-trip engineering supports not only code generation but also
the reverse engineering of UML diagrams from code. In this way, the system can evolve via
diagrams and via code in a round-trip manner.
Of course, like anything else, CASE should not be considered a silver bullet for project
development. Th e advanced CASE tools are complex applications that require signifi cant
training and experience to achieve real benefi ts. Our experience has shown that CASE is a
helpful way to support the communication and sharing of project diagrams and technical
specifi cations as long as it is used by trained developers who have applied CASE on past pro-
jects. All CASE tools use a CASE repository to store diagrams, models, and I/O designs and to
ensure consistency across iterations.
Standards
Project team members need to work together, and most project management soft ware and
CASE tools support them by providing access privileges to everyone working on the system.
However, without set procedures, collaboration can result in confusion. To make matters worse,
• Clearly defi ne plans for the project.
• Make sure that the team understands how the project is important to the organization.
• Develop detailed operating procedures and communicate these to the team members.
• Develop a project charter.
• Develop schedule commitments ahead of time.
• Forecast other priorities and their possible impact on the project.
Source: H. J. Thamhain and D. L. Wilemon, “Confl ict Management in Project Life Cycles,” Sloan Manage-
ment Review (Spring 1975).
FIGURE 2-23
Confl ict-Avoidance
Strategies
Environment and Infrastructure Management 77
people sometimes are reassigned in the middle of a project. It is important that their project
knowledge does not leave with them and that their replacements can get up to speed quickly.
One way to make certain that everyone is performing tasks in the same way and following
the same procedures is to create standards that the project team must follow. Standards can
include formal rules for naming fi les, forms that must be completed when goals are reached,
and programming guidelines. Figure 2-24 shows some examples of the types of standards that
a project can create. When a team forms standards and then follows them, the project can be
completed faster because task coordination becomes less complex.
Standards work best when they are created at the beginning of each major phase of the
project and communicated clearly to the entire project team. As the team moves forward, new
standards are added when necessary. Some standards (e.g., fi le naming conventions, status
reporting) are applied during the entire development process, whereas others (e.g., program-
ming guidelines) are appropriate only for certain tasks.
Documentation
Finally, during the inception phase of the infrastructure workfl ow, project teams establish
good documentation standards that include detailed information about the tasks of the Unifi ed
Process. Typically, the standards for the required documentation are set by the development
organization. Th e development team only needs to ascertain which documentation standards
are appropriate for the current systems development project. Oft en, the documentation is
stored in a project binder(s) that contains all the deliverables and all the internal communication
FIGURE 2-24
A Sampling of
Project Standards
Documentation standards The date and project name should appear as a header on
all documentation.
All margins should be set to 1 inch.
All deliverables should be added to the project binder and
recorded in its table of contents.
Coding standards All modules of code should include a header that lists the
programmer, last date of update, and a short description of the
purpose of the code.
Indentation should be used to indicate loops, if-then-else
statements, and case statements.
On average, every program should include one line of
comments for every fi ve lines of code.
Procedural standards Record actual task progress in the work plan every Monday
morning by 10 AM.
Report to project update meeting on Fridays at 3:30 PM.
All changes to a requirements document must be approved
by the project manager.
Specifi cation requirement standards Name of program to be created
Description of the program’s purpose
Special calculations that need to be computed
Business rules that must be incorporated into the program
Pseudocode
Due date
User interface design standards Labels will appear in boldface text, left-justifi ed, and followed by a colon.
The tab order of the screen will move from top left to bottom right.
Accelerator keys will be provided for all updatable fi elds.
Types of Standards Examples
78 Chapter 2 Project Management
that takes place—the history of the project. Th e good news is that Unifi ed Process has a set
of standard documentation that is expected. Th e documentation typically includes the system
request, the feasibility analysis, the original and later versions of the eff ort estimation, the
evolving workplan, and UML diagrams for the functional, structural, and behavioral models.
A poor project management practice is waiting until the last minute to create documentation;
this typically leads to an undocumented system that no one understands. Good project teams learn
to document a system’s history as it evolves while the details are still fresh in their memory. In most
CASE tools that support object-oriented systems development, some of the documentation can be
automated. For example, if the programming language chosen to implement the system in is Java,
then it is possible to automatically create HTML manual pages that will describe the classes being
implemented. Th is is accomplished through the javadoc18 tool that is part of the Java development
environment. Other tools enable the developer to automatically generate HTML documentation
for the UML diagrams, e.g., umldoc, which is part of the Poseidon for UML CASE tool.19 Even
though virtually all developers hate creating documentation and documentation takes valuable
time, it is a good investment that will pay off in the long run.
18 See Oracle, Javadoc Tool. Retrieved May 2014 from www.oracle.com. www.oracle.com/technetwork/java/javase/
documentation/index-jsp-135444.html.
19 See Gentleware, umldoc, an overview, retrieved May 2014 from /www.gentleware.com. www.gentleware.com/
fi leadmin/media/viewlets/text/UMLdoc.viewlet/UMLdoc_viewlet_swf.html.
Environment and Infrastructure Management 79
As Seattle University’s David Umphress has pointed
out, watching most organizations develop systems is
like watching reruns of Gilligan’s Island. At the begin-
ning of each episode, someone comes up with a
cockamamie scheme to get off the island, and it seems
to work for a while, but something goes wrong and
the castaways fi nd themselves right back where they
started—stuck on the island. Similarly, most companies
start new projects with grand ideas that seem to work,
only to make a classic mistake and deliver the project
behind schedule, over budget, or both. Here we sum-
marize four classic mistakes in the planning and project
management aspects of the project and discuss how to
avoid them:
1. Overly optimistic schedule: Wishful thinking can lead
to an overly optimistic schedule that causes analysis
and design to be cut short (missing key requirements)
and puts intense pressure on the programmers, who
produce poor code (full of bugs).
Solution: Don’t infl ate time estimates; instead,
explicitly schedule slack time at the end of each
phase to account for the variability in estimates.
2. Failing to monitor the schedule: If the team does not
regularly report progress, no one knows if the project
is on schedule.
Solution: Require team members to report progress (or
the lack of progress) honestly every week. There is no
penalty for reporting a lack of progress, but there are
immediate sanctions for a misleading report.
3. Failing to update the schedule: When a part of the
schedule falls behind (e.g., information gathering
uses all the slack in item 1 plus 2 weeks), a project
team often thinks it can make up the time later by
working faster. It can’t. This is an early warning that
the entire schedule is too optimistic.
Solution: Immediately revise the schedule and inform
the project sponsor of the new end date or use time-
boxing to reduce functionality or move it into future
versions.
4. Adding people to a late project: When a project
misses a schedule, the temptation is to add more
people to speed it up. This makes the project take
longer because it increases coordination problems
and requires staff to take time to explain what has
already been done.
Solution: Revise the schedule, use timeboxing, throw
away bug-fi lled code, and add people only to work
on an isolated part of the project.
Based upon Steve McConnell, Rapid Development (Redmond, WA:
Microsoft Press, 1996), pp. 29–50.
Avoiding Classic Planning MistakesPRACTICAL
TIP
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Explain the ways that projects are identifi ed and initiated.
Explain why it is important to link the information system to business needs of the organization.
Describe the purpose of the systems request and explain the contents of its sections.
Create a systems request for a proposed project.
Discuss the purpose of the feasibility study.
Describe the issues that are considered when evaluating a project’s technical feasibility.
Develop an economic feasibility assessment for a project.
Understand and evaluate the organizational feasibility of a project.
Explain how projects are selected.
Describe a task.
Create a standard work breakdown structure, a Gantt Chart, and a Network Diagram.
Perform PERT analysis and identify the critical path.
Estimate the system development eff ort using use-case points.
Create an evolutionary work breakdown structure.
Describe how iterative and incremental development using timeboxing addresses scope management.
Describe the characteristics of a “jelled” team.
Describe issues relating to motivating soft ware developers.
Describe the importance of CASE tools, standards, and documentation managing soft ware development projects.
KEY TERMS
Actor
Adjusted use-case
points (UCP)
Application program
interface (API)
Approval committee
Average actors
Average use case
Break-even point
Business need
Business requirement
Business value
Cash fl ow method
Champion
Compatibility
Complex actors
Complex use case
Computer-aided soft ware
engineering (CASE)
CASE repository
Cost–benefi t analysis
Critical path method
Critical task
Development costs
Documentation
Economic feasibility
Eff ort
Emerging Technology
Environmental factor (EF)
Environmental factor value
(EFactor)
Estimation
80 Chapter 2 Project Management
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
In Chapter 2, we look more closely at the completed system request that Max Ross and
his team develop for the Integrated Health Clinic Delivery system, including the business
needs, business requirements, and business values and constraints. We will also examine
the feasibility analysis that accompanies and justifi es the system request. Finally, we will
examine how the project eff ort was estimated, see how the project will be staff ed and
managed, and look at the Evolutionary Work Breakdown Structure for Version 1 of the
Integrated Health Clinic Delivery System.
As we progress through the text, examining how Patterson navigates through the sys-
tems analysis and design and development processes will help us understand real-world
implementation of the concepts presented.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
Evolutionary WBS
Familiarity with the
functional area
Familiarity with the
technology
Feasibility analysis
Feasibility study
First mover
Functional lead
Functionality
Gantt chart
Group cohesiveness
Intangible benefi ts
Intangible costs
Intangible value
Iterative workplan
Interpersonal skills
Methodology
Milestone
Motivation
Net present value (NPV)
Network Diagram
Node
Operational costs
Organizational feasibility
Organizational management
Person-hours multiplier
(PHM)
Program evaluation and
review technique (PERT)
Portfolio management
Project
Project binder
Project charter
Project initiation
Project management
Project management
soft ware
Project manager
Project size
Project sponsor
Reporting structure
Return on investment
(ROI)
Risk assessment
Risk management
Risks
Round-trip engineering
Scope creep
Simple actors
Simple use case
Special issues
Staffi ng plan
Stakeholder
Stakeholder analysis
Standards
Strategic alignment
System request
System users
Tangible benefi ts
Tangible value
Task
Task dependency
Technical complexity factor
(TCF)
Technical factor value
(TFactor)
Technical feasibility
Technical lead
Technical risk analysis
Technical skills
Timeboxing
Trade-off s
Unadjusted actor weight
total (UAW)
Unadjusted use-case points
(UUCP)
Unadjusted use-case weight
total (UUCW)
Use case
Use-case points
Work breakdown structure
(WBS)
Workplan
QUESTIONS
1. Give three examples of business needs for a system.
2. What is the purpose of an approval committee? Who
is usually on this committee?
3. Why should the system request be created by a busi-
ness person as opposed to an IS professional?
4. What is the diff erence between intangible value and
tangible value? Give three examples of each.
5. What are the purposes of the system request and the
feasibility analysis? How are they used in the project
selection process?
6. Describe two special issues that may be important to
list on a system request.
7. Describe the three techniques for feasibility analysis.
8. Describe a risky project in terms of technical feasibil-
ity. Describe a project that would not be considered
risky.
9. What are the steps for assessing economic feasibility?
Describe each step.
10. List two intangible benefi ts. Describe how these bene-
fi ts can be quantifi ed.
11. List two tangible benefi ts and two operational costs for
a system. How would you determine the values that
should be assigned to each item?
12. Explain the net present value and return on invest-
ment for a cost–benefi t analysis. Why would these
calculations be used?
13. What is the break-even point for the project? How is
it calculated?
14. What is stakeholder analysis? Discuss three stakehold-
ers that would be relevant for most projects.
15. Why do many projects end up having unreasonable
deadlines? How should a project manager react to
unreasonable demands?
16. What are the trade-off s that project managers must
manage?
17. Compare and contrast the Gantt chart with the net-
work diagram.
18. Some companies hire consulting fi rms to develop the
initial project plans and manage the project but use
their own analysts and programmers to develop the
system. Why do you think some companies do this?
19. What is a use-case point? For what is it used?
20. What process do we use to estimate systems develop-
ment based on use cases?
21. Name two ways to identify the tasks that need to be
accomplished over the course of a project.
22. What are the problems associated with conventional
WBSs?
23. What is an evolutionary WBS? How does it address
the problems associated with a conventional WBS?
24. What is an iterative workplan?
25. What is scope creep, and how can it be managed?
Questions 81
82 Chapter 2 Project Management
EXERCISES
A. Locate a news article in an IT trade magazine (e.g.,
Computerworld) about an organization that is imple-
menting a new computer system. Describe the tangi-
ble and intangible value that the organization is likely
to realize from the new system.
B. Car dealers have realized how profi table it can be to
sell automobiles using the Web. Pretend that you work
for a local car dealership that is part of a large chain
such as CarMax. Create a system request you might
use to develop a Web-based sales system. Remember
to list special issues that are relevant to the project.
C. Suppose that you are interested in buying a new com-
puter. Create a cost–benefi t analysis that illustrates
the return on investment that you would receive
from making this purchase. Computer-related web-
sites (e.g., Apple, Dell, HP) should have real tangible
costs that you can include in your analysis. Project
your numbers out to include a three-year period and
provide the net present value of the fi nal total.
D. Th e Amazon.com website originally sold books; then
the management of the company decided to extend
their Web-based system to include other products.
How would you have assessed the feasibility of this
venture when the idea fi rst came up? How risky would
you have considered the project that implemented this
idea? Why?
E. Interview someone who works in a large organization
and ask him or her to describe the approval process
that exists for approving new development projects.
What do they think about the process? What are the
problems? What are the benefi ts?
F. Visit a project management website, such as the Project
Management Institute (www.pmi.org). Most have links
to project management soft ware products, white papers,
and research. Examine some of the links for project
management to better understand a variety of Internet
sites that contain information related to this chapter.
G. Select a specifi c project management topic such as
CASE, project management soft ware, or timeboxing
and search for information on that topic using the
Web. Any search engine (e.g., Bing, Google) can pro-
vide a starting point for your eff orts.
H. Pretend that the career services offi ce at your univer-
sity wants to develop a system that collects student
résumés and makes them available to students and
recruiters over the Web. Students should be able to
input their résumé information into a standard résumé
template. Th e information then is presented in a
résumé format, and it also is placed in a database that
can be queried using an online search form. You have
been put in charge of the project. Develop a plan for
estimating the project. How long do you think it would
take for you and three other students to complete the
project? Provide support for the schedule that you
propose.
I. Refer to the situation in exercise H. You have been
told that recruiting season begins a month from today
and that the new system must be used. How would
you approach this situation? Describe what you can
do as the project manager to make sure that your team
does not burn out from unreasonable deadlines and
commitments.
J. Consider the system described in exercise H. Create
a workplan listing the tasks that will need to be com-
pleted to meet the project’s objectives. Create a Gantt
chart and a network diagram in a project management
tool (e.g., Microsoft Project) or using a spreadsheet
package to graphically show the high-level tasks of the
project.
K. Suppose that you are in charge of the project that is
described in exercise H and the project will be staff ed
by members of your class. Do your classmates have
all the right skills to implement such a project? If not,
how will you go about making sure that the proper
skills are available to get the job done?
L. Complete a use-case point worksheet to estimate the
eff ort to build the system described in exercises H, I, J,
and K. You will need to make assumptions regarding
the actors, the use cases, and the technical complexity
and environmental factors.
26. What is timeboxing, and why is it used?
27. Create a list of potential risks that could aff ect the
outcome of a project.
28. Describe the diff erences between a technical lead and
a functional lead. How are they similar?
29. Describe three technical skills and three interpersonal
skills that are very important to have on any project.
30. What are the best ways to motivate a team? What are
the worst ways?
31. List three techniques to reduce confl ict.
32. Describe three types of standards and provide exam-
ples of each.
33. What belongs in the project binder? How is the pro-
ject binder organized?
M. Consider the application that is used at your school to
register for classes. Complete a use-case point work-
sheet to estimate the eff ort to build such an applica-
tion. You will need to make some assumptions about
the application’s interfaces and the various factors that
aff ect its complexity.
N. Pretend that your instructor has asked you and two
friends to create a Web page to describe the course to
potential students and provide current class informa-
tion (e.g., syllabus, assignments, readings) to current
students. You have been assigned the role of leader, so
you will need to coordinate your activities and those
of your classmates until the project is completed.
Describe how you would apply the project manage-
ment techniques that you have learned in this chapter
in this situation. Include descriptions of how you
would create a workplan, staff the project, and coordi-
nate all activities—yours and those of your classmates.
O. Select two project management soft ware packages
and research them using the Web or trade magazines.
Describe the features of the two packages. If you were
a project manager, which one would you use to help
support your job? Why?
P. In 1997, Oxford Health Plans had a computer problem
that caused the company to overestimate revenue and
underestimate medical costs. Problems were caused
by the migration of its claims processing system from
the Pick operating system to a UNIX-based system
that uses Oracle database soft ware and hardware from
Pyramid Technology. As a result, Oxford’s stock price
plummeted, and fi xing the system became the number
one priority for the company. Suppose that you have
been placed in charge of managing the repair of the
claims processing system. Obviously, the project team
will not be in good spirits. How will you motivate
team members to meet the project’s objectives?
MINICASES
1. Th e Amberssen Specialty Company is a chain of twelve
retail stores that sell a variety of imported gift items,
gourmet chocolates, cheeses, and wines in the Toronto
area. Amberssen has an IS staff of three people who
have created a simple but eff ective information system
of networked point-of-sale registers at the stores and
a centralized accounting system at the company head-
quarters. Harry Hilman, the head of Amberssens IS
group, has just received the following memo from Bill
Amberssen, Sales Director (and son of Amberssen’s
founder).
Harry—it’s time Amberssen Specialty launched
itself on the Internet. Many of our competitors
are already there, selling to customers without the
expense of a retail storefront, and we should be
there too. I project that we could double or triple
our annual revenues by selling our products on
the Internet. I’d like to have this ready by Th anks-
giving, in time for the prime holiday gift -shopping
season. Bill
Aft er pondering this memo for several days, Harry
scheduled a meeting with Bill so that he could clarify
Bill’s vision of this venture. Using the standard con-
tent of a system request as your guide, prepare a list
of questions that Harry needs to have answered about
this project.
2. Th e Decker Company maintains a fl eet of ten service
trucks and crews that provide a variety of plumbing,
heating, and cooling repair services to residential cus-
tomers. Currently, it takes on average about six hours
before a service team responds to a service request.
Each truck and crew averages twelve service calls per
week, and the average revenue earned per service call
is $150. Each truck is in service fi ft y weeks per year.
Owing to the diffi culty in scheduling and routing,
there is considerable slack time for each truck and
crew during a typical week.
In an eff ort to more effi ciently schedule the trucks
and crews and improve their productivity, Decker
management is evaluating the purchase of a prewritten
routing and scheduling soft ware package. Th e benefi ts
of the system will include reduced response time to
service requests and more productive service teams,
but management is having trouble quantifying these
benefi ts.
One approach is to make an estimate of how much
service response time will decrease with the new system,
which then can be used to project the increase in the
number of service calls made each week. For example, if
the system permits the average service response time to
fall to four hours, management believes that each truck
will be able to make sixteen service calls per week on
average—an increase of four calls per week. With each
truck making four additional calls per week and the
average revenue per call at $150, the revenue increase
per truck per week is $600 (4 3 $150). With ten trucks in
service fi ft y weeks per year, the average annual revenue
increase will be $300,000 ($600 3 10 3 50).
Minicases 83
Decker Company management is unsure whether
the new system will enable response time to fall to four
hours on average or if it will be some other number.
Th erefore, management has developed the following
range of outcomes that may be possible outcomes of
the new system, along with probability estimates of
each outcome’s occurring.
New Response Time # Calls/Truck/Week Likelihood
2 hours 20 20%
3 hours 18 30%
4 hours 16 50%
Given these fi gures, prepare a spreadsheet model that
computes the expected value of the annual revenues to
be produced by this new system.
3. Emily Pemberton is an IS project manager facing a dif-
fi cult situation. Emily works for the First Trust Bank,
which has recently acquired the City National Bank.
Before the acquisition, First Trust and City National
were bitter rivals, fi ercely competing for market share
in the region. Following the acrimonious takeover,
numerous staff were laid off in many banking areas,
including IS. Key individuals were retained from both
banks’ IS areas, however, and were assigned to a new
consolidated IS department. Emily has been made pro-
ject manager for the fi rst signifi cant IS project since the
takeover, and she faces the task of integrating staff ers
from both banks on her team. Th e project they are
undertaking will be highly visible within the organi-
zation, and the time frame for the project is somewhat
demanding. Emily believes that the team can meet the
project goals successfully, but success will require that
the team become cohesive quickly and that potential
confl icts be avoided. What strategies do you suggest
that Emily implement in order to help ensure a suc-
cessfully functioning project team?
4. Tom, Jan, and Julie are IS majors at Great State Uni-
versity. Th ese students have been assigned a class
project by one of their professors, requiring them
to develop a new Web-based system to collect and
update information on the IS program’s alumni. Th is
system will be used by the IS graduates to enter job
and address information as they graduate and then
make changes to that information as they change jobs
and/or addresses. Th eir professor also has a number
of queries that she is interested in being able to imple-
ment. Based on their preliminary discussions with
their professor, the students have determined that the
only actor is an IS graduate. Th ey identifi ed one sim-
ple use case, four average use cases, and two complex
use cases. You need to assign reasonable values to
each of the technical complexity and environmental
factors. Calculate the eff ort for this project.
5. In looking for a capstone project for your fi nal MIS
course, you found a possible project. Th e master gar-
deners in Blint County have created a database of all
of the plants in their arboretum. Th e database is actu-
ally a spreadsheet created by one of the volunteers.
Along with providing a plant inventory, it is used to
print labels of all of the plants that the other master
gardeners grow for the annual plant. More than
5,000 plants are supplied each year by 100 garden-
ers from their home gardens. Because the type and
numbers of plants change each year and because the
members e-mail the information in varying formats,
label printing has become an onerous task. Pam,
who prints the labels each year, wants help in mak-
ing this task manageable. She provided an example
of a typical email as well as the type of information
she needs.
E-mail
Lilies—labels needed 32–
Lilium lancifolium / lilium tigrinum
Tiger Lily perennial light shade 4’
Ice plant (pink)—labels needed 3
Delosperma cooperi Hardy Ice Plant succulent
full sun 2–5”
Information for Labels
Botanical Name
Common Name
Plant Type
Light Requirement
Height and Width
In order to have this accepted as your project, you
need to form a team with the necessary skills and to
create a systems request. How would you approach
this project? What additional information do you
need from Pam in order to begin estimating the scope
of this project? Assuming that you have received this
information, create a systems request. Also create a
list of skills needed, the number of team members
required, and a project plan.
84 Chapter 2 Project Management
Analysis modeling answers the questions of who will use the system,
what the system will do, and where and when it will be used. During
analysis, detailed requirements are identifi ed and a system proposal
is created. Th e team then produces the functional model (use-case
diagram, activity diagrams, and use-case descriptions), structural
model (CRC cards and class diagram, and object diagrams), and
behavioral models (sequence diagrams, communication diagrams,
behavioral state machines, and a CRUDE matrix).
CHAPTER 3
Requirements
Determination
CHAPTER 4
Business Process
and Functional
Modeling
CHAPTER 5
Structural
Modeling
CHAPTER 6
Behavioral
Modeling
R
equirem
ents
D
efi nition
System
Proposal
U
se-C
ase
D
escriptions
O
bject
D
iagram
s
C
rude
M
atrix
U
se-C
ase
D
iagram
s
A
ctivity
D
iagram
s
C
R
C
C
ards
C
lass
D
iagram
s
Sequence
D
iagram
s
C
om
m
unication
D
iagram
s
B
ehavioral
State M
achines
P A R T O N E
Analysis Modeling
86
One of the fi rst activities of an analyst is to determine the business requirements for a new
system. Th is chapter begins by presenting the requirements defi nition, a document that lists
the new system’s capabilities. It then describes how to analyze requirements using require-
ments analysis strategies and how to gather requirements using interviews, JAD sessions,
questionnaires, document analysis, and observation. Th e chapter also describes a set of alter-
native requirements-documentation techniques and describes the system proposal document
that pulls everything together.
OBJECTIVES
■ Understand how to create a requirements defi nition
■ Become familiar with requirements-analysis techniques
■ Understand when to use each requirements-analysis technique
■ Understand how to gather requirements using interviews, JAD sessions, questionnaires,
document analysis, and observation
■ Understand the use of concept maps, story cards, and task lists as requirements-
documentation techniques
■ Understand when to use each requirements-gathering technique
■ Be able to begin creating a system proposal
INTRODUCTION
Th e systems development process aids an organization in moving from the current system
(oft en called the as-is system) to the new system (oft en called the to-be system). Th e output of
planning, discussed in Chapter 2, is the system request, which provides general ideas for the
to-be system, defi nes the project’s scope, and provides the initial workplan. Analysis takes the
general ideas in the system request and refi nes them into a detailed requirements defi nition
(this chapter), functional models (Chapter 4), structural models (Chapter 5), and behavioral
models (Chapter 6) that together form the system proposal. Th e system proposal also includes
revised project management deliverables, such as the feasibility analysis and the workplan
(Chapter 2).
Th e output of analysis, the system proposal, is presented to the approval committee, who
decides if the project is to continue. If approved, the system proposal moves into design, and
its elements (requirements defi nition and functional, structural, and behavioral models) are
used as inputs to the steps in design. Th is further refi nes them and defi nes in much more
detail how the system will be built.
Th e line between analysis and design is very blurry. Th is is because the deliverables
created during analysis are really the fi rst step in the design of the new system. Many of
the major design decisions for the new system are found in the analysis deliverables. It is
C H A P T E R 3
Requirements Determination
Requirements Determination 87
important to remember that the deliverables from analysis are really the fi rst step in the
design of the new system.
In many ways, because it is here that the major elements of the system fi rst emerge, the
requirements-determination step is the single most critical step of the entire system devel-
opment process. During requirements determination, the system is easy to change because
little work has been done yet. As the system moves through the system development process,
it becomes harder and harder to return to requirements determination and to make major
changes because of all of the rework that is involved. Several studies have shown that more
than half of all system failures are due to problems with the requirements.1 Th is is why the
iterative approaches of object-oriented methodologies are so eff ective—small batches of
requirements can be identifi ed and implemented in incremental stages, allowing the overall
system to evolve over time.
REQUIREMENTS DETERMINATION
Th e purpose of requirements determination is to turn the very high-level explanation of
the business requirements stated in the system request into a more precise list of require-
ments that can be used as inputs to the rest of analysis (creating functional, structural, and
behavioral models). Th is expansion of the requirements ultimately leads to the design of
the system.
Defi ning a Requirement
A requirement is simply a statement of what the system must do or what characteristic it
must have. During analysis, requirements are written from the perspective of the busi-
nessperson, and they focus on the “what” of the system. Because they focus on the needs
of the business user, they are usually called business requirements (and sometimes user
requirements). Later in design, business requirements evolve to become more technical,
and they describe how the system will be implemented. Requirements in design are writ-
ten from the developer’s perspective, and they are usually called system requirements.
We want to stress that there is no black-and-white line dividing a business requirement
and a system requirement—and some companies use the terms interchangeably. Th e impor-
tant thing to remember is that a requirement is a statement of what the system must do,
and requirements will change over time as the project moves from inception to elaboration
to construction. Requirements evolve from detailed statements of the business capabilities
that a system should have to detailed statements of the technical way the capabilities will be
implemented in the new system.
Requirements can be either functional or nonfunctional in nature. A functional require-
ment relates directly to a process a system has to perform or information it needs to contain.
For example, requirements stating that a system must have the ability to search for available
inventory or to report actual and budgeted expenses are functional requirements. Functional
requirements fl ow directly into the creation of functional, structural, and behavioral models
that represent the functionality of the evolving system (see Chapters 4, 5, and 6).
Nonfunctional requirements refer to behavioral properties that the system must have,
such as performance and usability. Th e ability to access the system using a Web browser is
considered a nonfunctional requirement. Nonfunctional requirements can infl uence the rest
of analysis (functional, structural, and behavioral models) but oft en do so only indirectly;
nonfunctional requirements are used primarily in design when decisions are made about the
database, the user interface, the hardware and soft ware, and the system’s underlying physical
architecture.
1 For example, see Th e Scope of Soft ware Development Project Failures (Dennis, MA: Th e Standish Group, 1995).
88 Chapter 3 Requirements Determination
Nonfunctional requirements describe a variety of characteristics regarding the system:
operational, performance, security, and cultural and political. Operational requirements
address issues related to the physical and technical requirements in which the system will
operate. Performance requirements address issues related to the speed, capacity, and reli-
ability of the system. Security requirements deal with issues with regard to who has access
to the system and under what specifi c circumstances. Cultural and political requirements
deal with issues related to the cultural, political factors and legal requirements that aff ect the
system. Th ese characteristics do not describe business processes or information, but they
are very important in understanding what the fi nal system should be like. Nonfunctional
requirements primarily aff ect decisions that will be made during the design of a system. We
will return to this topic later in the book when we discuss design (see Chapters 9, 10, and 11).
One area of information systems development that focused on diff erentiating functional
and nonfunctional requirements is soft ware quality. Th ere have been many diff erent models
proposed to measure the quality of soft ware. However, virtually all of them diff erentiate func-
tional and nonfunctional requirements. From a quality perspective, functional quality is related
to the degree that the soft ware meets the functional requirements, i.e., how much of the actual
problem is solved by the soft ware solution provided. Whereas, the nonfunctional requirements
are associated with the effi ciency, maintainability, portability, reliability, reusability, testability,
and usability quality dimensions. As stated above, the nonfunctional related dimensions are
associated primarily with the actual detailed design and implementation of the system.
When considering ISO 9000 compliance, quality dimensions are further decomposed into
those that the user can see (external) and those that the user cannot see (internal). Th e external
nonfunctional dimensions include effi ciency, reliability, and usability, whereas the internal
nonfunctional dimensions include maintainability, portability, reusability, and testability.
From a user perspective, the external dimensions are more important. If the system is simply
too diffi cult to use, regardless how well the system solves the problem, the user will simply
not use the system. In other words, from a user’s perspective, for an information system to be
successful, the system must not only meet the functional specifi cation, but it must also meet
the external nonfunctional specifi cations. From a developer perspective, the internal dimen-
sions are also important. For example, given that successful systems tend to be long-lived and
multiplatform, both the maintainability and portability dimensions can have strategic implica-
tions for the system being developed. Also, given the agile development approaches being used
in industry today, the development of reusable and testable soft ware is crucial.
Th ree additional topics that have infl uenced information system requirements are the
Sarbanes-Oxley Act, COBIT (Control OBjectives for Information and related Technology)
compliance and Capability Maturity Model compliance. Depending on the system being con-
sidered, these three topics could aff ect the defi nition of a system’s functional requirements,
nonfunctional requirements, or both. Th e Sarbanes-Oxley Act, for example, mandates addi-
tional functional and nonfunctional requirements. Th ese include additional security concerns
(nonfunctional) and specifi c information requirements that management must now provide
(functional). When developing fi nancial information systems, information system developers
should be sure to include Sarbanes-Oxley expertise in the development team. Moreover, a client
could insist on COBIT compliance or that a specifi c Capability Maturity Model level had been
reached in order for the fi rm to be considered as a possible vendor to supply the system under
consideration. Obviously, these types of requirements add to the nonfunctional requirements.
Further discussion of these topics is beyond the scope of this book.2
2 A concise discussion of the Sarbanes-Oxley Act is presented in G. P. Lander, What is Sarbanes-Oxley? (New York:
McGraw-Hill, 2004). A good reference for Sarbanes-Oxley Act-based security requirements is D. C. Brewer, Security
Controls for Sarbanes-Oxley Section 404 IT Compliance: Authorization, Authentication, and Access (Indianapolis, IN:
Wiley, 2006). For detailed information on COBIT, see www.isaca.org; for ISO 9000, see www.iso.org; and for details
on the Capability Maturity Model, see www.sei.cmu.edu/cmmi/.
Requirements Determination 89
Another recent topic that infl uences requirements for some systems is globalization. For
example, a global information supply chain generates a large number of additional nonfunc-
tional requirements. If the necessary operational environments do not exist for a mobile solu-
tion to be developed, it is important to adapt the solution to the local environment. Or, it may
not be reasonable to expect to deploy a high-technology-based solution in an area that does not
have the necessary power and communications infrastructure. In some cases, we may need to
consider supporting some parts of the global information supply chain with manual—rather
than automated—systems.
Manual systems have an entirely diff erent set of requirements that create diff erent per-
formance expectations and additional security concerns. Furthermore, cultural and political
concerns are potentially paramount. A simple example that aff ects the design of user inter-
faces is the proper use of color on forms (on a screen or paper). Diff erent cultures interpret
diff erent colors diff erently. In other words, in a global, multicultural business environment,
addressing cultural concerns goes well beyond simply having a multilingual user interface.
We must be able to adapt the global solution to the local realities. Friedman refers to these
concerns as glocalization.3 Otherwise, we will simply create another example of a failed infor-
mation system development project.
Requirements Defi nition
Th e requirements defi nition report—usually just called the requirements defi nition—is a
straightforward text report that simply lists the functional and nonfunctional requirements
in an outline format. Figure 3-1 shows a sample requirements defi nition for an appointment
system for a typical doctor’s offi ce. Notice it contains both functional and nonfunctional
requirements. Th e functional requirements include managing appointments, producing
schedules, and recording the availability of the individual doctors. Th e nonfunctional require-
ments include items such as the expected amount of time that it takes to store a new appoint-
ment, the need to support wireless printing, and which types of employees have access to the
diff erent parts of the system.
Th e requirements are numbered in a legal or outline format so that each requirement
is clearly identifi ed. Th e requirements are fi rst grouped into functional and nonfunctional
requirements; within each of those headings, they are further grouped by the type of nonfunc-
tional requirement or by function.
Sometimes business requirements are prioritized on the requirements defi nition. Th ey
can be ranked as having high, medium, or low importance in the new system, or they can
be labeled with the version of the system that will address the requirement (e.g., release 1,
release 2, release 3). Th is practice is particularly important when using object-oriented meth-
odologies since they deliver systems in an incremental manner.
Th e most obvious purpose of the requirements defi nition is to provide the information
needed by the other deliverables in analysis, which include functional, structural, and behav-
ioral models, and to support activities in design. Th e most important purpose of the require-
ments defi nition, however, is to defi ne the scope of the system. Th e document describes to
the analysts exactly what the system needs to end up doing. When discrepancies arise, the
document serves as the place to go for clarifi cation.
Determining Requirements
Determining requirements for the requirements defi nition is both a business task and an
information technology task. In the early days of computing, there was a presumption that
3 T. L. Friedman, Th e World is Flat: A Brief History of the Twenty-First Century, Updated and Expanded Edition. (New
York: Farrar, Straus, and Giroux, 2006.)
90 Chapter 3 Requirements Determination
the systems analysts, as experts with computer systems, were in the best position to defi ne
how a computer system should operate. Many systems failed because they did not adequately
address the true business needs of the users. Gradually, the presumption changed so that the
users, as the business experts, were seen as being the best position to defi ne how a computer
system should operate. However, many systems failed to deliver performance benefi ts because
users simply automated an existing ineffi cient system, and they failed to incorporate new
opportunities off ered by technology.
Th erefore, the most eff ective approach is to have both business people and analysts
working together to determine business requirements. Sometimes, however, users don’t
know exactly what they want, and analysts need to help them discover their needs. A set of
strategies has become popular to help analysts do problem analysis, root cause analysis, dura-
tion analysis, activity-based costing, informal benchmarking, outcome analysis, technology
analysis, and activity elimination. Analysts can use these tools when they need to guide the
users in explaining what is wanted from a system. Th ese strategies work similarly. Th ey help
users critically examine the current state of systems and processes (the as-is system), identify
exactly what needs to change, and develop a concept for a new system (the to-be system).
Functional Requirements
1. Manage Appointments
1.1. Patient makes new appointment.
1.2. Patient changes appointment.
1.3. Patient cancels appointment.
2. Produce Schedule
2.1. Office Manager checks daily schedule.
2.2. Office Manager prints daily schedule.
3. Record Doctor Availability
3.1. Doctor updates schedule
Nonfunctional Requirements
1. Operational Requirements
1.1. The system will operate in Windows environment.
1.2. The system should be able to connect to printers wirelessly.
1.3. The system should automatically back up at the end of each day.
2. Performance Requirements
2.1. The system will store a new appointment in 2 seconds or less.
2.2. The system will retrieve the daily appointment schedule in 2 seconds or less.
3. Security Requirements
3.1. Only doctors can set their availability.
3.2. Only a manager can produce a schedule.
4. Cultural and Political Requirements
4.1. No special cultural and political requirements are anticipated.
FIGURE 3-1
Sample Requirements
Defi nition
Requirements Determination 91
Although these strategies enable the analyst to help users create a vision for the new
system, they are not suffi cient for extracting information about the detailed business require-
ments that are needed to build it. Th erefore, analysts use a portfolio of requirements-gathering
techniques to acquire information from users. Th e analyst has many techniques from which
to choose: interviews, questionnaires, observation, joint application development (JAD), and
document analysis. Th e information gathered using these techniques is critically analyzed and
used to craft the requirements defi nition report.
Creating a Requirements Defi nition
Creating a requirements defi nition is an iterative and ongoing process whereby the analyst
collects information with requirements-gathering techniques (e.g., interviews, document
analysis), critically analyzes the information to identify appropriate business requirements
for the system, and adds the requirements to the requirements defi nition report. Th e require-
ments defi nition is kept up to date so that the project team and business users can refer to it
and get a clear understanding of the new system.
To create a requirements defi nition, the project team fi rst determines the kinds of func-
tional and nonfunctional requirements that they will collect about the system (of course, these
may change over time). Th ese become the main sections of the document. Next, the analysts
use a variety of requirements-gathering techniques to collect information, and they list the
business requirements that were identifi ed from that information. Finally, the analysts work
with the entire project team and the business users to verify, change, and complete the list and
to help prioritize the importance of the requirements that were identifi ed.
Th is process continues throughout analysis, and the requirements defi nition evolves
over time as new requirements are identifi ed and as the project moves into later phases of the
Unifi ed Process. Beware: Th e evolution of the requirements defi nition must be carefully man-
aged. Th e project team cannot keep adding to the requirements defi nition, or the system will
keep growing and growing and never get fi nished. Instead, the project team carefully identifi es
requirements and evaluates which ones fi t within the scope of the system. When a requirement
refl ects a real business need but is not within the scope of the current system or current release,
it is either added on a list of future requirements or given a low priority. Th e management of
requirements (and system scope) is one of the hardest parts of managing a project.
Real-World Problems with Requirements Determination
Avison and Fitzgerald provide us with a set of problems that can arise with regard to deter-
mining the set of requirements with which to be dealt.4 First, the analyst might not have
access to the correct set of users to uncover the complete set of requirements. Th is can lead to
requirements being missed, misrepresented, and/or overspecifi ed. Second, the specifi cation
of the requirements may be inadequate. Th is can be especially true with the lightweight tech-
niques associated with agile methodologies. Th ird, some requirements are simply unknowa-
ble at the beginning of a development process. However, as the system is developed, the users
and analysts will get a better understanding of both the domain issues and the applicable tech-
nology. Th is can cause new functional and nonfunctional requirements to be identifi ed and
current requirements to evolve or be canceled. Iterative and incremental-based development
methodologies, such as the Unifi ed Process and agile, can help in this case. Fourth, verifying
and validating of requirements can be very diffi cult. We take up this topic in the chapters
that deal with the creation of functional (Chapter 4), structural (Chapter 5), and behavioral
(Chapter 6) models.
4 See D. Avison and G. Fitzgerald, Information Systems Development: Methodologies, Techniques, & Tools, 4th Ed.
(London: McGraw-Hill, 2006).
92 Chapter 3 Requirements Determination
REQUIREMENTS ANALYSIS STRATEGIES
Before the project team can determine what requirements are appropriate for a given system,
there needs to be a clear vision of the kind of system that will be created and the level of
change that it will bring to the organization. Th e basic process of analysis is divided into three
steps: understanding the as-is system, identifying improvements, and developing require-
ments for the to-be system.
Sometimes the fi rst step (i.e., understanding the as-is system) is skipped or is performed in a
cursory manner. Th is happens when no current system exists, if the existing system and processes
are irrelevant to the future system, or if the project team is using a RAD or agile development
methodology in which the as-is system is not emphasized. Newer RAD, agile, and object-oriented
methodologies, such as phased development, prototyping, throwaway prototyping, extreme pro-
gramming, and Scrum (see Chapter 1) focus almost exclusively on improvements and the to-be
system requirements, and they spend little time investigating the current as-is system.
Requirements analysis strategies help the analyst lead users through the analysis steps
so that the vision of the system can be developed. Requirements analysis strategies and
requirements-gathering techniques go hand in hand. Analysts use requirements-gathering
techniques to collect information; requirements analysis strategies drive the kind of infor-
mation that is gathered and how it is ultimately analyzed. Th e requirements analysis strat-
egies and requirements gathering happen concurrently and are complementary activities.
To move the users from the as-is system to the to-be system, an analyst needs strong
critical thinking skills. Critical thinking is the ability to recognize strengths and weaknesses
and recast an idea in an improved form, and critical thinking skills are needed to really under-
stand issues and develop new business processes. Th ese skills are also needed to thoroughly
examine the results of requirements gathering, to identify business requirements, and to
translate those requirements into a concept for the new system.
Problem Analysis
Th e most straightforward (and probably the most commonly used) requirements-analysis
technique is problem analysis. Problem analysis means asking the users and managers to
identify problems with the as-is system and to describe how to solve them in the to-be
system. Most users have a very good idea of the changes they would like to see, and most
are quite vocal about suggesting them. Most changes tend to solve problems rather than
capitalize on opportunities, but the latter is possible as well. Improvements from problem
analysis tend to be small and incremental (e.g., provide more space in which to type the
customer’s address; provide a new report that currently does not exist).
Th is type of improvement oft en is very eff ective at improving a system’s effi ciency or
ease of use. However, it oft en provides only minor improvements in business value—the new
system is better than the old, but it may be hard to identify signifi cant monetary benefi ts from
the new system.
Root Cause Analysis
Th e ideas produced by problem analysis tend to be solutions to problems. All solutions make
assumptions about the nature of the problem, assumptions that might or might not be valid.
In our experience, users (and most people in general) tend to quickly jump to solutions with-
out fully considering the nature of the problem. Sometimes the solutions are appropriate, but
many times they address a symptom of the problem, not the true problem or root cause itself.5
5 Two good books that discuss the diffi culty in fi nding the root causes to problems are: E. M. Goldratt and
J. Cox, Th e Goal (Croton-on-Hudson, NY: North River Press, 1986); E. M. Goldratt, Th e Haystack Syndrome
(Croton-on-Hudson, NY: North River Press, 1990).
Requirements Analysis Strategies 93
For example, suppose a fi rm notices that its users report inventory stock-outs. Th e cost of
inventory stock-outs can be quite signifi cant. In this case, since they happen frequently, custom-
ers could fi nd another source for the items that they are purchasing from the fi rm. It is in the
fi rm’s interest to determine the underlying cause and not simply provide a knee-jerk reaction
such as arbitrarily increasing the amount of inventory kept on hand. In the business world,
the challenge lies in identifying the root cause—few real-world problems are simple. Th e users
typically propose a set of causes for the problem under consideration. Th e solutions that users
propose can address either symptoms or root causes, but without a careful analysis, it is diffi cult
to tell which one is addressed.
Root cause analysis, therefore, focuses on problems, not solutions. Th e analyst starts by
having the users generate a list of problems with the current system and then prioritize the
problems in order of importance. Starting with the most important, the users and/or the
analysts then generate all the possible root causes for the problems. Each possible root cause
is investigated (starting with the most likely or easiest to check) until the true root causes
are identifi ed. If any possible root causes are identifi ed for several problems, those should
be investigated fi rst, because there is a good chance they are the real root causes infl uencing
the symptom problems. In our example, there are several possible root causes:
■ Th e fi rm’s supplier might not be delivering orders to the fi rm in a timely manner.
■ Th ere could be a problem with the fi rm’s inventory controls.
■ Th e reorder level and quantities could be set wrong.
Sometimes, using a hierarchical chart to represent the causal relationships helps with the analysis.
As Figure 3-2 shows, there are many possible root causes that underlie the higher-level causes
identifi ed. Th e key point in root cause analysis is always to challenge the obvious.
Duration Analysis
Duration analysis requires a detailed examination of the amount of time it takes to perform
each process in the current as-is system. Th e analysts begin by determining the total amount
of time it takes, on average, to perform a set of business processes for a typical input. Th ey
then time each of the individual steps (or subprocesses) in the business process. Th e time to
Frequent
Inventory Stock-Outs
Order Approval
Late
Identifying Vendor
Delayed
Delay in Sending
Order to Vendor
Delays in Order
Processing
Late Recording of
Sales
Late Recording of
Purchases Received
Infrequent Manual
Inventory Reconciliation
Problems with
Inventory Controls
Reorder point set
too low
Reorder Quantity
(EOQ) set too low
Incorrect Reorder
Level and Quantities
FIGURE 3-2 Root Cause Analysis for Inventory Stock-Outs
complete the basic step is then totaled and compared to the total for the overall process. A
signifi cant diff erence between the two—and in our experience the total time oft en can be 10
or even 100 times longer than the sum of the parts—indicates that this part of the process is
badly in need of a major overhaul.
For example, suppose that the analysts are working on a home mortgage system and dis-
cover that on average, it takes thirty days for the bank to approve a mortgage. Th ey then look
at each of the basic steps in the process (e.g., data entry, credit check, title search, appraisal)
and fi nd that the total amount of time actually spent on each mortgage is about eight hours.
Th is is a strong indication that the overall process is badly broken, because it takes thirty days
to perform one day’s work.
Th ese problems probably occur because the process is badly fragmented. Many diff erent
people must perform diff erent activities before the process fi nishes. In the mortgage exam-
ple, the application probably sits on many people’s desks for long periods of time before it
is processed.
Processes in which many diff erent people work on small parts of the inputs are prime
candidates for process integration or parallelization. Process integration means changing the
fundamental process so that fewer people work on the input, which oft en requires changing
the processes and retraining staff to perform a wider range of duties. Process parallelization
means changing the process so that all the individual steps are performed at the same time.
For example, in the mortgage application case, there is probably no reason that the credit
check cannot be performed at the same time as the appraisal and title check.
Activity-Based Costing
Activity-based costing is a similar analysis; it examines the cost of each major process or step
in a business process rather than the time taken.6 Th e analysts identify the costs associated
with each of the basic functional steps or processes, identify the most costly processes, and
focus their improvement eff orts on them.
Assigning costs is conceptually simple. Analysts simply examine the direct cost of labor
and materials for each input. Materials costs are easily assigned in a manufacturing process,
whereas labor costs are usually calculated based on the amount of time spent on the input and
the hourly cost of the staff . However, as you may recall from a managerial accounting course,
there are indirect costs, such as rent, depreciation, and so on, that also can be included in
activity costs.
Informal Benchmarking
Benchmarking refers to studying how other organizations perform a business process in
order to learn how your organization can do something better. Benchmarking helps the
organization by introducing ideas that employees may never have considered but that have
the potential to add value.
Informal benchmarking is fairly common for customer-facing business processes (i.e.,
processes that interact with the customer). With informal benchmarking, the managers and
analysts think about other organizations or visit them as customers to watch how the business
process is performed. In many cases, the business studied may be a known leader in the indus-
try or simply a related fi rm.
6 Many books have been written on activity-based costing. Useful ones include K. B. Burk and D. W. Webster,
Activity-Based Costing (Fairfax, VA: American Management Systems, 1994); D. T. Hicks, Activity-Based Costing:
Making It Work for Small and Mid-sized Companies (New York: Wiley, 1998). Th e two books by Eli Goldratt men-
tioned previously (Th e Goal and Th e Haystack Syndrome) also off er unique insights into costing.
94 Chapter 3 Requirements Determination
Requirements-Gathering Techniques 95
Outcome Analysis
Outcome analysis focuses on understanding the fundamental outcomes that provide value to
customers. Although these outcomes sound as though they should be obvious, they oft en are
not. For example, consider an insurance company. One of its customers has just had a car
accident. What is the fundamental outcome from the customer’s perspective? Traditionally,
insurance companies have answered this question by assuming the customer wants to receive
the insurance payment quickly. To the customer, however, the payment is only a means to
the real outcome: a repaired car. Th e insurance company might benefi t by extending its view
of the business process past its traditional boundaries to include not paying for repairs but
performing the repairs or contracting with an authorized body shop to do them.
With this approach, system analysts encourage the managers and project sponsor to
pretend they are customers and to think carefully about what the organization’s products and
services enable the customers to do—and what they could enable the customer to do.
Technology Analysis
Many major changes in business since the turn of the century have been enabled by new
technologies. Technology analysis starts by having the analysts and managers develop a list
of important and interesting technologies. Th en the group systematically identifi es how
every technology could be applied to the business process and identifi es how the business
would benefi t. It is important to note the technology analysis in no way implies adopting
technology for technology’s sake. Rather the focus is on using new technologies to meet the
goals of the organization.
Activity Elimination
Activity elimination is exactly what it sounds like. Th e analysts and managers work together
to identify how the organization could eliminate each activity in the business process, how the
function could operate without it, and what eff ects are likely to occur. Initially, managers are
reluctant to conclude that processes can be eliminated, but this is a force-fi t exercise in that
they must eliminate each activity. In some cases, the results are silly; nonetheless, participants
must address every activity in the business process.
REQUIREMENTS-GATHERING TECHNIQUES
An analyst is very much like a detective (and business users are sometimes like elusive sus-
pects). He or she knows that there is a problem to be solved and therefore must look for clues
that uncover the solution. Unfortunately, the clues are not always obvious (and are oft en
missed), so the analyst needs to notice details, talk with witnesses, and follow leads just as
Sherlock Holmes would have done. Th e best analysts thoroughly gather requirements using a
variety of techniques and make sure that the current business processes and the needs for the
new system are well understood before moving into design. Analysts don’t want to discover
later that they have key requirements wrong—such surprises late in the development process
can cause all kinds of problems.
Th e requirements-gathering process is used for building political support for the pro-
ject and establishing trust and rapport between the project team building the system and
the users who ultimately will choose to use or not use the system. Involving someone in the
process implies that the project teams view that person as an important resource and value
his or her opinions. All the key stakeholders (the people who can aff ect the system or who
will be aff ected by the system) must be included in the requirements-gathering process. Th e
stakeholders might include managers, employees, staff members, and even some customers
and suppliers. If a key person is not involved, that individual might feel slighted, which can
cause problems during implementation (e.g., How could they have developed the system
without my input?).
Th e second challenge of requirements gathering is choosing the way(s) information is
collected. Th ere are many techniques for gathering requirements that vary from asking people
questions to watching them work. In this section, we focus on the fi ve most commonly used
techniques: interviews, JAD sessions (a special type of group meeting), questionnaires, docu-
ment analysis, and observation. Each technique has its own strengths and weaknesses, many of
which are complementary, so most projects use a combination of techniques.7
Interviews
An interview is the most commonly used requirements-gathering technique. Aft er all, it is
natural—if you need to know something, you usually ask someone. In general, interviews are
conducted one-on-one (one interviewer and one interviewee), but sometimes, owing to time
constraints, several people are interviewed at the same time. Th ere are fi ve basic steps to the inter-
view process: selecting interviewees, designing interview questions, preparing for the interview,
conducting the interview, and postinterview follow-up.8
Th e fi rst step in interviewing is to create an interview schedule listing who will be interviewed,
when, and for what purpose (see Figure 3-3). Th e schedule can be an informal list that is
used to help set up meeting times or a formal list that is incorporated into the workplan. Th e
people who appear on the interview schedule are selected based on the analyst’s information
needs. Th e project sponsor, key business users, and other members of the project team can
help the analyst determine who in the organization can best provide important information
about requirements. Th ese people are listed on the interview schedule in the order in which
they should be interviewed.
People at diff erent levels of the organization have varying perspectives on the system, so
it is important to include both managers who manage the processes and staff who actually
perform the processes to gain both high-level and low-level perspectives on an issue. Also,
the kinds of interview subjects needed can change over time. For example, at the start of the
project, the analyst has a limited understanding of the as-is business process. It is common
to begin by interviewing one or two senior managers to get a strategic view and then to move
to midlevel managers who can provide broad, overarching information about the business
process and the expected role of the system being developed. Once the analyst has a good
understanding of the big picture, lower-level managers and staff members can fi ll in the exact
details of how the process works. Like most other things about systems analysis, this is an
iterative process—starting with senior managers, moving to midlevel managers, then staff
members, back to midlevel managers, and so on, depending upon what information is needed
along the way.
It is quite common for the list of interviewees to grow, oft en by 50 to 75 percent. As peo-
ple are interviewed, more information that is needed and additional people who can provide
the information will probably be identifi ed.
7 Some excellent books that address the importance of gathering requirements and various techniques include
Alan M. Davis, Soft ware Requirements: Objects, Functions, & States, Revision (Englewood Cliff s, NJ: Prentice Hall,
1993); Gerald Kotonya and Ian Sommerville, Requirements Engineering (Chichester, England: Wiley, 1998); Dean
Leffi ngwell and Don Widrig, Managing Soft ware Requirements: A Unifi ed Approach (Reading, MA: Addison-Wesley,
2000).
8 A good book on interviewing is that by Brian James, Th e Systems Analysis Interview (Manchester, England: NCC
Blackwell, 1989).
1. Select
Interviewees
96 Chapter 3 Requirements Determination
Requirements-Gathering Techniques 97
Th ere are three types of interview questions: closed-ended questions, open-ended questions,
and probing questions. Closed-ended questions are those that require a specifi c answer. Th ey
are similar to multiple-choice or arithmetic questions on an exam (see Figure 3-4). Closed-
ended questions are used when an analyst is looking for specifi c, precise information (e.g.,
how many credit card requests are received per day). In general, precise questions are best.
For example, rather than asking, Do you handle a lot of requests? it is better to ask, How many
requests do you process per day? Closed-ended questions enable analysts to control the inter-
view and obtain the information they need. However, these types of questions don’t uncover
why the answer is the way it is, nor do they uncover information that the interviewer does not
think to ask for ahead of time.
Open-ended questions are those that leave room for elaboration on the part of the inter-
viewee. Th ey are similar in many ways to essay questions that you might fi nd on an exam (see
Figure 3-4 for examples). Open-ended questions are designed to gather rich information and
give the interviewee more control over the information that is revealed during the interview.
Sometimes the information that the interviewee chooses to discuss uncovers information that is
just as important as the answer (e.g., if the interviewee talks only about other departments when
asked for problems, it may suggest that he or she is reluctant to admit his or her own problems).
Th e third type of question is the probing question. Probing questions follow up on what
has just been discussed in order to learn more, and they oft en are used when the interviewer is
unclear about an interviewee’s answer. Th ey encourage the interviewee to expand on or to con-
fi rm information from a previous response, and they signal that the interviewer is listening and
is interested in the topic under discussion. Many beginning analysts are reluctant to use probing
questions because they are afraid that the interviewee might be off ended at being challenged or
because they believe it shows that they didn’t understand what the interviewee said. When done
politely, probing questions can be a powerful tool in requirements gathering.
In general, an interviewer should not ask questions about information that is readily
available from other sources. For example, rather than asking what information is used to
perform to a task, it is simpler to show the interviewee a form or report (see the section on
document analysis) and ask what information on it is used. Th is helps focus the interviewee
on the task and saves time, because the interviewee does not need to describe the information
detail—he or she just needs to point it out on the form or report.
No type of question is better than another, and a combination of questions is usually used
during an interview. At the initial stage of an IS development project, the as-is process can
2. Design
Interview Questions
FIGURE 3-3
Sample Interview
Schedule
Andria McClellan Director, Accounting Strategic vision for new Mon., March 1
accounting system 8:00–10:00 AM
Jennifer Draper Manager, Accounts Current problems with Mon., March 1
Receivable accounts receivable 2:00–3:15 PM
process; future goals
Mark Goodin Manager, Accounts Current problems with Mon., March 1
Payable accounts payable 4:00–5:15 PM
process; future goals
Anne Asher Supervisor, Data Entry Accounts receivable and Wed., March 3
payable processes 10:00–11:00 AM
Fernando Merce Data Entry Clerk Accounts receivable and Wed., March 3
payable processes 1:00–3:00 PM
Purpose of
Name Position Interview Meeting
be unclear, so the interview process begins with unstructured interviews, interviews that seek
broad and roughly defi ned information. In this case, the interviewer has a general sense of the
information needed but has few closed-ended questions to ask. Th ese are the most challeng-
ing interviews to conduct because they require the interviewer to ask open-ended questions
and probe for important information on the fl y.
As the project progresses, the analyst comes to understand the business process much better
and needs very specifi c information about how business processes are performed (e.g., exactly
how a customer credit card is approved). At this time, the analyst conducts structured interviews,
in which specifi c sets of questions are developed before the interviews. Th ere usually are more
closed-ended questions in a structured interview than in the unstructured approach.
No matter what kind of interview is being conducted, interview questions must be
organized into a logical sequence so that the interview fl ows well. For example, when trying
to gather information about the current business process, it can be useful to move in logical
order through the process or from the most important issues to the least important.
Th ere are two fundamental approaches to organizing the interview questions: top down
or bottom up (see Figure 3-5). With the top-down interview, the interviewer starts with broad,
general issues and gradually works toward more-specifi c ones. With the bottom-up interview,
the interviewer starts with very specifi c questions and moves to broad questions. In practice,
analysts mix the two approaches, starting with broad, general issues, moving to specifi c ques-
tions, and then returning to general issues.
Th e top-down approach is an appropriate strategy for most interviews (it is certainly the
most common approach). Th e top-down approach enables the interviewee to become accus-
tomed to the topic before he or she needs to provide specifi cs. It also enables the interviewer
to understand the issues before moving to the details because the interviewer might not have
suffi cient information at the start of the interview to ask very specifi c questions. Perhaps most
importantly, the top-down approach enables the interviewee to raise a set of big-picture issues
before becoming enmeshed in details, so the interviewer is less likely to miss important issues.
One case in which the bottom-up strategy may be preferred is when the analyst already
has gathered a lot of information about issues and just needs to fi ll in some holes with details.
Bottom-up interviewing may be appropriate if lower-level staff members feel threatened or
unable to answer high-level questions. For example, How can we improve customer service?
might be too broad a question for a customer service clerk, whereas a specifi c question is readily
answerable (e.g., How can we speed up customer returns?). In any event, all interviews should
begin with noncontroversial questions and then gradually move into more contentious issues
aft er the interviewer has developed some rapport with the interviewee.
Closed-ended questions • How many telephone orders are received per day?
• How do customers place orders?
• What information is missing from the monthly sales report?
Open-ended questions • What do you think about the current system?
• What are some of the problems you face on a daily basis?
• What are some of the improvements you would like to see in a
new system?
Probing questions • Why?
• Can you give me an example?
• Can you explain that in a bit more detail?
FIGURE 3-4
Three Types of
Questions
Types of Questions Examples
98 Chapter 3 Requirements Determination
Requirements-Gathering Techniques 99
It is important to prepare for the interview in the same way that you would prepare to give
a presentation. Th e interviewer should have a general interview plan listing the questions to
be asked in the appropriate order, should anticipate possible answers and provide follow-up
with them, and should identify segues between related topics. Th e interviewer should con-
fi rm the areas in which the interviewee has knowledge so as not to ask questions that the
interviewee cannot answer. Review the topic areas, the questions, and the interview plan,
and clearly decide which have the greatest priority in case time runs short.
In general, structured interviews with closed-ended questions take more time to prepare
than unstructured interviews. Some beginning analysts prefer unstructured interviews, think-
ing that they can wing it. Th is is very dangerous and oft en counterproductive, because any
information not gathered in the fi rst interview will require follow-up eff orts, and most users
do not like to be interviewed repeatedly about the same issues.
Th e interviewer should be sure to prepare the interviewee as well. When the interview
is scheduled, the interviewee should be told the reason for the interview and the areas that
will be discussed far enough in advance so that he or she has time to think about the issues
and organize his or her thoughts. Th is is particularly important when the interviewer is
an outsider to the organization and for lower-level employees, who oft en are not asked for
their opinions and who may be uncertain about why they are being interviewed.
Th e fi rst goal is to build rapport with the interviewee, so that he or she trusts the inter-
viewer and is willing to tell the whole truth, not just give the answers that he or she thinks
are wanted. Th e interviewer should appear to be a professional and unbiased, independent
seeker of information. Th e interview should start with an explanation of why the inter-
viewer is there and why he or she has chosen to interview the person; then the interviewer
should move into the planned interview questions.
It is critical to carefully record all the information that the interviewee provides. In our
experience, the best approach is to take careful notes—write down everything the interviewee
says, even if it does not appear immediately relevant. Th e interviewer shouldn’t be afraid to ask
the person to slow down or to pause while writing, because this is a clear indication that the
interviewee’s information is important. One potentially controversial issue is whether or not
to tape-record an interview. Recording ensures that the interviewer does not miss important
3. Prepare for the
Interview
4. Conduct the
Interview
FIGURE 3-5 Top-Down and Bottom-Up Questioning Strategies
High-level:
Very general
Top-Down
Bottom-Up
Medium-level:
Moderately specific
Low-level:
Very specific
How
can order
processing
be improved?
How can we reduce
the number of times that
customers return items
they’ve ordered?
How can we reduce the number of
errors in order processing (e.g., shipping
the wrong products)?
points, but it can be intimidating for the interviewee. Most organizations have policies or
generally accepted practices about the recording of interviews, so they should be determined
before an interview. If the interviewer is worried about missing information and cannot tape
the interview, then he or she can bring along a second person to take detailed notes.
As the interview progresses, it is important to understand the issues that are discussed.
If the interviewer does not understand something, he or she should ask for clarifi cation.
Th e interviewer should not be afraid to ask dumb questions, because the only thing worse
than appearing dumb is to be dumb by not understanding something. If the interviewer
doesn’t understand something during the interview, he or she certainly won’t understand it
aft erwards. Jargon should be recognized and defi ned; any jargon not understood should be
clarifi ed. One good strategy to increase understanding during an interview is to periodically
summarize the key points that the interviewee is communicating. Th is avoids misunder-
standings and also demonstrates that the interviewer is listening.
Finally, facts should be separated from opinion. Th e interviewee may say, for example,
We process too many credit card requests. Th is is an opinion, and it is useful to follow this
up with a probing question requesting support for the statement (e.g., Oh, how many do you
process in a day?). It is helpful to check the facts because any diff erences between the facts
and the interviewee’s opinions can point out key areas for improvement. Suppose the inter-
viewee complains about a high or increasing number of errors, but the logs show that errors
have been decreasing. Th is suggests that errors are viewed as a very important problem that
should be addressed by the new system, even if they are declining.
As the interview draws to a close, the interviewee should have time to ask questions or
provide information that he or she thinks is important but was not part of the interview plan.
In most cases, the interviewee has no additional concerns or information, but in some cases
this leads to unanticipated, but important, information. Likewise, it can be useful to ask the
interviewee if there are other people who should be interviewed. Th e interview should end on
time (if necessary, some topics can be omitted or another interview can be scheduled).
As a last step in the interview, the interviewer should briefl y explain what will happen.
Th e interviewer shouldn’t prematurely promise certain features in the new system or a spe-
cifi c delivery date, but he or she should reassure the interviewee that his or her time was well
spent and very helpful to the project.
Aft er the interview is over, the analyst needs to prepare an interview report that describes the
information from the interview (Figure 3-6). Th e report contains interview notes, information
that was collected over the course of the interview and is summarized in a useful format. In
general, the interview report should be written within forty-eight hours of the interview,
because the longer the interviewer waits, the more likely he or she is to forget information.
Oft en, the interview report is sent to the interviewee with a request to read it and inform
the analyst of clarifi cations or updates. Th e interviewee needs to be convinced that the inter-
viewer genuinely wants his or her corrections to the report. Usually there are few changes, but
the need for any signifi cant changes suggests that a second interview will be required. Never
distribute someone’s information without prior approval.
Joint Application Development (JAD)
JAD is an information-gathering technique that allows the project team, users, and management
to work together to identify requirements for the system. IBM developed the JAD technique in
the late 1970s, and it is oft en the most useful method for collecting information from users.9
9 More information on JAD can be found in J. Wood and D. Silver, Joint Application Development (New York:
Wiley, 1989); Alan Cline, “Joint Application Development for Requirements Collection and Management,” http://
www.carolla.com/wp-jad.htm.
5. Post-Interview
Follow-up
100 Chapter 3 Requirements Determination
Requirements-Gathering Techniques 101
Capers Jones claims that JAD can reduce scope creep by 50 percent and prevent the system’s
requirements from being too specifi c or too vague, both of which cause trouble during later stages
of the development process.10
JAD is a structured process in which ten to twenty users meet together under the direc-
tion of a facilitator skilled in JAD techniques. Th e facilitator sets the meeting agenda and
guides the discussion but does not join in the discussion as a participant. He or she does not
provide ideas or opinions on the topics under discussion so as to remain neutral during the
session. Th e facilitator must be an expert in both group-process techniques and systems-
analysis and design techniques. One or two scribes assist the facilitator by recording notes,
making copies, and so on. Oft en the scribes use computers and CASE tools to record infor-
mation as the JAD session proceedings.
Th e JAD group meets for several hours, several days, or several weeks until all the issues
have been discussed and the needed information is collected. Most JAD sessions take place
in a specially prepared meeting room, away from the participants’ offi ces so that they are not
interrupted. Th e meeting room is usually arranged in a U-shape so that all participants can
easily see each other. At the front of the room (the open part of the U), are a whiteboard, fl ip
chart, and/or overhead projector for use by the facilitator leading the discussion.
FIGURE 3-6 Interview Report
Interview Notes Approved by: Linda Estey
Person Interviewed: Linda Estey,
Director, Human Resources
Interviewer: Barbara Wixom
Purpose of Interview:
• Understand reports produced for Human Resources by the current system
• Determine information requirements for future system
Summary of Interview:
• Sample reports of all current HR reports are attached to this report. The information that is not
used and missing information are noted on the reports.
• Two biggest problems with the current system are:
1. The data are too old (the HR Department needs information within two days of month end;
currently, information is provided to them after a three-week delay)
2. The data are of poor quality (often reports must be reconciled with departmental HR
database)
• The most common data errors found in the current system include incorrect job level information and
missing salary information.
Open Items:
• Get current employee roster report from Mary Skudrna (extension 4355).
• Verify calculations used to determine vacation time with Mary Skudrna.
• Schedule interview with Jim Wack (extension 2337) regarding the reasons for data quality
problems.
Detailed Notes: See attached transcript.
10 See Kevin Strehlo, “Catching up with the Jones and ‘Requirement’ Creep,” Infoworld (July 29, 1996); Kevin Strehlo,
“Th e Makings of a Happy Customer: Specifying Project X,” Infoworld (November 11, 1996).
JAD suff ers from the traditional problems associated with groups: Sometimes people
are reluctant to challenge the opinions of others (particularly their boss), a few people oft en
dominate the discussion, and not everyone participates. In a fi ft een-member group, for exam-
ple, if everyone participates equally, then each person can talk for only four minutes each
hour and must listen for the remaining fi ft y-six minutes—not a very effi cient way to collect
information.
A new form of JAD called electronic JAD, or e-JAD, attempts to overcome these prob-
lems by using groupware. In an e-JAD meeting room, each participant uses special soft ware
on a networked computer to send anonymous ideas and opinions to everyone else. In this
way, all participants can contribute at the same time without fear of reprisal from people
with diff ering opinions. Initial research suggests that e-JAD can reduce the time required
to run JAD sessions by 50 to 80 percent.11 A good JAD approach follows a set of fi ve steps.
JAD participants are selected in the same way as are interview participants, based on the
information they can contribute in order to provide a broad mix of organizational levels and
to build political support for the new system. Th e need for all JAD participants to be away
from their offi ce at the same time can be a major problem. Th e offi ce might need to be closed
or operate with a skeleton staff until the JAD sessions are complete.
11 For more information on e-JAD, see A. R. Dennis, G. S. Hayes, and R. M. Daniels, “Business Process Modeling
with Groupware,” Journal of Management Information Systems 15, no. 4 (1999): 115–142.
1. Select Participants
102 Chapter 3 Requirements Determination
Interpersonal skills are skills that enable you to develop
rapport with others, and they are very important for
interviewing. They help you to communicate with others
effectively. Some people develop good interpersonal
skills at an early age; they simply seem to know how to
communicate and interact with others. Other people are
less lucky and need to work hard to develop their skills.
Interpersonal skills, like most skills, can be learned.
Here are some tips:
• Don’t worry, be happy. Happy people radiate con-
fi dence and project their feelings on others. Try inter-
viewing someone while smiling and then interviewing
someone else while frowning and see what happens.
• Pay attention. Pay attention to what the other person
is saying (which is harder than you might think). See
how many times you catch yourself with your mind
on something other than the conversation at hand.
• Summarize key points. At the end of each major
theme or idea that someone explains, repeat the key
points back to the speaker (e.g., Let me make sure I
understand. The key issues are. . . .”). This demon-
strates that you consider the information important,
and it also forces you to pay attention (you can’t
repeat what you didn’t hear).
• Be succinct. When you speak, be succinct. The goal
in interviewing (and in much of life) is to learn, not to
impress. The more you speak, the less time you give
to others.
• Be honest. Answer all questions truthfully, and if you
don’t know the answer, say so.
• Watch body language (yours and theirs). The way a
person sits or stands conveys much information. In
general, a person who is interested in what you are
saying sits or leans forward, makes eye contact, and
often touches his or her face. A person leaning away
from you or with an arm over the back of a chair is
uninterested. Crossed arms indicate defensiveness or
uncertainty, and steepling (sitting with hands raised
in front of the body with fi ngertips touching) indi-
cates a feeling of superiority.
3-1 Developing Interpersonal SkillsPRACTICAL
TIP
Requirements-Gathering Techniques 103
Ideally, the participants who are released from regular duties to attend the JAD sessions
should be the very best people in that business unit. However, without strong management
support, JAD sessions can fail because those selected to attend the JAD session are people who
are less likely to be missed (i.e., the least competent people).
Th e facilitator should be someone who is an expert in JAD or e-JAD techniques and,
ideally, someone who has experience with the business under discussion. In many cases, the
JAD facilitator is a consultant external to the organization because the organization might not
have a recurring need for JAD or e-JAD expertise. Developing and maintaining this expertise
in-house can be expensive.
JAD sessions can run from as little as half a day to several weeks, depending upon the size and
scope of the project. In our experience, most JAD sessions tend to last fi ve to ten days, spread
over a three-week period. Most e-JAD sessions tend to last one to four days in a one-week
period. JAD and e-JAD sessions usually go beyond collecting information and move into anal-
ysis. For example, the users and the analysts collectively can create analysis deliverables, such as
the functional models or the requirements defi nition.
JAD sessions usually are designed and structured using the same principles as inter-
views. Most JAD sessions are designed to collect specifi c information from users, and this
requires developing a set of questions before the meeting. One diff erence between JAD
and interviewing is that all JAD sessions are structured—they must be carefully planned.
In general, closed-ended questions are seldom used because they do not spark the open
and frank discussion that is typical of JAD. In our experience, it is better to proceed top
down in JAD sessions when gathering information. Typically thirty minutes is allocated to
each separate agenda item, and frequent breaks are scheduled throughout the day because
participants tire easily.
As with interviewing, it is important to prepare the analysts and participants for a JAD
session. Because the sessions can go beyond the depth of a typical interview and are usually
conducted off -site, participants may be more concerned about how to prepare. It is impor-
tant that the participants understand what is expected of them. If the goal of the JAD session,
for example, is to develop an understanding of the current system, then participants can
bring procedure manuals and documents with them. If the goal is to identify improvements
for a system, then before they come to the JAD session they can think about how they would
improve the system.
Most JAD sessions follow a formal agenda, and most have formal ground rules that defi ne appro-
priate behavior. Common ground rules include following the schedule, respecting others’ opin-
ions, accepting disagreement, and ensuring that only one person talks at a time.
Th e role of a JAD facilitator can be challenging. Many participants come to a JAD session
with strong feelings about the system to be discussed. Channeling these feelings so that the ses-
sion moves forward in a positive direction and getting participants to recognize and accept—but
not necessarily agree on—opinions and situations diff erent from their own requires signifi cant
expertise in systems analysis and design, JAD, and interpersonal skills. Few systems analysts
attempt to facilitate JAD sessions without being trained in JAD techniques, and most apprentice
with a skilled JAD facilitator before they attempt to lead their fi rst session.
Th e JAD facilitator performs three key functions. First, he or she ensures that the group
sticks to the agenda. Th e only reason to digress from the agenda is when it becomes clear to
the facilitator, project leader, and project sponsor that the JAD session has produced some
new information that is unexpected and requires the JAD session (and perhaps the project)
to move in a new direction. When participants attempt to divert the discussion away from the
4. Conducting
a JAD Session
2. Design a JAD
Session
3. Preparing for a
JAD Session
agenda, the facilitator must be fi rm but polite in leading discussion back to the agenda and
getting the group back on track.
Second, the facilitator must help the group understand the technical terms and jargon
that surround the system-development process and help the participants understand the
specifi c analysis techniques used. Participants are experts in their area, or their part of
the business, but they are not experts in systems analysis. Th e facilitator must, therefore,
minimize the learning required and teach participants how to eff ectively provide the right
information.
Th ird, the facilitator records the group’s input on a public display area, which can be a
whiteboard, fl ip chart, or computer display. He or she structures the information that the
group provides and helps the group recognize key issues and important solutions. Th e facil-
itator must remain neutral at all times and simply help the group through the process. Th e
moment the facilitator off ers an opinion on an issue, the group will see him or her not as a
neutral party but rather as someone who could be attempting to sway the group into some
predetermined solution.
However, this does not mean that the facilitator should not try to help the group resolve
issues. For example, if two items appear to be the same to the facilitator, the facilitator should
not say, “I think these may be similar.” Instead, the facilitator should ask, “Are these similar?”
If the group decides they are, the facilitator can combine them and move on. However, if
the group decides they are not similar (despite what the facilitator believes), the facilitator
should accept the decision and move on. Th e group is always right, and the facilitator has
no opinion.
As with interviews, a JAD post-session report is prepared and circulated among session
attendees. Th e post-session report is essentially the same as the interview report in Figure 3-6.
Because the JAD sessions are longer and provide more information, it usually takes a week or
two aft er the JAD session before the report is complete.
Questionnaires
A questionnaire is a set of written questions used to obtain information from individ-
uals. Questionnaires are oft en used when there is a large number of people from whom
information and opinions are needed. In our experience, questionnaires are a common
technique with systems intended for use outside the organization (e.g., by customers or
vendors) or for systems with business users spread across many geographic locations.
Most people automatically think of paper when they think of questionnaires, but today
more questionnaires are being distributed in electronic form, either via e-mail or on the
Web. Electronic distribution can save a signifi cant amount of money as compared to dis-
tributing paper questionnaires. A good process to use when using questionnaires follows
four steps.
As with interviews and JAD sessions, the fi rst step is to identify the individuals to whom the
questionnaire will be sent. However, it is not usual to select every person who could provide
useful information. Th e standard approach is to select a sample, or subset, of people who
are representative of an entire group. Sampling guidelines are discussed in most statistics
books, and most business schools include courses that cover the topic, so we do not discuss it
here. Th e important point in selecting a sample, however, is to realize that not everyone who
receives a questionnaire will actually complete it. On average, only 30 to 50 percent of paper
and e-mail questionnaires are returned. Response rates for Web-based questionnaires tend to
be signifi cantly lower (oft en only 5 to 30 percent).
104 Chapter 3 Requirements Determination
1. Select Participants
5. Post-JAD Follow-up
Requirements-Gathering Techniques 105
Managing Problems in JAD Sessions
I have run more than a hundred JAD sessions and have
learned several standard “facilitator tricks.” Here are some
common problems and some ways to deal with them.
• Domination. The facilitator should ensure that no one
person dominates the group discussion. The only way
to deal with someone who dominates is head on. Dur-
ing a break, approach the person, thank him or her for
his or her insightful comments, and ask the person to
help you make sure that others also participate.
• Noncontributors. Drawing out people who have par-
ticipated very little is challenging because you want
to bring them into the conversation so that they will
contribute again. The best approach is to ask a direct
factual question that you are certain they can answer.
And it helps to ask the question in a long way to give
them time to think. For example, “Pat, I know you’ve
worked shipping orders a long time. You’ve probably
been in the shipping department longer than anyone
else. Could you help us understand exactly what hap-
pens when an order is received in shipping?”
• Side discussions. Sometimes participants engage in
side conversations and fail to pay attention to the
group. The easiest solution is simply to walk close
to the people and continue to facilitate right in front
of them. Few people will continue a side conversion
when you are two feet from them and the entire
group’s attention is on you and them.
• Agenda merry-go-round. The merry-go-round occurs
when a group member keeps returning to the same
issue every few minutes and won’t let go. One solu-
tion is to let the person have fi ve minutes to ramble
on about the issue while you carefully write down
every point on a fl ip chart or computer fi le. This fl ip
chart or fi le is then posted conspicuously on the
wall. When the person brings up the issue again, you
interrupt them, walk to the paper and ask them what
to add. If they mention something already on the list,
you quickly interrupt, point out that it is there, and
ask what other information to add. Don’t let them
repeat the same point, but write any new information.
• Violent agreement. Some of the worst disagreements
occur when participants really agree on the issues
but don’t realize that they agree because they are
using different terms. An example is arguing whether
a glass is half empty or half full; they agree on the
facts but can’t agree on the words. In this case, the
facilitator has to translate the terms into different
words and fi nd common ground so the parties rec-
ognize that they really agree.
• Unresolved confl ict. In some cases, participants
don’t agree and can’t understand how to determine
what alternatives are better. You can help by structur-
ing the issue. Ask for criteria by which the group will
identify a good alternative (e.g., “Suppose this idea
really did improve customer service. How would I
recognize the improved customer service?”). Then
once you have a list of criteria, ask the group to
assess the alternatives using them.
• True confl ict. Sometimes, despite every attempt, par-
ticipants just can’t agree on an issue. The solution is
to postpone the discussion and move on. Document
the issue as an open issue and list it prominently on
a fl ip chart. Have the group return to the issue hours
later. Often the issue will have resolved itself by then
and you haven’t wasted time on it. If the issue cannot
be resolved later, move it to the list of issues to be
decided by the project sponsor or some other more
senior member of management.
• Humor. Humor is one of the most powerful tools a
facilitator has and thus must be used judiciously. The
best JAD humor is always in context; never tell jokes
but take the opportunity to fi nd the humor in the
situation.
Alan Dennis
PRACTICAL
TIP
Because the information on a questionnaire cannot be immediately clarifi ed for a confused
respondent, developing good questions is critical for questionnaires. Questions on question-
naires must be very clearly written and leave little room for misunderstanding, so closed-ended
questions tend to be most commonly used. Questions must clearly enable the analyst to sep-
arate facts from opinions. Opinion questions oft en ask respondents the extent to which they
agree or disagree (e.g., Are network problems common?), whereas factual questions seek more
2. Designing a
Questionnaire
precise values (e.g., How oft en does a network problem occur: once an hour, once a day, once
a week?). See Figure 3-7 for guidelines on questionnaire design.
Perhaps the most obvious issue—but one that is sometimes overlooked—is to have a
clear understanding of how the information collected from the questionnaire will be analyzed
and used. Th is issue must be addressed before the questionnaire is distributed, because it is
too late aft erward.
Questions should be relatively consistent in style, so that the respondent does not have to
read instructions for each question before answering it. It is generally good practice to group
related questions together to make them simpler to answer. Some experts suggest that ques-
tionnaires should start with questions important to respondents, so that the questionnaire
immediately grabs their interest and induces them to answer it. Perhaps the most important
step is to have several colleagues review the questionnaire and then pretest it with a few people
drawn from the groups to whom it will be sent. It is surprising how oft en seemingly simple
questions can be misunderstood.
Th e key issue in administering the questionnaire is getting participants to complete the
questionnaire and send it back. Dozens of marketing research books have been written about
ways to improve response rates. Commonly used techniques include clearly explaining why
the questionnaire is being conducted and why the respondent has been selected, stating a date
by which the questionnaire is to be returned, off ering an inducement to complete the ques-
tionnaire (e.g., a free pen), and off ering to supply a summary of the questionnaire responses.
Systems analysts have additional techniques to improve response rates inside the organiza-
tion, such as personally handing out the questionnaire and personally contacting those who
have not returned them aft er a week or two, as well as requesting the respondents’ supervisors
to administer the questionnaires in a group meeting.
It is helpful to process the returned questionnaires and develop a questionnaire report soon aft er
the questionnaire deadline. Th is ensures that the analysis process proceeds in a timely fashion and
that respondents who requested copies of the results receive them promptly.
Document Analysis
Project teams oft en use document analysis to understand the as-is system. Under ideal cir-
cumstances, the project team that developed the existing system will have produced docu-
mentation that was then updated by all subsequent projects. In this case, the project team can
start by reviewing the documentation and examining the system itself.
Unfortunately, many systems are not well documented because project teams fail to
document their projects along the way, and when the projects are over, there is no time to
go back and document. Th erefore, there might not be much technical documentation about
the current systems available, or it might not contain updated information about recent sys-
tem changes. However, many helpful documents do exist in an organization: paper reports,
3. Administering
the Questionnaire
4. Questionnaire
Follow-up
106 Chapter 3 Requirements Determination
• Begin with nonthreatening and interesting questions.
• Group items into logically coherent sections.
• Do not put important items at the very end of the questionnaire.
• Do not crowd a page with too many items.
• Avoid abbreviations.
• Avoid biased or suggestive items or terms.
• Number questions to avoid confusion.
• Pretest the questionnaire to identify confusing questions.
• Provide anonymity to respondents.
FIGURE 3-7
Good Questionnaire
Design
memorandums, policy manuals, user-training manuals, organization charts, forms, and, of
course, the user interface with the existing system.
But these documents tell only part of the story. Th ey represent the formal system that the
organization uses. Quite oft en, the real, or informal, system diff ers from the formal one, and these
diff erences, particularly large ones, give strong indications of what needs to be changed. For
example, forms or reports that are never used should probably be eliminated. Likewise, boxes or
questions on forms that are never fi lled in (or are used for other purposes) should be rethought.
See Figure 3-8 for an example of how a document can be interpreted.
Th e most powerful indication that the system needs to be changed is when users
create their own forms or add additional information to existing ones. Such changes clearly
demonstrate the need for improvements to existing systems. Th us, it is useful to review both
blank and completed forms to identify these deviations. Likewise, when users access multiple
reports to satisfy their information needs, it is a clear sign that new information or new infor-
mation formats are needed.
Name: Buffy Pat Smith
Pet’s Name: Buffy Collie 7/6/99
Address: 100 Central Court. Apartment 10
Toronto, Ontario K7L 3N6
Phone Number: 555-3400
416-
Do you have insurance: yes
Insurance Company: Pet’s Mutual
Policy Number: KA-5493243
CENTRAL VETERINARY CLINIC
Patient Information Card
The staff had to add additional
information about the type of animal
and the animal’s date of birth. This
information should be added to the
new form in the to-be system.
The customer made a mistake.
This should be labeled
Owner’s Name to prevent
confusion.
The customer did not include
area code in the phone
number. This should be made
more clear.
FIGURE 3-8
Performing a
Document Analysis
Requirements-Gathering Techniques 107
Observation
Observation, the act of watching processes being performed, is a powerful tool for gathering
information about the as-is system because it enables the analyst to see the reality of a situa-
tion, rather than listening to others describe it in interviews or JAD sessions. Several research
studies have shown that many managers really do not remember how they work and how
they allocate their time. (Quick, how many hours did you spend last week on each of your
courses?) Observation is a good way to check the validity of information gathered from indi-
rect sources such as interviews and questionnaires.
In many ways, the analyst becomes an anthropologist as he or she walks through the
organization and observes the business system as it functions. Th e goal is to keep a low pro-
fi le, to not interrupt those working, and to not infl uence those being observed. Nonetheless,
it is important to understand that what analysts observe may not be the normal day-to-day
routine because people tend to be extremely careful in their behavior when they are being
watched. Even though normal practice may be to break formal organizational rules, the
observer is unlikely to see this. (Remember how you drove the last time a police car followed
you?) Th us, what you see might not be what you get.
Observation is oft en used to supplement interview information. Th e location of a person’s
offi ce and its furnishings give clues to the person’s power and infl uence in the organization and
can be used to support or refute information given in an interview. For example, an analyst
might become skeptical of someone who claims to use the existing computer system exten-
sively if the computer is never turned on while the analyst visits. In most cases, observation
supports the information that users provide in interviews. When it does not, it is an important
signal that extra care must be taken in analyzing the business system.
Selecting the Appropriate Techniques
Each of the requirements-gathering techniques discussed earlier has strengths and weak-
nesses. No one technique is always better than the others, and in practice most projects use a
combination of techniques. Th us, it is important to understand the strengths and weaknesses
of each technique and when to use each (see Figure 3-9). One issue not discussed is that of the
analysts’ experience. In general, document analysis and observation require the least amount
of training, whereas JAD sessions are the most challenging.
Type of Information Th e fi rst characteristic is the type of information. Some techniques are
more suited for use at diff erent stages of the analysis process, whether understanding the as-is
system, identifying improvements, or developing the to-be system. Interviews and JAD are
commonly used in all three stages. In contrast, document analysis and observation usually are
most helpful for understanding the as-is, although occasionally they provide information about
FIGURE 3-9 Table of Requirements-Gathering Techniques
Type of information As-is, improvements, As-is, improvements, As-is, improvements As-is As-is
to-be to-be
Depth of information High High Medium Low Low
Breadth of information Low Medium High High Low
Integration of information Low High Low Low Low
User involvement Medium High Low Low Low
Cost Medium Low to Medium Low Low Low to Medium
Joint Application Document
Interviews Design Questionnaires Analysis Observation
108 Chapter 3 Requirements Determination
current problems that need to be improved. Questionnaires are oft en used to gather informa-
tion about the as-is system as well as general information about improvements.
Depth of Information Th e depth of information refers to how rich and detailed the infor-
mation is that the technique usually produces and the extent to which the technique is useful
for obtaining not only facts and opinions but also an understanding of why those facts and
opinions exist. Interviews and JAD sessions are very useful for providing a good depth of rich
and detailed information and helping the analyst to understand the reasons behind them. At
the other extreme, document analysis and observation are useful for obtaining facts, but little
beyond that. Questionnaires can provide a medium depth of information, soliciting both facts
and opinions with little understanding of why they exist.
Breadth of Information Breadth of information refers to the range of information and infor-
mation sources that can be easily collected using the chosen technique. Questionnaires and
document analysis are both easily capable of soliciting a wide range of information from a large
number of information sources. In contrast, interviews and observation require the analyst to
visit each information source individually and, therefore, take more time. JAD sessions are in
the middle because many information sources are brought together at the same time.
Integration of Information One of the most challenging aspects of requirements gather-
ing is integrating the information from diff erent sources. Simply put, diff erent people can
provide confl icting information. Combining this information and attempting to resolve
diff erences in opinions or facts is usually very time consuming because it means contacting
each information source in turn, explaining the discrepancy, and attempting to refi ne the
information. In many cases, the individual wrongly perceives that the analyst is challenging
his or her information, when in fact it is another user in the organization who is doing so.
Th is can make the user defensive and make it hard to resolve the diff erences.
All techniques suff er integration problems to some degree, but JAD sessions are designed
to improve integration because all information is integrated when it is collected, not aft er-
ward. If two users provide confl icting information, the confl ict becomes immediately obvi-
ous, as does the source of the confl ict. Th e immediate integration of information is the single
most important benefi t of JAD that distinguishes it from other techniques, and this is why
most organizations use JAD for important projects.
User Involvement User involvement refers to the amount of time and energy the intended
users of the new system must devote to the analysis process. It is generally agreed that as users
become more involved in the analysis process, the chance of success increases. However, user
involvement can have a signifi cant cost, and not all users are willing to contribute valuable
time and energy. Questionnaires, document analysis, and observation place the least burden
on users, whereas JAD sessions require the greatest eff ort.
Cost Cost is always an important consideration. In general, questionnaires, document
analysis, and observation are low-cost techniques (although observation can be quite time
consuming). Th e low cost does not imply that they are more or less eff ective than the other
techniques. Interviews and JAD sessions generally have moderate costs. In general, JAD ses-
sions are much more expensive initially, because they require many users to be absent from
their offi ces for signifi cant periods of time, and they oft en involve highly paid consultants.
However, JAD sessions signifi cantly reduce the time spent in information integration and
thus can cost less in the long term.
Combining Techniques In practice, requirements gathering combines a series of diff erent tech-
niques. Most analysts start by using interviews with senior manager(s) to gain an understanding
of the project and the big-picture issues. From these interviews, it becomes clear whether large
or small changes are anticipated. Th ese interviews are oft en followed with analysis of documents
Requirements-Gathering Techniques 109
12 See B. Henderson-Sellers, A. Simons, and H. Younessi, Th e OPEN Toolbox of Techniques (Harlow, England:
Addison-Wesley, 1998).
13 For more information on concept mapping, see J. D. Novak and D. B. Gowin, Learning How to Learn (Cambridge,
UK: Cambridge University Press, 1984); J. D. Novak, Learning, Creating, and Using Knowledge: Concept MapsTM as
Facilitative Tools in Schools and Corporations (Mahwah, NJ: Lawrence Erlbaum Associates, Publishers, 1998). Also, a
free concept mapping tool is available from the Institute of Human and Machine Cognition at cmap.ihmc.us.
110 Chapter 3 Requirements Determination
and policies to gain some understanding of the as-is system. Usually interviews come next to
gather the rest of the information needed for the as-is picture.
In our experience, identifying improvements is most commonly done using JAD sessions
because the JAD session enables the users and key stakeholders to work together through an
analysis technique and come to a shared understanding of the possibilities for the to-be sys-
tem. Occasionally, these JAD sessions are followed by questionnaires sent to a much wider set
of users or potential users to see whether the opinions of those who participated in the JAD
sessions are widely shared.
Developing the concept for the to-be system is oft en done through interviews with senior
managers, followed by JAD sessions with users of all levels to make sure that the key needs of
the new system are well understood.
ALTERNATIVE REQUIREMENTS DOCUMENTATION TECHNIQUES
Some other very useful requirements-gathering and documentation techniques include
throwaway prototyping, use cases, role-playing CRC cards with use-case-based scenarios,
concept mapping, and recording user stories on story cards and task lists. Th rowaway pro-
totyping was described in Chapter 1. In essence, throwaway prototypes are created to better
understand some aspect of the new system. In many cases, they are used to test out some
technical aspect of a nonfunctional requirement, such as connecting a client workstation to a
server. If you have never done this before, it will be a lot easier to develop a very small example
system to test out the necessary design of the connection from the client workstation to the
server instead of trying to do it the fi rst time with the full-blown system. Th rowaway proto-
typing is very useful in designing user interfaces (see Chapter 10).
Use cases, as described in Chapter 1, are the fundamental approach that the Unifi ed Process
and Unifi ed Modeling Language (UML) use to document and gather functional requirements.
We describe them in Chapter 4. Role-playing CRC cards with use-case-based scenarios are
very useful when creating functional (see Chapter 4), structural (see Chapter 5), and behavioral
(see Chapter 6) models. We describe this approach in Chapter 5. Th e remainder of this section
describes the use of concept mapping recording user stories on story cards and task lists.
Concept Maps
Concept maps represent meaningful relationships between concepts. Th ey are useful for
focusing individuals on the small number of key ideas on which they should concentrate.
A concept map is essentially a node-and-arc representation, where the nodes represent the
individual requirements and the arcs represent the relationships among the requirements.
Each arc is labeled with a relationship name. Concept maps also have been recommended as
a possible technique to support modeling requirements for object-oriented systems develop-
ment and knowledge-management systems.12 Concept mapping is an educational psychology
technique that has been used in schools, corporations, and health care agencies to facilitate
learning, understanding, and knowledge creation.13 Th e advantage of the concept-mapping
approach to representing requirements over the typical textual approach (see Figure 3-1) is
that a concept map is not limited to a hierarchical representation. Concept maps allow the rela-
tionships among the functional and nonfunctional requirements to be explicitly represented.
Figure 3-10 shows a concept map that portrays the information contained in the requirements
R
eq
ui
re
m
en
ts
C
ul
tu
ra
l a
nd
Po
lit
ic
al
R
eq
ui
re
m
en
ts
N
on
fu
nc
ti
on
al
R
eq
ui
re
m
en
ts
Se
cu
ri
ty
R
eq
ui
re
m
en
ts
O
pe
ra
ti
on
al
R
eq
ui
re
m
en
ts
Pe
rf
or
m
an
ce
R
eq
ui
re
m
en
ts
B
ac
ku
p
Sc
he
du
le
D
ai
ly
Su
pp
or
t W
ir
el
es
s
Pr
in
ti
ng
St
or
e
N
ew
A
pp
oi
nt
m
en
t
in
2
S
ec
on
ds
o
r
Le
ss
R
et
ri
ev
e
D
ai
ly
A
pp
oi
nt
m
en
t
Sc
he
du
le
in
2
Se
co
nd
s
or
L
es
s
O
pe
ra
te
in
W
in
do
w
s
En
vi
ro
nm
en
t
M
ak
e
A
pp
oi
nt
m
en
t
C
an
ce
l A
pp
oi
nt
m
en
t Pr
in
t
Sc
he
du
le
U
pd
at
e
Sc
he
du
le
C
he
ck
S
ch
ed
ul
e
C
ha
ng
e
A
pp
oi
nt
m
en
t
M
an
ag
e
A
pp
oi
nt
m
en
ts
Pr
od
uc
e
Sc
he
du
le
Fu
nc
ti
on
al
R
eq
ui
re
m
en
ts
R
ec
or
d
D
oc
to
r
A
va
ila
bi
lit
y
O
nl
y
M
an
ag
er
s
C
an
Pr
od
uc
e
Sc
he
du
le
O
nl
y
D
oc
to
rs
S
et
A
va
ila
bi
lit
y
im
pa
ct
s
im
pa
ct
s
im
pa
ct
s
im
pa
ct
s
im
pa
ct
s
im
pa
ct
s
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
in
cl
ud
e
FI
G
U
R
E
3
-1
0
Sa
m
p
le
R
eq
u
ir
em
en
ts
C
o
n
ce
p
t
M
ap
111
defi nition shown in Figure 3-1. By using a concept map to represent the requirements instead
of the textual approach, the relationship between the functional and nonfunctional require-
ments can be made explicit. For example, the two security requirements Only Doctors Set
Availability and Only Managers Can Produce Schedule are explicitly linked to the Record
Doctor Availability and Produce Schedule functional requirements, respectively. Th is is very
diffi cult to represent in a text-only version of the requirements defi nition. Also, by having
the user and analyst focus on the graphical layout of the map, additional requirements can
be discovered. One obvious issue with this approach is that if the number of requirements
becomes many and the relationships between them become complex, then the number of
nodes and arcs will become so intertwined that the advantage of being able to explicitly see the
relationships will be lost. However, by combining both text and concept-map representations,
it is possible to leverage the strength of both textual and graphical representations to more
completely represent the requirements.
User Stories
User stories, along with their associated story cards and task lists, are associated with the
agile development approaches. User stories have been shown to be very useful in gathering
requirements in a nonthreatening manner that respects the user’s point of view. Th ey are
typically captured using story cards (index cards) and are recorded on a task list (or from a
Scrum perspective, on the product backlog). Both story cards and task lists are considered
to be lightweight approaches to documenting and gathering requirements.14 Stories capture
both functional and nonfunctional requirements. For example, with regard to the doctor’s
offi ce appointment example, a functional requirement-based story could be:
As a secretary, I want to be able to schedule appointments for our patients so that we can
meet our patients’ needs.
While an operational nonfunctional requirement-based story could be:
As a secretary, I want to be able to print the daily schedule using wireless technology so
that all printing can be performed using a shared printer without having to deal with
printer cables connecting all of the computers to the printer.
Once the story is written down, it is discussed to determine the amount of eff ort it will take
to implement it. During the discussion, a task list is created for the story. If the story is
deemed to be too large—e.g., there are too many tasks on the task list—the story is split up
into multiple stories each being recorded on its own story card and the tasks are allocated
across the new stories. In many shops, once a set of tasks has been identifi ed with a story,
the story and its tasks are taped on a wall together so that all members of the development
team can see the requirements. Th e story can be prioritized by importance by placing a rat-
ing on the card. Th e story can also be evaluated for the level of risk associated with it. Th e
importance level and amount of risk associated with the story can be used to help choose
which requirements to implement fi rst. Th e advantage of using story cards and task lists
to document requirements is that they are very low tech, high touch, easily updatable, and
very portable.
112 Chapter 3 Requirements Determination
14 For more information on story cards and task lists see M. Cohn, User Stories Applied: For Agile Soft ware
Development (Boston, MA: Addison-Wesley, 2004); B. Rinzler, Telling Stories: A Short Path to Writing Better
Soft ware Requirements (Indianapolis, IN: Wiley, 2009); M. Lippert, S. Roock, H. Wolf, eXtreme Programming
in Action: Practical Experiences from Real World Projects (Chichester, England: Wiley & Sons, Ltd., 2002);
C. Larman, Agile & Iterative Development: A Manager’s Guide (Boston, MA: Addison-Wesley, 2004).
1. Table of Contents
2. Executive Summary
A summary of all the essential information in the proposal so that a busy executive can read it
quickly and decide what parts of the proposal to read in more depth.
3. System Request
The revised system request form (see Chapter 2).
4. Workplan
The original workplan, revised after having completed analysis (see Chapter 2).
5. Feasibility Analysis
A revised feasibility analysis, using the information from analysis (see Chapter 2).
6. Requirements Defi nition
A list of the functional and nonfunctional business requirements for the system (this chapter).
7. Functional Model
An activity diagram, a set of use-case descriptions, and a use-case diagram that illustrate the basic
processes or external functionality that the system needs to support (see Chapter 4).
8. Structural Models
A set of CRC cards, class diagram, and object diagrams that describe the structural aspects of the
to-be system (see Chapter 5). This may also include structural models of the current as-is system that
will be replaced.
9. Behavioral Models
A set of sequence diagrams, communication diagrams, behavioral-state machines, and a CRUDE
matrix that describe the internal behavior of the to-be system (see Chapter 6). This may include
behavioral models of the as-is system that will be replaced.
10. Appendices
These contain additional material relevant to the proposal, often used to support the recommended
system. This might include results of a questionnaire survey or interviews, industry reports and
statistics, and so on.
FIGURE 3-11
System Proposal
Template
15 Depending on the client, much more detailed specifi cations may be required; for example the Department
of Defense, NASA, IEEE/ANSI, and the Naval Research Laboratory all have very specifi c formats that must be
followed. For more information on these more detailed specifi cations, see A. M Davis, Soft ware Requirements,
Revision (Upper Saddle River, NJ: Prentice Hall, 1993); G. Kotonya and I. Sommerville, Requirements Engineering
(Chichester, England: Wiley, 1998); R. H. Th ayer and M. Dorfman (eds.), Soft ware Requirements Engineering, 2nd Ed.
(Los Alamitos, CA: IEEE Computer Society Press, 1997).
THE SYSTEM PROPOSAL
A system proposal brings together into a single comprehensive document the material created
during planning and analysis. Th e system proposal typically includes an executive summary,
the system request, the workplan, the feasibility analysis, the requirements defi nition, and the
evolving models that describe the new system. Th e evolving models include functional models
(see Chapter 4), structural models (see Chapter 5), and behavioral models (see Chapter 6).15 Th e
executive summary provides all critical information in a very concise form. It can be thought
of as a summary of the complete proposal. Its purpose is to allow a busy executive to quickly
read through it and determine which parts of the proposal he or she needs to go through more
thoroughly. Th e executive summary is typically no more than a single page long. Figure 3-11
provides a template for a system proposal and references to where the other sections of the
proposal are described.
The System Proposal 113
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Create a requirements defi nition.
Diff erentiate between a functional and a nonfunctional requirement.
Discuss the problem analysis requirements strategy.
Discuss the root cause analysis requirements strategy.
Discuss the duration analysis requirements strategy.
Discuss the activity-based costing analysis requirements strategy.
Discuss the informal benchmarking analysis requirements strategy.
Discuss the outcome analysis requirements strategy.
Discuss the technology analysis requirements strategy.
Discuss the activity elimination requirements strategy.
Discuss how to use interviews to gather requirements.
Discuss how to use joint application development to gather requirements.
Discuss how to use questionnaires to gather requirements.
Discuss how to use document analysis to gather requirements.
Discuss how to use observation to gather requirements.
Describe how to use concept maps to document requirements.
Describe how to use story cards and task lists to document requirements.
Describe the purpose and contents of system proposal.
KEY TERMS
Activity elimination
Activity-based costing
Analysis
As-is system
Benchmarking
Bottom-up interview
Breadth of analysis
Business requirements
Closed-ended question
Concept mapping
Concept maps
Critical thinking skills
Document analysis
Duration analysis
Electronic JAD (e-JAD)
Facilitator
Formal system
Functional requirements
Ground rules
Informal benchmarking
114 Chapter 3 Requirements Determination
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Chapter 3 introduced requirements determination for object-oriented systems develop-
ment projects. Determining the system’s requirements is the most important activity in
the systems development process. A requirement is WHAT the system must do or WHAT
characteristics it must have. If the requirements are not fully or correctly defi ned, the sys-
tem developed is unlikely to meet the needs of the user. In other words, if the requirements
are wrong, the system will be wrong.
In this chapter’s installment of the Patterson Superstore case, we see the require-
ments analysis and requirement-gathering techniques that the analysts used to determine
requirements for Version 1 of the Integrated Health Clinic Delivery System. We also see the
functional and nonfunctional requirements that were developed and an initial draft of the
developing systems proposal for the project. Th is systems proposal will be fi nalized aft er
the functional (Chapter 4), structural (Chapter 5), and behavioral (Chapter 6) modeling of
the system has been completed.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
Informal system
Interpersonal skills
Interview
Interview notes
Interview report
Interview schedule
JAD (joint application
development)
Nonfunctional requirements
Observation
Open-ended question
Outcome analysis
Parallelization
Process Integration
Post-session report
Potential business value
Probing question
Problem analysis
Project cost
Questionnaire
Requirement
Requirements defi nition
Requirements
determination
Risk
Root cause
Root cause analysis
Sample
Scribe
Story cards
Structured interview
System proposal
System requirements
Task lists, 144
Technology analysis
To-be system
Top-down interview
Unstructured interview
User stories
Walkthrough
QUESTIONS
1. What are the key deliverables that are created during
analysis? What is the fi nal deliverable from analysis,
and what does it contain?
2. What is the diff erence between an as-is system and a
to-be system?
3. What is the purpose of the requirements defi nition?
4. What are the three basic steps of the analysis process?
Which step is sometimes skipped or done in a cursory
fashion? Why?
5. Compare and contrast problem analysis and root
cause analysis. Under what conditions would you use
problem analysis? Under what conditions would you
use root cause analysis?
6. Compare and contrast duration analysis and activity-
based costing.
7. Describe the fi ve major steps in conducting interviews.
8. Explain the diff erences among a closed-ended ques-
tion, an open-ended question, and a probing question.
When would you use each?
9. Explain the diff erences between unstructured inter-
views and structured interviews. When would you use
each approach?
10. Explain the diff erence between a top-down and
bottom-up interview approach. When would you use
each approach?
11. How are participants selected for interviews and JAD
sessions?
12. How can you diff erentiate between facts and opinions?
Why can both be useful?
13. Describe the fi ve major steps in conducting JAD
sessions.
14. How does a JAD facilitator diff er from a scribe?
15. What are the three primary things that a facilitator
does in conducting the JAD session?
16. What is e-JAD, and why might a company be inter-
ested in using it?
17. How does designing questions for questionnaires diff er
from designing questions for interviews or JAD sessions?
18. What are typical response rates for questionnaires,
and how can you improve them?
19. What is document analysis?
20. How does the formal system diff er from the informal
system? How does document analysis help you under-
stand both?
21. What are the key aspects of using observation in the
information-gathering process?
22. Explain factors that can be used to select information-
gathering techniques.
23. What is the primary advantage that concept maps
have over traditional textual requirements documents
techniques?
24. What are some of the advantages of using story cards
and task lists as a requirements-gathering and docu-
mentation technique?
25. What information is typically included in a system
proposal?
26. What is the purpose of the executive summary of the
system proposal?
EXERCISES
A. Review the Amazon.com website. Develop the
requirements defi nition for the site. Create a list
of functional business requirements that the system
meets. What diff erent kinds of nonfunctional business
requirements does the system meet? Provide exam-
ples for each kind.
B. Suppose you are going to build a new system that auto-
mates or improves the interview process for the career
Exercises 115
services department of your school. Develop a require-
ments defi nition for the new system. Include both
functional and nonfunctional system requirements.
Pretend you will release the system in three diff erent
versions. Prioritize the requirements accordingly.
C. Describe in very general terms the as-is business
process for registering for classes at your university.
Collaborate with another student in your class, and
evaluate the process using problem analysis and root
cause analysis. Based on your work, list some improve-
ments that you have identifi ed.
D. Describe in very general terms the as-is business pro-
cess for applying for admission at your university.
Collaborate with another student in your class, and
evaluate the process using informal benchmarking.
Based on your work, list some improvements that you
have identifi ed.
E. Describe in very general terms the as-is business
process for registering for classes at your university.
Collaborate with another student in your class, and
evaluate the process using activity elimination. Based
on your work, list some improvements that you have
identifi ed.
F. Suppose your university is having a dramatic increase
in enrollment and is having diffi culty fi nding enough
seats in courses for students. Perform a technology
analysis to identify new ways to help students com-
plete their studies and graduate.
G. Suppose you are the analyst charged with developing a
new system for the university bookstore so that students
can order books online and have them delivered to their
dorms or off -campus housing. What requirements-
gathering techniques will you use? Describe in detail
how you would apply the techniques.
H. Suppose you are the analyst charged with developing
a new system to help senior managers make bet-
ter strategic decisions. What requirements-gathering
techniques will you use? Describe in detail how you
would apply the techniques.
I. Find a partner and interview each other about what
tasks each did in the last job you held (full-time,
part-time, past, or current). If you haven’t worked
before, then assume your job is being a student.
Before you do this, develop a brief interview plan.
Aft er your partner interviews you, identify the type
of interview, interview approach, and types of ques-
tions used.
J. Find a group of students and run a sixty-minute
JAD session on improving alumni relations at your
university. Develop a brief JAD plan, select two tech-
niques that will help identify improvements, and then
develop an agenda. Conduct the session using the
agenda, and write your post-session report.
K. Find a questionnaire on the Web that has been created
to capture customer information. Describe the pur-
pose of the survey, the way questions are worded, and
how the questions have been organized. How can it be
improved? How will the responses be analyzed?
L. Develop a questionnaire that will help gather infor-
mation regarding processes at a popular restaurant
or the college cafeteria (e.g., ordering, customer ser-
vice). Give the questionnaire to ten to fi ft een students,
analyze the responses, and write a brief report that
describes the results.
M. Contact the career services department at your uni-
versity, and fi nd all the pertinent documents designed
to help students fi nd permanent and/or part-time
jobs. Analyze the documents and write a brief report.
MINICASES
1. Th e State Firefi ghter’s Association has a membership
of 15,000. Th e purpose of the organization is to pro-
vide some fi nancial support to the families of deceased
member fi refi ghters and to organize a conference
each year bringing together fi refi ghters from all over
the state. Members are billed dues and calls annually.
Calls are additional funds required to take care of
payments made to the families of deceased members.
Th e bookkeeping work for the association is handled
by the elected treasurer, Bob Smith, although it is
widely known that his wife, Laura, does all the work.
Bob runs unopposed each year at the election, because
no one wants to take over the tedious and time-
consuming job of tracking memberships. Bob is paid
a stipend of $8,000 per year, but his wife spends well
over twenty hours per week on the job. Th e organiza-
tion, however, is not happy with their performance.
A computer system is used to track the billing and
receipt of funds. Th is system was developed in 1984
by a computer science student and his father. Th e
system is a DOS-based system written using dBase 3.
Th e most immediate problem facing the treasurer and
116 Chapter 3 Requirements Determination
his wife is the fact that the soft ware package no longer
exists, and there is no one around who knows how to
maintain the system. One query, in particular, takes
seventeen hours to run. Over the years, they have just
avoided running this query, although the information
in it would be quite useful. Questions from mem-
bers concerning their statements cannot easily be
answered. Usually Bob or Laura just jots down the
inquiry and returns a call with the answer. Sometimes
it takes three to fi ve hours to fi nd the information
needed to answer the question. Oft en, they have to
perform calculations manually because the system
was not programmed to handle certain types of que-
ries. When member information is entered into the
system, each fi eld is presented one at a time, which
makes it very diffi cult to return to a fi eld and correct
a value that was entered. Sometimes a new member is
entered but disappears from the records. Th e report
of membership used in the conference materials does
not alphabetize members by city. Only cities are listed
in the correct order.
What requirements analysis strategy or strategies
would you recommend for this situation? Explain
your answer.
2. Brian Callahan, IS project manager, is just about ready
to depart for an urgent meeting called by Joe Campbell,
manager of manufacturing operations. A major project
sponsored by Joe recently cleared the approval hurdle,
and Brian helped bring the project through project
initiation. Now that the approval committee has given
the go-ahead, Brian has been working on the project’s
analysis plan.
One evening, while playing golf with a friend who
works in the manufacturing operations department,
Brian learned that Joe wants to push the project’s time
frame up from Brian’s original estimate of thirteen
months. Brian’s friend overheard Joe say, “I can’t see
why that IS project team needs to spend all that time
analyzing things. Th ey’ve got two weeks scheduled
just to look at the existing system! Th at seems like a
real waste. I want that team to get going on building
my system.”
Because Brian has a little inside knowledge about
Joe’s agenda for this meeting, he has been considering
how to handle Joe. What do you suggest Brian tell Joe?
3. Barry has recently been assigned to a project team that
will be developing a new retail store management sys-
tem for a chain of submarine sandwich shops. Barry
has several years of experience in programming, but
he has not done much analysis in his career. He was a
little nervous about the new work he would be doing,
but he was confi dent he could handle any assignment
he was given.
One of Barry’s fi rst assignments was to visit one
of the submarine sandwich shops and prepare an
observation report on how the store operates. Barry
planned to arrive at the store around noon, but he
chose a store in an area of town he was unfamiliar
with, and due to traffi c delays and diffi culty in fi nd-
ing the store, he did not arrive until 1:30. Th e store
manager was not expecting him and refused to let a
stranger behind the counter until Barry had her contact
the project sponsor (the director of store management)
at company headquarters to verify who he was and
what his purpose was.
Aft er fi nally securing permission to observe, Barry
stationed himself prominently in the work area
behind the counter so that he could see everything.
Th e staff had to maneuver around him as they went
about their tasks, but there were only minor occa-
sional collisions. Barry noticed that the store staff
seemed to be going about their work very slowly and
deliberately, but he supposed that was because the
store wasn’t very busy. At fi rst, Barry questioned each
worker about what he or she was doing, but the store
manager eventually asked him not to interrupt their
work so much—he was interfering with their service
to the customers.
By 3:30, Barry was a little bored. He decided to leave,
fi guring he could get back to the offi ce and prepare
his report before 5:00 that day. He was sure his team
leader would be pleased with his quick completion
of his assignment. As he drove, he refl ected, “Th ere
really won’t be much to say in this report. All they
do is take the order, make the sandwich, collect the
payment, and hand over the order. It’s really simple!”
Barry’s confi dence in his analytical skills soared as he
anticipated his team leader’s praise.
Back at the store, the store manager shook her
head, commenting to her staff , “He comes here at the
slowest time of day on the slowest day of the week. He
never even looked at all the work I was doing in the
back room while he was here—summarizing yester-
day’s sales, checking inventory on hand, making up
resupply orders for the weekend . . . plus he never even
considered our store-opening and -closing procedures.
I hate to think that the new store management system
is going to be built by someone like that. I’d better
contact Chuck [the director of store management]
and let him know what went on here today.”
Minicases 117
118 Chapter 3 Requirements Determination
Evaluate Barry’s conduct of the observation
assignment.
4. Anne has been given the task of conducting a survey
of sales clerks who will be using a new order-entry sys-
tem being developed for a household products catalog
company. Th e goal of the survey is to identify the clerks’
opinions on the strengths and weaknesses of the current
system. Th ere are about 50 clerks who work in three
diff erent cities, so a survey seemed like an ideal way of
gathering the needed information from the clerks.
Anne developed the questionnaire carefully and
pretested it on several sales supervisors who were
available at corporate headquarters. Aft er revising it
based on their suggestions, she sent a paper version
of the questionnaire to each clerk, asking that it be
returned within one week. Aft er one week, she had
only three completed questionnaires returned. Aft er
another week, Anne received just two more completed
questionnaires. Feeling somewhat desperate, Anne
then sent out an e-mail version of the questionnaire,
again to all the clerks, asking them to respond to
the questionnaire by e-mail as soon as possible. She
received two e-mail questionnaires and three mes-
sages from clerks who had completed the paper ver-
sion expressing annoyance at being bothered with the
same questionnaire a second time. At this point, Anne
has just a 14 percent response rate, which she is sure
will not please her team leader. What suggestions do
you have that could have improved Anne’s response
rate to the questionnaire?
Functional models describe business processes and the interaction of an information sys-
tem with its environment. In object-oriented systems development, two types of models are
used to describe the functionality of an information system: use cases and activity diagrams.
Use cases are used to describe the basic functions of the information system. Activity dia-
grams support the logical modeling of business processes and workfl ows. Both can be used to
describe the current as-is system and the to-be system being developed. Th is chapter describes
business process and functional modeling as a means to document and understand require-
ments and to understand the functional or external behavior of the system.
OBJECTIVES
■ Understand the process used to identify business processes and use cases.
■ Understand the process used to create use-case diagrams.
■ Understand the process used to model business processes with activity diagrams.
■ Understand the rules and style guidelines for activity diagrams.
■ Understand the process used to create use-case descriptions.
■ Understand the rules and style guidelines for use-case descriptions.
■ Be able to create functional models of business processes using use-case diagrams,
activity diagrams, and use-case descriptions.
INTRODUCTION
Th e previous chapter discussed popular requirements-gathering techniques, such as inter-
viewing, JAD, and observation. Using these techniques, the analyst determined the require-
ments and created a requirements defi nition. Th e requirements defi nition defi ned what the
system is to do. In this chapter, we discuss how the information that is gathered using these
techniques is organized and presented in the form of use-case and activity diagrams and
use-case descriptions. Because Unifi ed Modeling Language (UML) has been accepted as
the standard notation by the Object Management Group (OMG), almost all object-oriented
development projects today use these models to document and organize the requirements
that are obtained during the analysis workfl ow.1
119
1 Other, similar techniques that are commonly used in non-UML projects are task modeling and scenario-based
design. For task modeling, see Ian Graham, Migrating to Object Technology (Reading, MA: Addison-Wesley, 1995);
Ian Graham, Brian Henderson-Sellers, and Houman Younessi, Th e OPEN Process Specifi cation (Reading, MA: Addi-
son-Wesley, 1997). For scenario-based design, see John M. Carroll, Scenario-Based Design: Envisioning Work and
Technology in System Development (New York: Wiley, 1995).
C H A P T E R 4
Business Process and
Functional Modeling
120 Chapter 4 Business Process and Functional Modeling
As pointed out in Chapter 1, all object-oriented systems development approaches are
use-case driven, architecture-centric, and iterative and incremental. A use case is a formal way
of representing the way a business sys tem interacts with its environment. Essentially, a use
case is a high-level overview of the business processes in a business information system. From
a practical perspective, use cases represent the entire basis for an object-oriented system. Use
cases can document the current system (i.e., as-is system) or the new system being developed
(i.e., to-be system). Given that object-oriented systems are use-case driven, use cases also
form the foundation for testing (see Chapter 12) and user-interface design (see Chapter 10).
Two forms of use-case driven testing are walkthroughs (described later in this chapter) and
role-playing (described in Chapter 5).
From an architecture-centric perspective, use-case modeling supports the creation of
an external or functional view of a business process in that it shows how the users view the
process rather than the internal mechanisms by which the process and supporting systems
operate. Th e structural and behavioral architecture-based views are described in Chapters
5 and 6, respectively. Finally, all object-oriented systems development approaches are
developed in an incremental and iterative manner. Even though we present the three
architectural views in a sequential manner, this is done primarily for pedagogical rea-
sons. You will find that you will need to not only iterate across the business process and
functional models (described in this chapter), you will also have to iterate across all
three architectural views to fully capture and represent the requirements for a business
information system.
Activity diagrams are typically used to augment our understanding of the business pro-
cesses and our use-case model. Technically, an activity diagram can be used for any type of
process-modeling activity.2 In this chapter, we describe their use in the context of business
process modeling. Process models depict how a business system operates. Th ey illustrate
the processes or activities that are performed and how objects (data) move among them.
A process model can be used to document a current system (i.e., as-is system) or a new
system being developed (i.e., to-be system), whether computerized or not. Many diff erent
process-modeling techniques are in use today.3
Activity diagrams and use cases are logical models—models that describe the busi-
ness domain’s activities without suggesting how they are conducted. Logical models are
sometimes referred to as problem domain models. Reading a use-case or activity diagram,
in principle, should not indicate if an activity is computerized or manual, if a piece of
information is collected by paper form or via the Web, or if information is placed in
a filing cabinet or a large database. These physical details are defined during design
when the logical models are refined into physical models. These models provide infor-
mation that is needed to ultimately build the system. By focusing on logical activities
first, analysts can focus on how the business should run without being distracted with
implementation details.
2 We actually used an activity diagram to describe a simple process in Chapter 1 (see Figure 1-1).
3 Another commonly used process-modeling technique is IDEF0. IDEF0 is used extensively throughout the
U.S. federal government. For more information about IDEF0, see FIPS 183: Integration Defi nition for Function
Modeling (IDEF0), Federal Information Processing Standards Publications (Washington, DC: U.S. Department
of Commerce, 1993). From an object-oriented perspective, a good book that uses the UML to address business
process modeling is Hans-Erik Eriksson and Magnus Penker, Business Modeling with UML (New York: Wiley,
2000). Finally, a new process modeling technique is BPMN (Business Process Modeling Notation). A good
book that compares the notation and use of BPMN to UML’s activity diagram is Martin Schedlbauer, Th e Art of
Business Process Modeling: Th e Business Analysts Guide to Process Modeling with UML & BPMN (Sudbury, MA: Th e
Cathris Group, 2010).
Last h1 121Business Process Identifi cation with Use Cases and Use-Case Diagrams 121
As a fi rst step, the project team gathers requirements from the users (see Chapter 3).
Using the gathered requirements, the project team then identifi es the business processes and
their environment using use cases and use-case diagrams. Next, users work closely with the
team to model the business processes in the form of activity diagrams, and the team docu-
ments the business processes described in the use-case and activity diagrams by creating a
use-case description for each use case. Finally, the team verifi es and validates the business
processes by ensuring that all three models (use-case diagram, activity diagram(s), and use-
case descriptions) agree with one another. Once the current understanding of the business
processes is documented in the functional models, the team is ready to move on to structural
modeling (see Chapter 5).
In this chapter, we fi rst describe business process identifi cation using use cases and
use-case diagrams. Second, we describe business process modeling with activity dia-
grams. Th ird, we describe use-case descriptions, their elements, and a set of guidelines for
creating them. Fourth, we describe the process of verifi cation and validation of the business
process and functional models.
BUSINESS PROCESS IDENTIFICATION WITH USE CASES
AND USE-CASE DIAGRAMS
In the previous chapter, we learned about strategies and techniques that are useful in iden-
tifying the diff erent business processes of a system so that a requirements defi nition could
be created. In this section, we learn how to begin modeling business processes with use
cases and the use-case diagram. An analyst can employ use cases and the use-case diagram
to better understand the functionality of the system at a very high level. Typically, because
a use-case diagram provides a simple, straightforward way of communicating to the users
exactly what the system will do, a use-case diagram is drawn when gathering and defi ning
requirements for the system. In this manner, the use-case diagram can encourage the users
to provide additional high-level requirements. A use-case diagram illustrates in a very simple
way the main functions of the system and the diff erent kinds of users that will interact with
it. Figure 4-1 describes the basic syntax rules for a use-case diagram. Figure 4-2 presents
a use-case diagram for the doctor’s offi ce appointment system introduced in the previous
chapter. We can see from the diagram that patients, doctors, and management personnel
will use the appointment system to manage appointments, record availability, and produce
schedules, respectively.
Elements of Use-Case Diagrams
Th e elements of a use-case diagram include actors, use cases, subject boundaries, and a set of
relationships among actors, actors and use cases, and use cases. Th ese relationships consist
of association, include, extend, and generalization relationships. Each of these elements is
described next.
Actors Th e stick fi gures on the diagram represent actors (see Figure 4-1). An actor is not a
specifi c user but instead is a role that a user can play while interacting with the system. An
actor can also represent another system in which the current system interacts. In this case,
the actor optionally can be represented by a rectangle containing <> and the name
of the system. Basically, actors represent the principal elements in the environment in which
122 Chapter 4 Business Process and Functional Modeling
An actor:
■ Is a person or system that derives benefit from and is external to the subject.
■ Is depicted as either a stick figure (default) or, if a nonhuman actor is involved, a
rectangle with <> in it (alternative).
■ Is labeled with its role.
■ Can be associated with other actors using a specialization/superclass association,
denoted by an arrow with a hollow arrowhead.
■ Is placed outside the subject boundary.
A use case:
■ Represents a major piece of system functionality.
■ Can extend another use case.
■ Can include another use case.
■ Is placed inside the system boundary.
■ Is labeled with a descriptive verb–noun phrase.
A subject boundary:
■ Includes the name of the subject inside or on top.
■ Represents the scope of the subject, e.g., a system or an individual
business process.
An include relationship:
■ Represents the inclusion of the functionality of one use case within another.
■ Has an arrow drawn from the base use case to the used use case.
An extend relationship:
■ Represents the extension of the use case to include optional behavior.
■ Has an arrow drawn from the extension use case to the base use case.
A generalization relationship:
■ Represents a specialized use case to a more generalized one.
■ Has an arrow drawn from the specialized use case to the base use case.
An association relationship:
■ Links an actor with the use case(s) with which it interacts.
<>
Actor/Role
Subject
Actor/Role
Use Case
<>
<>
* *
FIGURE 4-1 Syntax for Use-Case Diagram
the system operates. Actors can provide input to the system, receive output from the system,
or both. Th e diagram in Figure 4-2 shows that three actors will interact with the appointment
system (a patient, a doctor, and management).
Sometimes an actor plays a specialized role of a more general type of actor. For example,
there may be times when a new patient interacts with the system in a way that is somewhat
diff erent from a general patient. In this case, a specialized actor (i.e., new patient) can be
placed on the model, shown using a line with a hollow triangle at the end of the more- general
Business Process Identifi cation with Use Cases and Use-Case Diagrams 123
actor (i.e., patient). Th e specialized actor inherits the behavior of the more general actor and
extends it in some way (see Figure 4-3).
Association Use cases are connected to actors through association relationships; these rela-
tionships show with which use cases the actors interact (see Figure 4-1). A line drawn from
an actor to a use case depicts an association. Th e association typically represents two-way
communication between the use case and the actor. If the communication is only one way,
then a solid arrowhead can be used to designate the direction of the fl ow of information.
Appointment System
Patient
Produce Schedules
Manage
Appointments
Management
Doctor
Record
Availability
* *
* *
* *
Appointment System
Patient
New Patient
Produce Schedules
Manage
Appointments
Management
Doctor
Record
Availability
* *
* *
* *
FIGURE 4-3
Use-Case Diagram
with a Specialized
Actor
FIGURE 4-2
Use-Case Diagram
for the Appointment
System
124 Chapter 4 Business Process and Functional Modeling
For example, in Figure 4-2 the Patient actor communicates with the Manage Appointments
use case. Because there are no arrowheads on the association, the communication is two-
way. Finally, it is possible to represent the multiplicity of the association. Figure 4-2 shows
an asterisk (*) at either end of the association between the Patient and the Manage Appoint-
ments use case. Th is simply indicates that an individual patient (instance of the Patient actor)
executes the Manage Appointments use case as many times as he or she wishes and that it is
possible for the appointment part of the Manage Appointments use case to be executed by
many diff erent patients. In most cases, this type of many-to-many relationship is appropri-
ate. However, it is possible to restrict the number of patients who can be associated with the
Manage Appointments use case. We discuss the multiplicity issue in detail in the next chapter
in regard to class diagrams.
Use Case A use case, depicted by an oval in the UML, is a major process that the system
performs and that benefi ts an actor or actors in some way (see Figure 4-1); it is labeled using
a descriptive verb–noun phrase. We can tell from Figure 4-2 that the system has three primary
use cases: Manage Appointments, Produce Schedule, and Record Availability.
Th ere are times when a use case includes, extends, or generalizes the functionality of
another use case in the diagram. Th ese are shown using include, extend, and generalization
relationships. To increase the ease of understanding a use-case diagram, higher-level use cases
are normally drawn above the lower-level ones. It may be easier to understand these relation-
ships with the help of examples. Let’s assume that every time a patient makes an appointment,
the patient is asked to verify payment arrangements. However, it is occasionally necessary to
actually make new payment arrangements. Th erefore, we may want to have a use case called
Make Payment Arrangements that extends the Manage Appointments use case to include
this additional functionality. In Figure 4-4, an arrow labeled with extend was drawn from the
Make Payment Arrangements use case to the Manage Appointment use case to denote this
special use-case relationship. Th e Make Payment Arrangements use case was drawn lower
than the Manage Appointments use case.
Similarly, there are times when a single use case contains common functions that are
used by other use cases. For example, suppose there is a use case called Manage Schedule that
performs some routine tasks needed to maintain the doctor’s offi ce appointment schedule,
and the two use cases Record Availability and Produce Schedule both perform the routine
tasks. Figure 4-4 shows how we can design the system so that Manage Schedule is a shared
use case that is used by others. An arrow labeled with include is used to denote the include
relationship, and the included use case is drawn below the use cases that contain it. Notice
that the arrows are drawn from the Record Availability and Produce Schedule use cases to
the common Manage Schedule use case.
Finally, there are times when it makes sense to use a generalization relationship to
simplify the individual use cases. For example in Figure 4-4, the Manage Appointments use
case has been specialized to include a use case for an Old Patient and a New Patient. Th e
Make Old Patient Appt use case inherits the functionality of the Manage Appointments
use case (including the Make Payment Arrangements use-case extension) and extends its
own functionality with the Update Patient Information use case. Th e Make New Patient
Appt use case also inherits all the functionality of the generic Manage Appointments use
case and calls the Create New Patient use case, which includes the functionality neces-
sary to insert the new patient into the patient database. Th e generalization relationship
is represented as an unlabeled hollow arrow with the more general use case being higher
than the lower use cases. Also, notice that we have added a second specialized actor, Old
Patient, and that the Patient actor is now simply a generalization of the Old and New
Patient actors.
Last h1 125Business Process Identifi cation with Use Cases and Use-Case Diagrams 125
Subject Boundary Th e use cases are enclosed within a subject boundary, which is a box
that defi nes the scope of the system and clearly delineates what parts of the diagram are
external or internal to it (see Figure 4-1). One of the more diffi cult decisions to make is
where to draw the subject boundary. A subject boundary can be used to separate a soft –
ware system from its environment, a subsystem from other subsystems within the soft ware
system, or an individual process in a soft ware system. Th ey also can be used to separate
an information system, including both soft ware and internal actors, from its environment.
Care should be taken to decide what the scope of the information system is to be.
Th e name of the subject can appear either inside or on top of the box. Th e subject
boundary is drawn based on the scope of the system. In the appointment system, we assumed
that the Management and Doctor actors are outside of the scope of the system; that is, they
use the system. We could have included a receptionist as an actor. However, in this case, we
assumed that the receptionist is an internal actor who is part of the Manage Appointments
Appointment System
Patient
New Patient
Old Patient
Produce Schedules
Update Patient
Information
Make Payment
Arrangements
Make Old
Patient Appt
Make New
Patient Appt
Create New
Patient
Manage
Appointments
Management
Doctor
Record
Availability
Manage
Schedule
<
>
<>
<
>
<>
<>
* *
*
*
*
*
*
*
FIGURE 4-4 Extend and Include Relationships
126 Chapter 4 Business Process and Functional Modeling
use case with which the Patient actor interacts. Th erefore, the receptionist is not drawn on
the diagram.4
Identifying the Major Use Cases
Th e fi rst step is to review the requirements defi nition (see Figure 3-1). Th is helps the analyst
to get a complete overview of the underlying business process being modeled.
Th e second step is to identify the subject’s boundaries. Th is helps the analyst to identify the
scope of the system. However, as we work through the development process, the boundary of
the system most likely will change.
Th e third step is to identify the primary actors and their goals. Th e primary actors involved
with the system come from a list of stakeholders and users. Recall that a stakeholder is a per-
son, group, or organization that can aff ect (or will be aff ected by) a new system, whereas an
actor is a role that a stakeholder or user plays, not a specifi c user (e.g., doctor, not Dr. Jones).
Th e goals represent the functionality that the system must provide the actor for the system
to be a success. Identifying the tasks that each actor must perform can facilitate this. For
example, does the actor need to create, read, update, delete, or execute (CRUDE)5 any infor-
mation currently in the system, are there any external changes of which an actor must inform
the system, or is there any information that the system should give the actor? Steps 2 and 3
are intertwined. As actors are identifi ed and their goals are uncovered, the boundary of the
system will change.
Th e fourth step is to simply identify the business processes and major use cases. Rather than
jumping into one use case and describing it completely at this point, we only want to identify the
use cases. Identifying only the major use cases at this time prevents the users and analysts from
forgetting key business processes and helps the users explain the overall set of business processes
for which they are responsible. It is important at this point to understand and defi ne acronyms
and jargon so that the project team and others from outside the user group can clearly understand
the use cases. Again, the requirements defi nition is a very useful beginning point for this step.
Th e fi fth step is to carefully review the current set of use cases. It may be necessary to split
some of them into multiple use cases or merge some of them into a single use case. Also,
based on the current set, a new use case may be identifi ed. You should remember that iden-
tifying use cases is an iterative process, with users oft en changing their minds about what a
use case is and what it includes. It is very easy to get trapped in the details at this point, so
you need to remember that the goal at this step is to only identify the major use cases. For
example, in the doctor’s offi ce example in Figure 4-2, we defi ned one use case as Manage
Appointments. Th is use case included the cases for both new patients and existing patients,
as well as for when a patient changes or cancels an appointment. We could have defi ned each
of these activities (makes an appointment, changes an appointment, or cancels an appoint-
ment) as separate use cases, but this would have created a huge set of small use cases.
Th e trick is to select the right size so that you end up with three to nine use cases in each
system. If the project team discovers many more than eight use cases, this suggests that the use
cases are too small or that the system boundary is too big. If more than nine use cases exist, the
4 In other non-UML approaches to object-oriented systems development, it is possible to represent external actors
along with internal actors. In this example, the receptionist would be considered an internal actor (see Graham,
Migrating to Object Technology, and Graham, Henderson-Sellers, and Younessi, Th e OPEN Process Specifi cation).
5 We describe the use of CRUDE analysis and matrices in Chapter 6.
1. Review Require –
ments Defi nition
2. Identify Subject’s
Boundaries
3. Identify Primary
Actors & Goals
4. Identify Business
Processes & Major
Use Cases
5. Review Current
Set of Use Cases
Last h1 127Business Process Identifi cation with Use Cases and Use-Case Diagrams 127
use cases should be grouped together into packages (i.e., logical groups of use cases) to make the
diagrams easier to read and keep the models at a reasonable level of complexity. It is simple at
that point to sort the use cases and group together these small use cases into larger use cases that
include several small ones or to change the system boundaries.6
Creating a Use-Case Diagram
Basically, drawing the use-case diagram is straightforward once use cases have been detailed.
Th e actual use-case diagram encourages the use of information hiding. Th e only parts drawn
on the use-case diagram are the system boundary, the use cases themselves, the actors, and the
various associations between these components. Th e major strength of the use-case diagram
is that it provides the user with an overview of the business processes. However, remember
that any time a use case changes, it could aff ect the use case diagram. Th ere are four major
steps in drawing a use-case diagram.
First, we place and draw the use cases on the diagram. Th ese are taken directly from the major
use cases previously identifi ed. Special use-case associations (include, extend, or generaliza-
tion) are also added to the model at this point. Be careful in laying out the diagram. Th ere is
no formal order to the use cases, so they can be placed in whatever fashion is needed to make
the diagram easy to read and to minimize the number of lines that cross. It oft en is necessary
to redraw the diagram several times with use cases in diff erent places to make the diagram
easy to read. Also, for understandability purposes, there should be no more than three to nine
use cases on the model counting use cases that have been factored out and now are associated
with another use case through the include, extend, or generalization relationships.
Second, the actors are placed and drawn on the diagram. To minimize the number of lines
that cross on the diagram, the actors should be placed near the use cases with which they are
associated.
Th ird, the subject boundary is drawn. Th is forms the border of the subject, separating use
cases (i.e., the subject’s functionality) from actors (i.e., the roles of the external users).
Th e fourth and last step is to add associations by drawing lines to connect the actors to the use
cases with which they interact. No order is implied by the diagram, and the items added along the
way do not have to be placed in a particular order; therefore, it might help to rearrange the symbols
a bit to minimize the number of lines that cross, making the diagram less confusing.
Campus Housing Example Identify the actors and major use cases for the following high-
level business processes in a housing system run by the campus housing service. Th e campus
housing service helps students fi nd apartments. Apartment owners complete information
forms about the available rental units (e.g., location, number of bedrooms, monthly rent),
which are then entered into a database. Students can search this database via the Web to fi nd
apartments that meet their needs (e.g., a two-bedroom apartment for $400 or less per month
within a half mile of campus) and contact the apartment owners directly to see the apartment
and possibly rent it. Apartment owners call the service to delete their listing when they have
rented their apartment(s).
As a fi rst step, we identify the primary actors, major business processes, and major use
cases. In this case, the primary actors are the apartment owners and the students. Th e goal of
6 For those familiar with structured analysis and design, packages serve a similar purpose as the leveling and
balancing processes used in data fl ow diagramming. Packages are described in Chapter 7.
1. Place & Draw
Use Cases
2. Place & Draw
Actors
3. Draw Subject
Boundary
4. Add Associations
128 Chapter 4 Business Process and Functional Modeling
the primary actors is both sides of a rental transaction, i.e., to rent the apartments. Th e major
business processes and use cases to allow the actors to realize their goal are to maintain the
available rental unit information for the apartment owners and to fi nd appropriate rental units
to consider for the students. Using the identifi ed actors and use cases and following the pro-
cess described above, the use-case diagram in Figure 4-5 was created. Notice that the diagram
only includes two use cases and two actors. In this case, the Maintain Available Rental Unit
Information use case actually includes two separate subprocesses. Th e apartment owners can
add a rental unit that has become available, and they can delete a rental unit that has been rented
and is no longer available. A student can search the Search Available Rental Units use case by
using three separate criteria: distance from campus, number of bedrooms, and monthly rent.
Th ese criteria can be used individually or by any combination of the three. We will return to
this example in the next section of the chapter. However, before we move on, we next describe
a slightly more involved system for a university library.
Library Example Th e functional requirements for an automated university library circulation
system include the need to support searching, borrowing, and book-maintenance activities. Th e
system should support searching by title, author, keywords, and ISBN. Searching the library’s
collection database should be available on terminals in the library and available to potential
borrowers via the Web. If the book of interest is currently checked out, a valid borrower should
be allowed to request the book to be returned. Once the book has been checked back in, the
borrower requesting the book should be notifi ed of the book’s availability.
Th e borrowing activities are built around checking books out and returning books by bor-
rowers. Th ere are three types of borrowers: students, faculty or staff , and guests. Regardless of
the type of borrower, the borrower must have a valid ID card. If the borrower is a student, hav-
ing the system check with the registrar’s student database validates the ID card. If the borrower
is a faculty or staff member, having the system check with the personnel offi ce’s employee data-
base validates the ID card. If the borrower is a guest, the ID card is checked against the library’s
own borrower database. If the ID card is valid, the system must also check to determine whether
the borrower has any overdue books or unpaid fi nes. If the ID card is invalid, the borrower has
overdue books, or the borrower has unpaid fi nes, the system must reject the borrower’s request
to check out a book, otherwise the borrower’s request should be honored. If a book is checked
out, the system must update the library’s collection database to refl ect the book’s new status.
Th e book-maintenance activities deal with adding and removing books from the library’s
book collection. Th is requires a library manager to both logically and physically add and remove
the book. Books being purchased by the library or books being returned in a damaged state
typically cause these activities. If a book is determined to be damaged when it is returned and it
needs to be removed from the collection, the last borrower will be assessed a fi ne. However, if the
book can be repaired, depending on the cost of the repair, the borrower might not be assessed a
fi ne. Every Monday, the library sends reminder e-mails to borrowers who have overdue books.
If a book is overdue more than two weeks, the borrower is assessed a fi ne. Depending on how
long the book remains overdue, the borrower can be assessed additional fi nes every Monday.
Campus Housing System
Apartment
Owner
Maintain Available
Rental Unit
Information* *
* *
Student
Search Available
Rental Units
FIGURE 4-5
Campus Housing
Use-Case Diagram
To begin we need to identify the major use cases and create a use-case diagram that
represents the high-level business processes in the business situation just described. Based on
the steps to identify the major use cases, we need to review the requirements defi nition and
identify the boundaries (scope) of the problem. Based on the description of the problem, it is
obvious that the system to be created is limited to managing the library’s book collection. Th e
next thing we need to do is to identify the primary actors and business processes that need
to be supported by the system. Based on the functional requirements described, the primary
actors are borrowers and librarians, whereas the primary business processes are borrowing
books, returning books, searching the book collection, maintaining the book collection, and
processing overdue books. Now that we have identifi ed all of the actors and major use cases, we
can draw the use-case diagram that represents an overview of the library’s book collection manage-
ment system (see Figure 4-6). Notice the addition of two nonhuman actors (Personnel Offi ce
and Registrar Offi ce).
BUSINESS PROCESS MODELING WITH ACTIVITY DIAGRAMS
Business process models describe the diff erent activities that, when combined, support a busi-
ness process. Business processes typically cut across functional departments (e.g., the creation
of a new product involves many diff erent activities that combine the eff orts of many employees
Library Book
Collection
Management
System
Maintain Book
Collection
Process Overdue
Books
Librarian
Borrow Books
* *
* *
*
*
*
Borrower
*
*
* *
*
<>
Personnel Office
*
<>
Registrar Office
Search Collection
Return Books
FIGURE 4-6
Library Book
Collection Management
System Use-Case
Diagram
Business Process Modeling with Activity Diagrams 129
130 Chapter 4 Business Process and Functional Modeling
in many departments). From an object-oriented perspective, these processes cut across mul-
tiple objects. Many of the earlier object-oriented systems development approaches tended to
ignore business process modeling. However, today we realize that modeling business pro-
cesses themselves is a very constructive activity that can be used to make sense of the gathered
requirements (see Chapter 3). Th e one potential problem of building business process models,
from an object-oriented systems development perspective, is that they tend to reinforce a
functional decomposition mindset. However, as long as they are used properly, business pro-
cess models are very powerful tools for communicating the analyst’s current understanding
of the requirements to the user.
Martin Schedlbauer provides a set of best practices to follow when modeling business
processes.7
■ Be realistic, because it is virtually impossible to identify everything that is included
in a business process at this point in the evolution of the system. Even if we could
identify everything, everything is not equally important.
■ Be agile because even though we might not identify every single feature of a
business process, the features that we do identify should be identifi ed in a rigorous
manner.
■ All modeling is a collaborative/social activity. Th erefore, business process mode-
ling must be performed with teams, not by individuals. When an individual creates a
model, the chance of mixing up or omitting important tasks is greatly increased.
■ Do not use a CASE tool to do the modeling but use whiteboards instead.
However, once the process is understood, it is a good idea to use a CASE tool to doc-
ument the process.
■ Process modeling should be done in an iterative manner. As you better understand
a business process, you will need to return to the documented version of the process
and revise it.
■ When modeling a business process, stay focused on that specifi c process.
If tasks associated with other business processes are identifi ed, simply record
them on a to-do list and get back to the business process that you are currently
modeling.
■ Remember that a business process model is an abstraction of reality. By that, we
mean that you should not include every minor task in the current description of the
business process. Remember, you cannot aff ord to lose sight of the proverbial forest
for the sake of detailed understanding of a single tree. Too many details at this point
in the evolution of the system can cause confusion and actually prevent you from
solving the underlying problem being addressed by the new system.
Activity diagrams are used to model the behavior in a business process independent
of objects. Activity diagrams can be used to model everything from a high-level business
workfl ow that involves many diff erent use cases, to the details of an individual use case, all
the way down to the specifi c details of an individual method. In a nutshell, activity diagrams
can be used to model any type of process.8 In this chapter, we restrict our coverage of activity
diagrams to documenting and modeling high-level business processes.
7 Martin Schedlbauer, Th e Art of Business Process Modeling: Th e Business Analysts Guide to Process Modeling with
UML & BPMN (Sudbury, MA: Th e Cathris Group, 2010).
8 Technically speaking, activity diagrams combine process-modeling ideas from many diff erent techniques, includ-
ing event models, statecharts, and Petri nets. However, UML 2.0’s activity diagram has more in common with
Petri nets than the other process-modeling techniques. For a good description of using Petri nets to model busi-
ness workfl ows, see Wil van der Aalst and Kees van Hee, Workfl ow Management: Models, Methods, and Systems
(Cambridge, MA: MIT Press, 2002).
Business Process Modeling with Activity Diagrams 131
Elements of an Activity Diagram
Activity diagrams portray the primary activities and the relationships among the activities in
a process. Figure 4-7 shows the syntax of an activity diagram. Figure 4-8 presents a simple
activity diagram that represents the Manage Appointments use case of the appointment sys-
tem for the doctor’s offi ce example.9
Actions and Activities Actions and activities are performed for some specifi c business
reason. Actions and activities can represent manual or computerized behavior. Th ey are
depicted in an activity diagram as a rounded rectangle (see Figure 4-7). Th ey should have a
name that begins with a verb and ends with a noun (e.g., Get Patient Information or Make
Payment Arrangements). Names should be short, yet contain enough information so that the
reader can easily understand exactly what they do. Th e only diff erence between an action and
an activity is that an activity can be decomposed further into a set of activities and/or actions,
whereas an action represents a simple nondecomposable piece of the overall behavior being
modeled. Typically, only activities are used for business process or workfl ow modeling. In
most cases, each activity is associated with a use case. Th e activity diagram in Figure 4-8
shows a set of separate but related activities for the Manage Appointments use case (see
Figures 4-2, 4-3, and 4-4): Get Patient Information, Update Patient Information, Create New
Patient, Make Payment Arrangements, Make New Appointment, Change Appointment,
and Cancel Appointment. Notice that the Make Payment Arrangements and Make New
Appointment activities appear twice in the diagram: once for an “old” patient and once for
a “new” patient.
Object Nodes Activities and actions typically modify or transform objects. Object nodes
model these objects in an activity diagram. Object nodes are portrayed in an activity diagram
as rectangles (see Figure 4-7). Th e name of the class of the object is written inside the rectan-
gle. Essentially, object nodes represent the fl ow of information from one activity to another
activity. Th e simple appointment system portrayed in Figure 4-8 shows object nodes fl owing
from Get Patient Information activity.
Control Flows and Object Flows Th ere are two diff erent types of fl ows in activity dia-
grams: control and object (see Figure 4-7). Control fl ows model the paths of execution
through a business process. A control fl ow is portrayed as a solid line with an arrowhead
on it showing the direction of fl ow. Control fl ows can be attached only to actions or activ-
ities. Figure 4-8 portrays a set of control fl ows through the doctor’s offi ce’s appointment
system. Object fl ows model the fl ow of objects through a business process. Because activi-
ties and actions modify or transform objects, object fl ows are necessary to show the actual
objects that fl ow into and out of the actions or activities. An object fl ow is depicted as a
dashed line with an arrowhead on it showing the direction of fl ow. An individual object
fl ow must be attached to an action or activity on one end and an object node on the other
end. Figure 4-8 portrays a set of control and object fl ows through the appointment system
of a doctor’s offi ce.
9 Owing to the actual complexity of the syntax of activity diagrams, we follow a minimalist philosophy in our
coverage [see John M. Carrol, Th e Nurnberg Funnel: Designing Minimalist Instruction for Practical Computer Skill
(Cambridge, MA: MIT Press, 1990)]. However, the material contained in this section is based on the Unifi ed Mod-
eling Language: Superstructure Version 2.5, ptc/2010-11-14 (www.uml.org). Additional useful references include
Michael Jesse Chonoles and James A. Schardt, UML 2 for Dummies (Indianapolis, IN: Wiley, 2003); Hans-Erik
Eriksson, Magnus Penker, Brian Lyons, and David Fado, UML 2 Toolkit (Indianapolis: Wiley, 2004); Kendall Scott,
Fast Track UML 2.0 (Berkeley, CA: Apress, 2004). For a complete description of all diagrams, see www.uml.org.
132 Chapter 4 Business Process and Functional Modeling
An action:
■ Is a simple, nondecomposable piece of behavior.
■ Is labeled by its name.
An activity:
■ Is used to represent a set of actions.
■ Is labeled by its name.
Activity
Action
An object node:
■ Is used to represent an object that is connected to a set of object flows.
■ Is labeled by its class name.
A decision node:
■ Is used to represent a test condition to ensure that the control flow or object flow
only goes down one path.
■ Is labeled with the decision criteria to continue down the specific path.
A control flow:
■ Shows the sequence of execution.
A final-activity node:
■ Is used to stop all control flows and object flows in an activity (or action).
An initial node:
■ Portrays the beginning of a set of actions or activities.
A merge node:
■ Is used to bring back together different decision paths that were created using a
decision node.
A fork node:
Is used to split behavior into a set of parallel or concurrent flows of activities (or
A swimlane:
A join node:
Is used to bring back together a set of parallel or concurrent flows of activities (or
An object flow:
■ Shows the flow of an object from one activity (or action) to another activity (or action).
A final-flow node:
■ Is used to stop a specific control flow or object flow.
Swimlane
[Decision
Criteria]
[Decision
Criteria]
Is used to break up an activity diagram into rows and columns to assign the
individual activities (or actions) to the individuals or objects that are responsible
for executing the activity (or action)
Is labeled with the name of the individual or object responsible
Class Name
actions)
actions)
FIGURE 4-7 Syntax for an Activity Diagram
Control Nodes Th ere are seven diff erent types of control nodes in an activity diagram: initial,
fi nal-activity, fi nal-fl ow, decision, merge, fork, and join (see Figure 4-7). An initial node por-
trays the beginning of a set of actions or activities. An initial node is shown as a small fi lled-in
circle. A fi nal-activity node is used to stop the process being modeled. Any time a fi nal-activity
Business Process Modeling with Activity Diagrams 133
Get Patient Information
Appt
Request Info
Appt
Request Info
Create New Patient
Update Patient Information
[New Patient][Old Patient]
[Create] [Change]
Cancel Appointment Change AppointmentCreate Appointment
Make Payment Arrangements
Create Appointment
Make Payment Arrangements
[New Info]
[New Arrange]
[Cancel]
FIGURE 4-8 Activity Diagram for the Manage Appointments Use Case
node is reached, all actions and activities are ended immediately, regardless of whether they are
completed. A fi nal-activity node is represented as a circle surrounding a small, fi lled-in circle,
making it resemble a bull’s-eye. A fi nal-fl ow node is similar to a fi nal-activity node, except that
it stops a specifi c path of execution through the business process but allows the other concur-
rent or parallel paths to continue. A fi nal-fl ow node is shown as a small circle with an X in it.
134 Chapter 4 Business Process and Functional Modeling
Th e decision and merge nodes support modeling the decision structure of a business pro-
cess. Th e decision node is used to represent the actual test condition that determines which of the
paths exiting the decision node is to be traversed. In this case, each exiting path must be labeled
with a guard condition. A guard condition represents the value of the test for that particular
path to be executed. For example, in Figure 4-8, the decision node immediately below the Get
Patient Information activity has two mutually exclusive paths that could be executed: one for
old, or previous, patients and the other for new patients. Th e merge node is used to bring back
together multiple mutually exclusive paths that have been split based on an earlier decision (e.g.,
the old- and new-patient paths in Figure 4-8 are brought back together near the bottom of the
diagram). However, sometimes, for clarity, it is better not to use a merge node. For example, in
Figure 4-9, which of the two activity diagrams, both representing an overview level of an order
process, is easier to understand, the one on the left or the one on the right? Th e one on the left
contains a merge node for the More Items on Order question, but the one on the right does
not. In a sense, the decision node is playing double duty in the diagram on the right: It also
serves as a merge node. Technically speaking, we should not omit the merge node; however,
[Item Available] [Item Not Available]
[More Items
on Order]
[No More Items on Order]
Place Order
Process Order
Back Order ItemProcess Item Process Item
[Item Available] [Item Not Available]
[More Items
on Order]
[No More Items on Order]
Place Order
Process Order
Back Order Item
FIGURE 4-9 Two Very Similar Activity Diagrams
Business Process Modeling with Activity Diagrams 135
sometimes being technically correct according to the UML’s diagramming rules actually causes
the diagram to become confusing. From a business process modeling perspective, a good deal
of common sense can go a long way.
Th e fork and join nodes allow parallel and concurrent processes to be modeled (see
Figure 4-7). Th e fork node is used to split the behavior of the business process into multiple
parallel or concurrent fl ows. Unlike the decision node, the paths are not mutually exclusive
(i.e., both paths are executed concurrently). For example, in Figure 4-10, the fork node
firstParent secondParent
GetJelly GetBread
GetDrink GetDessert
GetPeanutButter
CreateSandwich
CreateLunch
GetLunchBox
PutLunchInBox
FIGURE 4-10
Activity Diagram for
Making a School Box
Lunch
136 Chapter 4 Business Process and Functional Modeling
is used to show that two concurrent, parallel processes are to be executed. In this case, each
process is executed by two separate processors (parents). Th e purpose of the join node is sim-
ilar to that of the merge node. Th e join node simply brings back together the separate parallel
or concurrent fl ows in the business process into a single fl ow.
Swimlanes Activity diagrams can model a business process independent of any object imple-
mentation. However, there are times when it helps to break up an activity diagram in such a
way that it can be used to assign responsibility to objects or individuals who would actually
perform the activity. Th is is especially useful when modeling a business workfl ow and is
accomplished through the use of swimlanes. In Figure 4-10, the swimlanes are used to break
up among two parents the making of a school lunch comprising a peanut butter and jelly
sandwich, a drink, and dessert. In this case, we use vertical swimlanes. We could also draw the
activity diagram using more of a left -to-right orientation instead of a top-down orientation.
In that case, the swimlanes are drawn horizontally.
In an actual business workfl ow, there would be activities that should be associated with
roles of individuals involved in the business workfl ow (e.g., employees or customers) and the
activities to be accomplished by the information system being created. Th is association of
activities with external roles, internal roles, and the system is very useful when creating the
use-case descriptions described later in this chapter.
Guidelines for Creating Activity Diagrams
Scott Ambler suggests the following guidelines when creating activity diagrams:10
■ Because an activity diagram can be used to model any kind of process, you should
set the context or scope of the activity being modeled. Once you have determined
the scope, you should give the diagram an appropriate title.
■ You must identify the activities, control fl ows, and object fl ows that occur between
the activities.
■ You should identify any decisions that are part of the process being modeled.
■ You should attempt to identify any prospects for parallelism in the process.
■ You should draw the activity diagram.
When drawing an activity diagram, the diagram should be limited to a single initial node
that starts the process being modeled. Th is node should be placed at the top or top left of the
diagram, depending on the complexity of the diagram. For most business processes, there
should only be a single fi nal-activity node. Th is node should be placed at the bottom or bot-
tom right of the diagram (see Figures 4-8, 4-9, and 4-10). Because most high-level business
processes are sequential, not parallel, the use of a fi nal-fl ow node should be limited.
When modeling high-level business processes or workfl ows, only the more important
decisions should be included in the activity diagrams. In those cases, the guard conditions
associated with the outfl ows of the decision nodes should be mutually exclusive. Th e outfl ows
and guard conditions should form a complete set (i.e., all potential values of the decision are
associated with one of the fl ows).
As in decision modeling, forks and joins should be included only to represent the more
important parallel activities in the process. For example, an alternative version of Figure 4-10
10 Th e guidelines presented here are based on work done by Scott Ambler. For more details, see Scott W. Ambler, Th e
Object Primer: Th e Application Developer’s Guide to Object Orientation and the UML, 2nd Ed. (Cambridge, England:
Cambridge University Press/SIGS Books, 2001); Scott W. Ambler, Th e Elements of UML Style (Cambridge, England:
Cambridge University Press, 2003).
Business Process Modeling with Activity Diagrams 137
might not include the forks and joins associated with the Get Jelly, Get Bread, Get Peanut
Butter, Get Drink, and Get Dessert activities. Th is would greatly simplify the diagram.11
When laying out the activity diagram, line crossings should be minimized to enhance
the readability of the diagram. Th e activities on the diagram should also be laid out in a left –
to-right and/or top-to-bottom order based on the order in which the activities are executed.
For example, in Figure 4-10, the Create Sandwich activity takes place before the Create Lunch
activity.
Swimlanes should be used only to simplify the understanding of an activity diagram.
Furthermore, the swimlanes should enhance the readability of a diagram. For example,
when using a horizontal orientation for swimlanes, the top swimlane should represent the
most important object or individual involved with the process. Th e order of the remain-
ing swimlanes should be based on minimizing the number of fl ows crossing the diff erent
swimlanes. Also, when there are object fl ows among the activities associated with the dif-
ferent individuals (swimlanes) executing the activities of the process, it is useful to show the
actual object fl owing from one individual to another individual by including an object node
between the two individuals (i.e., between the two swimlanes). Th is, of course, aff ects how the
swimlanes should be placed on the diagram.
Finally, any activity that does not have any outfl ows or any infl ows should be challenged.
Activities with no outfl ows are referred to as black-hole activities. If the activity is truly an
end point in the diagram, the activity should have a control fl ow from it to a fi nal-activity or
fi nal-fl ow node. An activity that does not have any infl ow is known as a miracle activity. In
this case, the activity is missing an infl ow either from the initial node of the diagram or from
another activity.
Creating Activity Diagrams
Th ere are fi ve steps in creating an activity diagram to document and model a business process.
First, you must choose a business process that was previously identifi ed to model. To do this,
you should review the requirements defi nition (see Figure 3-1) and the use-case diagram (see
Figures 4-2, 4-3, and 4-4) created to represent the requirements. You should also review all
of the documentation created during the requirements-gathering process (see Chapter 3),
e.g., reports created that documented interviews or observations, any output from any JAD
sessions, any analysis of any questionnaires used, and any story cards or task lists created. In
most cases, the use cases on the use-case diagram will be the best place to start. For example,
in the appointment system, we had identifi ed three primary use cases: Manage Appointments,
Produce Schedule, and Record Doctor Availability. We also identifi ed a whole set of minor use
cases (these will be useful in identifying the elements of the activity diagram).
Second, identify the set of activities necessary to support the business process. For exam-
ple, in Figure 3-1, three processes are identifi ed as being part of the Manage Appointments
business process. Also, by reviewing the use-case diagram (see Figure 4-4), we see that fi ve
minor use cases are associated with the Manage Appointments major use case. Based on this
information, we can identify a set of activities. In this case, the activities are Update Patient
Information, Make Payment Arrangements, Create New Patient, Create Appointment,
Cancel Appointment, and Change Appointment.
Th ird, identify the control fl ows and nodes necessary to document the logic of the business
process. For example, in Figure 4-4, the Make Payment Arrangements and Update Patient
11 In fact, the only reason we depicted the diagram in Figure 4-10 with the multiple fork and join nodes was to
demonstrate that it could be done.
1. Choose a Business
Process
2. Identify Activities
3. Identify Control
Flows & Nodes
138 Chapter 4 Business Process and Functional Modeling
Information use cases are extensions to the Manage Appointments and Make Old Patient
Appt uses cases. We know that these use cases are executed only in certain circumstances.
From this we can infer that the activity diagram must include some decision and merge
nodes. Based on the requirements defi nition (see Figure 3-1), we can infer another set of deci-
sion and merge nodes based on the Create Appointment, Cancel Appointment, and Change
Appointment activities identifi ed in the previous step.
Fourth, identify the object fl ows and nodes necessary to support the logic of the business
process. Typically object nodes and fl ows are not shown on many activity diagrams used
to model a business process. Th e primary exception is if information captured by the
system in one activity is used in an activity that is performed later, but not immediately
aft er the activity that captured the information. In the appointment example, it is obvious
that we need to be able to determine whether the patient is an old or new patient and the
type of action that the patient would like to have performed (create, cancel, or change an
appointment). It is obvious that a new patient cannot cancel or change an appointment
because the patient is by defi nition a new patient. Obviously, we need to capture this type
of information at the beginning of the business process and use it when required. For
example, in the appointment problem, we need to have a Get Patient Information activity
that captures the appropriate information and makes it available at the appropriate time
in the process.
Fift h, lay out and draw the activity diagram to document the business process. For esthetic
and understandability reasons, just as when drawing a use-case diagram, you should attempt
to minimize potential line crossings. Based on the previous steps and carefully laying
out the diagram, the activity diagram in Figure 4-8 was created to document the Manage
Appointments business process.
Campus Housing Example Th e fi rst step in detailing the identifi ed business processes
(Maintain Available Rental Unit Information and Search Available Rental Units) is to choose
one of them. In this example, we are going to focus on the Maintain Available Rental Unit
Information associated with the apartment owners. Based on the earlier description, there are
two separate activities (subprocesses): one to add a rental unit and one to delete a rental unit.
To add a rental unit, the apartment owner must provide the campus housing service with
the location of the apartment, the number of bedrooms in the apartment, and the monthly
rent of the apartment. To delete an apartment, the apartment owners must tell the campus
housing service that the specifi c apartment has been rented and is no longer available. Using
this information, the activity diagram that represents the logical description of the Maintain
Available Rental Unit Information use case is portrayed in Figure 4-11. Notice that there is
absolutely no reference to fi lling out a form, entering the information into a database, or
searching the database. Th ere are actually many diff erent potential ways in which the apart-
ment information could be captured, e.g., on a manual form, on a computerized form, on a
Web form, and via a mobile interface. In fact, we might want to be able to support all of them.
Also, there are many diff erent ways in which the information can be stored. However, at this
stage of development, all design and implementation details should be ignored. We are only
interested in capturing the functional requirements. Once we have successfully modeled the
functional requirements, we can move on to the nonfunctional requirements, design, and
implementation details. We will return to this example in the next section of the chapter.
However, before we move on, we next describe the activity diagram for the Borrow Books use
case of the university library problem.
4. Identify Object
Flows & Nodes
5. Lay Out & Draw
Diagram
Business Process Modeling with Activity Diagrams 139
Library Example As with the Campus Housing example, the fi rst step is to choose a business
process to model. In this case, we want to create an activity diagram for the Borrow Books use
case (see Figure 4-6). Th e functional requirements for this use case were:
Th e borrowing activities are built around checking books out and returning books by
borrowers. Th ere are three types of borrowers: students, faculty or staff , and guests.
Regardless of the type of borrower, the borrower must have a valid ID card. If the bor-
rower is a student, having the system check with the registrar’s student database vali-
dates the ID card. If the borrower is a faculty or staff member, having the system check
with the personnel offi ce’s employee database validates the ID card. If the borrower
is a guest, the ID card is checked against the library’s own borrower database. If the
ID card is valid, the system must also check to determine whether the borrower has
any overdue books or unpaid fi nes. If the ID card is invalid, the borrower has overdue
books, or the borrower has unpaid fi nes, the system must reject the borrower’s request
to check out a book, otherwise the borrower’s request should be honored.
Th e second step to model a business process is to identify the activities that make up the pro-
cess. Based on the requirements for the Borrow Books use case, we can identify three major
activities: Validate ID Card, Check for Overdue Books and Fines, and Check Out Books. Th e
third step is to identify the control fl ows and control nodes necessary to model the decision
logic of the business process. In this case, there obviously will have to be an initial node, a
fi nal-fl ow node, and a set of decision and merge nodes for each decision to be made. Th e
fourth step is to identify the object fl ows and object nodes necessary to complete the descrip-
tion of the business process. In this case, there really is no need to include object nodes and
fl ows. Finally, we can lay out the diagram (see Figure 4-12).
FIGURE 4-11
Campus Housing
Maintain Available
Rental Unit
Information Activity
Diagram
[Apartment Rented][Apartment Available]
Delete Rental UnitCapture Location
Capture Number of
Bedrooms
Capture Monthly
Rent
Add Rental Unit
140 Chapter 4 Business Process and Functional Modeling
BUSINESS PROCESS DOCUMENTATION WITH USE CASES
AND USE-CASE DESCRIPTIONS
Use-case diagrams provided a bird’s-eye view of the basic functionality of the business pro-
cesses contained in the evolving system. Activity diagrams, in a sense, open up the black box of
each business process by providing a more-detailed graphical view of the underlying activities
that support each business process. Use-case descriptions provide a means to more fully doc-
ument the diff erent aspects of each individual use case.12 Th e use-case descriptions are based
on the identifi ed requirements, use-case diagram, and the activity diagram descriptions of the
business processes. Use-case descriptions contain all the information needed to document the
functionality of the business processes.13
Use cases are the primary drivers for all the UML diagramming techniques. A use case
communicates at a high level what the system needs to do, and all the UML diagramming tech-
niques build on this by presenting the use-case functionality in a diff erent way for a diff erent
purpose. Use cases are the building blocks by which the system is designed and built.
12 For a more detailed description of use-case modeling, see Alistair Cockburn, Writing Eff ective Use Cases (Reading,
MA: Addison-Wesley, 2001).
13 Nonfunctional requirements, such as reliability requirements and performance requirements, are oft en docu-
mented outside of the use case through more traditional requirements documents. See Gerald Kotonya and Ian
Sommerville, Requirements Engineering (Chichester, England: Wiley, 1998); Benjamin L. Kovitz, Practical Soft ware
Requirements: A Manual of Content & Style (Greenwich, CT: Manning, 1999); Dean Leffi ngwell and Don Widrig,
Managing Soft ware Requirements: A Unifi ed Approach (Reading, MA: Addison-Wesley, 2000); Richard H. Th ayer,
M. Dorfman, and Sidney C. Bailin (eds.), Soft ware Requirements Engineering, 2nd Ed. (Los Alamitos, CA: IEEE
Computer Society, 1997).
[Valid Card]
[No Overdue Books & No Fines]
Validate ID Card
Check Out Books
Check for Overdue Books and Fines
FIGURE 4-12
Activity Diagram for
the Borrow Books
Use Case
Business Process Documentation with Use Cases and Use-Case Descriptions 141
Use cases capture the typical interaction of the system with the system’s users (end
users and other systems). Th ese interactions represent the external, or functional, view of the
system from the perspective of the user. Each use case describes one and only one function
in which users interact with the system. Although a use case may contain several paths that
a user can take while interacting with the system, each possible execution path through the
use case is referred to as a scenario. Another way to look at a scenario is as if a scenario is
an instantiation of a specifi c use case. Scenarios are used extensively in behavioral modeling
(see Chapter 6). Finally, by identifying all scenarios and trying to execute them through
role-playing CRC cards (see Chapter 5), you will be testing the clarity and completeness of
your evolving understanding of the system being developed.14
When creating use-case descriptions, the project team must work closely with the users to
fully document the functional requirements. Organizing the functional requirements and docu-
menting them in a use-case description are a relatively simple process, but it takes considerable
practice to ensure that the descriptions are complete enough to use in structural (Chapter 5) and
behavioral (Chapter 6) modeling. Th e best place to begin is to review the use-case and activity
diagrams. Th e key thing to remember is that each use case is associated with one and only one
role that users have in the system. For example, a receptionist in a doctor’s offi ce may play mul-
tiple roles—he or she can make appointments, answer the telephone, fi le medical records, wel-
come patients, and so on. It is possible that multiple users will play the same role. Th erefore, use
cases should be associated with the roles played by the users and not with the users themselves.
Types of Use Cases
Th ere are many diff erent types of use cases. We suggest two separate dimensions on which to
classify a use case based on the purpose of the use case and the amount of information that
the use case contains: overview versus detail and essential versus real.
An overview use case is used to enable the analyst and user to agree on a high-level over-
view of the requirements. Typically, overview use cases are created very early in the process
of understanding the system requirements, and they document only basic information about
the use case, such as its name; ID number; primary actor; type; a brief description; and the
relationships among the actors, actors and use cases, and use cases. Th ese can easily be created
immediately aft er the creation of the use-case diagram.
Once the user and the analyst agree upon a high-level overview of the requirements, the
overview use cases are converted to detail use cases. A detail use case typically documents, as
far as possible, all the information needed for the use case. Th ese can be based on the activities
and control fl ows contained in the activity diagrams.
An essential use case is one that describes only the minimum essential issues necessary to
understand the required functionality. A real use case goes farther and describes a specifi c set
of steps. For example, an essential use case in a doctor offi ce might say that the receptionist
should attempt to match the patient’s desired appointment times with the available times,
whereas a real use case might say that the receptionist should look up the available dates on
the calendar using Google Calendar to determine if the requested appointment times were
available. Th e primary diff erence is that essential use cases are implementation independent,
whereas real use cases are detailed descriptions of how to use the system once it is imple-
mented. Th us, real use cases tend to be used only in the design, implementation, and testing.
Elements of a Use-Case Description
A use-case description contains all the information needed to build the structural (Chapter 5)
and behavioral (Chapter 6) diagrams that follow, but it expresses the information in a less-formal
14 For presentation purposes, we defer discussion of role-playing to Chapter 5.
142 Chapter 4 Business Process and Functional Modeling
way that is usually simpler for users to understand. Figure 4-13 shows a sample use-case
description.15 A use-case description has three basic parts: overview information, relationships,
and the fl ow of events.
Overview Information Th e overview information identifi es the use case and provides basic
background information about the use case. Th e use-case name should be a verb–noun phrase
(e.g., Make Old Patient Appt). Th e use-case ID number provides a unique way to fi nd every
use case and also enables the team to trace design decisions back to a specifi c requirement.
Th e use-case type is either overview or detail and essential or real. Th e primary actor is usually
the trigger of the use case—the person or thing that starts the execution of the use case. Th e
primary purpose of the use case is to meet the goal of the primary actor. Th e brief description
is typically a single sentence that describes the essence of the use case.
Th e importance level can be used to prioritize the use cases. Th e importance level enables
the users to explicitly prioritize which business functions are most important and need to be
part of the fi rst version of the system and which are less important and can wait until later
versions if necessary. Th e importance level can use a fuzzy scale, such as high, medium, and
low (e.g., in Figure 4-13 we have assigned an importance level of high to the Make Old Patient
Appt use case). It can also be done more formally using a weighted average of a set of criteria.
For example, Larman16 suggests rating each use case over the following criteria using a scale
from zero to fi ve:
■ Th e use case represents an important business process.
■ Th e use case supports revenue generation or cost reduction.
■ Technology needed to support the use case is new or risky and therefore requires
considerable research.
■ Functionality described in the use case is complex, risky, and/or time critical.
Depending on a use case’s complexity, it may be useful to consider splitting its imple-
mentation over several diff erent versions.
■ Th e use case could increase understanding of the evolving design relative to the
eff ort expended.
A use case may have multiple stakeholders that have an interest in the use case. Each use
case lists each of the stakeholders with each one’s interest in the use case (e.g., Old Patient and
Doctor). Th e stakeholders’ list always includes the primary actor (e.g., Old Patient).
Each use case typically has a trigger—the event that causes the use case to begin (e.g., Old
Patient calls and asks for a new appointment or asks to cancel or change an existing appoint-
ment). A trigger can be an external trigger, such as a customer placing an order or the fi re
alarm ringing, or it can be a temporal trigger, such as a book being overdue at the library or
the need to pay the rent.
Relationships Use-case relationships explain how the use case is related to other use cases
and users. Th ere are four basic types of relationships: association, extend, include, and gener-
alization. An association relationship documents the communication that takes place between
the use case and the actors that use the use case. An actor is the UML representation for the
15 Currently there is no standard set of elements for a use case. Th e elements described in this section are based on
recommendations contained in Alistair Cockburn, Writing Eff ective Use Cases (Reading, MA: Addison-Wesley,
2001); Craig Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis and Design and
the Unifi ed Process, 2nd Ed. (Upper Saddle River, NJ: Prentice Hall, 2002); Brian Henderson-Sellers and Bhuvan
Unhelkar, OPEN Modeling with UML (Reading, MA: Addison-Wesley, 2000). Also see Graham, Migrating to Object
Technology.
16 Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis and Design.
Business Process Documentation with Use Cases and Use-Case Descriptions 143
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Make Old Patient Appt 2 Low
Old Patient
Old Patient – wants to make, change, or cancel an appointment
Doctor – wants to ensure patient’s needs are met in a timely manner
Use Case Type: Detail, Essential
Patient calls and asks for a new appointment or asks to cancel or change an existing appointment
Old Patient
Update Patient Information
Manage Appointments
1. The Patient contacts the office regarding an appointment.
2. The Patient provides the Receptionist with his or her name and address.
3. If the Patient’s information has changed
Execute the Update Patient Information use case.
4. If the Patient’s payment arrangements has changed
Execute the Make Payments Arrangements use case.
5. The Receptionist asks Patient if he or she would like to make a new appointment, cancel an existing appointment, or change
an existing appointment.
S-1: New Appointment
1. The Receptionist asks the Patient for possible appointment times.
2. The Receptionist matches the Patient’s desired appointment times with available dates and
times and schedules the new appointment.
S-2: Cancel Appointment
1. The Receptionist asks the Patient for the old appointment time.
2. The Receptionist finds the current appointment in the appointment file and cancels it.
S-3: Change Appointment
1. The Receptionist performs the S-2: cancel appointment subflow.
2. The Receptionist performs the S-1: new appointment subflow.
S-1, 2a1: The Receptionist proposes some alternative appointment times based on what is available in the
appointment schedule.
S-1, 2a2: The Patient chooses one of the proposed times or decides not to make an appointment.
6. The Receptionist provides the results of the transaction to the Patient.
This use case describes how we make an appointment as well as changing or canceling
an appointment for a previously seen patient.
Relationships:
Association:
Include:
Extend:
Generalization:
If the patient wants to make a new appointment,
the S-1: new appointment subflow is performed.
If the patient wants to cancel an existing appointment,
the S-2: cancel appointment subflow is performed.
If the patient wants to change an existing appointment,
the S-3: change appointment subflow is performed.
FIGURE 4-13 Sample Use-Case Description
role that a user plays in the use case. For example, in Figure 4-13, the Make Old Patient Appt
use case is associated with the actor Old Patient (see Figure 4-4). In this case, a patient makes
an appointment. All actors involved in the use case are documented with the association
relationship.
TEMPLATE
can be found at
www.wiley.com
/college/dennis
144 Chapter 4 Business Process and Functional Modeling
An include relationship represents the mandatory inclusion of another use case. Th e
include relationship enables functional decomposition—the breaking up of a complex use
case into several simpler ones. For example, in Figure 4-4, the Manage Schedule use case was
considered to be complex and complete enough to be factored out as a separate use case that
could be executed by the Produce Schedules and Record Availability use cases. Th e include
relationship also enables parts of use cases to be reused by creating them as separate use cases.
An extend relationship represents the extension of the functionality of the use case to
incorporate optional behavior. In Figure 4-13, the Make Old Patient Appt use case condi-
tionally uses the Update Patient Information use case. Th is use case is executed only if the
patient’s information has changed.
Th e generalization relationship allows use cases to support inheritance. For example,
the use case in Figure 4-4, the Manage Appointments use case, was specialized so that
a new patient would be associated with the Make New Patient Appt and an old patient
could be associated with a Make Old Patient Appt. Th e common, or generalized, behav-
ior that both the Make New Patient Appointment and Make Old Patient Appointment
use cases contain would be placed in the generalized Manage Appointments use case.
In other words, the Make New Patient Appointment and Make Old Patient Appointment
use cases would inherit the common functionality from the Manage Appointments use
case. Th e specialized behavior would be placed in the appropriate specialized use case.
For example, the extend relationship to the Update Patient Information use case would be
placed with the specialized Make Old Patient Appointment use case.
Flow of Events Finally, the individual steps within the business process are described. Th ree
diff erent categories of steps, or fl ows of events, can be documented: normal fl ow of events,
subfl ows, and alternative, or exceptional, fl ows:
■ Th e normal fl ow of events includes only steps that normally are executed in a use
case. Th e steps are listed in the order in which they are performed. In Figure 4-13,
the patient and the receptionist have a conversation regarding the patient’s name,
address, and action to be performed.
■ In some cases, the normal fl ow of events should be decomposed into a set of subfl ows
to keep the normal fl ow of events as simple as possible. In Figure 4-13, we have
identifi ed three subfl ows: Create Appointment, Cancel Appointment, and Change
Appointment. Each of the steps of the subfl ows is listed. Th ese subfl ows are based on
the control fl ow logic in the activity diagram representation of the business process
(see Figure 4-7). Alternatively, we could replace a subfl ow with a separate use case
that could be incorporated via the include relationships (see the earlier discussion).
However, this should be done only if the newly created use case makes sense by itself.
For example, in Figure 4-13, does it make sense to factor out a Create Appointment,
Cancel Appointment, and/or Change Appointment use case? If it does, then the spe-
cifi c subfl ow(s) should be replaced with a call to the related use case, and the use case
should be added to the include relationship list.
■ Alternative or exceptional fl ows are ones that do happen but are not considered to be
the norm. Th ese must be documented. For example, in Figure 4-13, we have identi-
fi ed two alternative or exceptional fl ows. Th e fi rst one simply addresses the situation
that occurs when the set of requested appointment times is not available. Th e second
one is simply a second step to the alternative fl ow. Like the subfl ows, the primary
purpose of separating out alternate or exceptional fl ows is to keep the normal fl ow of
events as simple as possible. Again, as with the subfl ows, it is possible to replace the
alternate or exceptional fl ows with separate use cases that could be integrated via the
extend relationship (see the earlier discussion).
Business Process Documentation with Use Cases and Use-Case Descriptions 145
When should events be factored out from the normal fl ow of events into subfl ows? When
should subfl ows and/or alternative or exceptional fl ows be factored out into separate use cases?
Or when should things simply be left alone? Th e primary criteria should be based on the level of
complexity that the use case entails. Th e more diffi cult it is to understand the use case, the more
likely events should be factored out into subfl ows, or subfl ows and/or alternative or exceptional
fl ows should be factored out into separate use cases that are called by the current use case. Th is,
of course, creates more use cases. Th erefore, the use-case diagram will become more cluttered.
In other words, the choice that the analyst must make is to have a more complex use-case dia-
gram with simpler use cases or have a simpler use-case diagram with more complex use cases.
Practically speaking, we must decide which makes more sense. Th is varies greatly, depending
on the problem and the client. Remember, we are trying to represent, in a manner as complete
and concise as possible, our understanding of the business processes that we are investigating
so that the client can validate the requirements that we are modeling. Th erefore, there really is
no single right answer. It really depends on the analyst, the client, and the problem.
Optional Characteristics Other characteristics of use cases can be documented by use-case
descriptions. Th ese include the level of complexity of the use case; the estimated amount of
time it takes to execute the use case; the system with which the use case is associated; specifi c
data fl ows between the primary actor and the use case; any specifi c attribute, constraint, or
operation associated with the use case; any preconditions that must be satisfi ed for the use
case to execute; or any guarantees that can be made based on the execution of the use case. As
we noted at the beginning of this section, there is no standard set of characteristics of a use
case that must be captured. We suggest that the information contained in Figure 4-13 is the
minimal amount to be captured.
Guidelines for Creating Use-Case Descriptions17
Th e essence of a use case is the fl ow of events. Writing the fl ow of events in a manner that is
useful for later stages of development generally comes with experience.
First, write each individual step in the form subject–verb–direct object and, optionally,
preposition–indirect object. Th is form has become known as SVDPI sentences. Th is form of
sentence has proved to be useful in identifying classes and operations (see Chapter 5). For
example, in Figure 4-13, the fi rst step in the normal fl ow of events, the Patient contacts the
offi ce regarding an appointment, suggests the possibility of three classes of objects: Patient,
Offi ce, and Appointment. Th is approach simplifi es the process of identifying the classes in
the structural model (see Chapter 5). SVDPI sentences cannot be used for all steps, but they
should be used whenever possible.
Second, make clear who or what is the initiator of the action and who or what is the
receiver of the action in each step. Normally, the initiator should be the subject of the sen-
tence and the receiver should be the direct object of the sentence. For example, in Figure
4-13, the second step, Patient provides the Receptionist with his or her name and address,
clearly portrays the Patient as the initiator and the Receptionist as the receiver.
Th ird, write the step from the perspective of an independent observer. To accomplish
this, each step might have to be written fi rst from the perspective of both the initiator and
the receiver. Based on the two points of view, the bird’s-eye view version can then be written.
For example, in Figure 4-13, the Patient provides the Receptionist with his or her name and
address, neither the patient’s nor the receptionist’s perspective is represented.
Fourth, write each step at the same level of abstraction. Each step should make about
the same amount of progress toward completing the use case as each of the other steps.
17 Th ese guidelines are based on Cockburn, Writing Eff ective Use Cases, and Graham, Migrating to Object Technology.
146 Chapter 4 Business Process and Functional Modeling
On high-level use cases, the amount of progress could be very substantial, whereas in a
low-level use case, each step could represent only incremental progress. For example, in
Figure 4-13, each step represents about the same amount of eff ort to complete.
Fift h, ensure that the use case contains a sensible set of actions. Each use case should
represent a transaction. Th erefore, each use case should comprise four parts:
1. Th e primary actor initiates the execution of the use case by sending a request
(and possibly data) to the system.
2. Th e system ensures that the request (and data) is valid.
3. Th e system processes the request (and data) and possibly changes its own internal state.
4. Th e system sends the primary actor the result of the processing.
For example, in Figure 4-13, the patient requests an appointment (steps 1 and 2), the
receptionist determines whether any of the patient’s information has changed or not (step 3),
the receptionist determines whether the patient’s payments arrangements have changed or
not (step 4), the receptionist sets up the appointment transaction (step 5), and the receptionist
provides the results of the transaction to the patient (step 6).
Th e sixth guideline is the KISS principle. If the use case becomes too complex and/or too
long, the use case should be decomposed into a set of use cases. Furthermore, if the normal fl ow
of events of the use case becomes too complex, subfl ows should be used. For example, in Figure
4-13, the fi ft h step in the normal fl ow of events was suffi ciently complex to decompose it into
three separate subfl ows. However, care must be taken to avoid the possibility of decomposing too
much. Most decomposition should be done with classes (see Chapter 5).
Th e seventh guideline deals with repeating steps. Normally, in a programming language,
we put loop defi nition and controls at the beginning of the loop. However, because the use-
case steps are written in simple English, it is normally better to simply write Repeat steps A
through E until some condition is met aft er step E. It makes the use case more readable to
people unfamiliar with programming.
Creating Use Case Descriptions
Use cases provide a bird’s-eye view of the business processes contained in the evolving system.
Th e use-case diagram depicts the communication path between the actors and the system.
Use cases and their use-case description documentation tend to be used to model both
the contexts of the system and the detailed requirements for the system. Even though the pri-
mary purpose of use cases is to document the functional requirements of the system, they also
are used as a basis for testing the evolving system. In this section, we provide a set of steps that
can be used to guide the actual creation of a use-case description for each use case in the use-
case diagram based on the requirements defi nition and the use-case and activity diagrams.18
Th ese steps are performed in order, but of course the analyst oft en cycles among them in an
iterative fashion as he or she moves from one use case to another use case.
Th e fi rst step is to choose one of the use cases to document with a use-case description. Using
the importance level of the use case can help do this. For example, in Figure 4-13, the Make Old
Patient Appt use case has an importance level of high. As such, it should be one of the earlier use
18 Th e approach in this section is based on the work of Cockburn, Writing Eff ective Use Cases; Graham, Migrating to
Object Technology; George Marakas and Joyce Elam, “Semantic Structuring in Analyst Acquisition and Representa-
tion of Facts in Requirements Analysis,” Information Systems Research 9, no. 1 (1998): 37–63; Alan Dennis, Glenda
Hayes, and Robert Daniels, “Business Process Modeling with Group Support Systems,” Journal of Management
Information Systems 15, no. 4 (1999): 115–142.
1. Choose a
Use Case
Business Process Documentation with Use Cases and Use-Case Descriptions 147
cases to be expanded. Th e criteria suggested by Larman19 can also be used to set the prioritization
of the use cases, as noted earlier. An alternative approach suggests that each use case should be
voted on by each member of the development team. In this approach, each team member is given
a set of “dots” that they can use to vote on the use cases. Th ey can use all of their dots to vote for a
single use case, or they can spread them over a set of use cases. Th e use cases then can be ranked
in order of the number of dots received. Use case descriptions are created for the individual use
cases based on the rank order.20
Th e second step is to create an overview description of the use case; that is, name the primary
actor, set the type for the use case, list all of the identifi ed stakeholders and their interests in
the use case, identify the level of importance of the use case, give a brief description of the use
case, give the trigger information for the use case, and list the relationships in which the use
case participates.
Th e third step is to fi ll in the steps of the normal fl ow of events required to describe each use
case. Th e steps focus on what the business process does to complete the use case, as opposed
to what actions the users or other external entities do. In general, the steps should be listed
in the order in which they are performed, from fi rst to last. Remember to write the steps in
an SVDPI form whenever possible. In writing the use case, remember the seven guidelines
described earlier. Th e goal at this point is to describe how the chosen use case operates. One of
the best ways to begin to understand how an actor works through a use case is to visualize per-
forming the steps in the use case—i.e., role play. Th e techniques of visualizing how to interact
with the system and of thinking about how other systems work (informal benchmarking) are
important techniques that help analysts and users understand how systems work and how to
write a use case. Both techniques (visualization and informal benchmarking) are common in
practice. It is important to remember that at this point in the development of a use case, we
are interested only in the typical successful execution of the use case. If we try to think of all of
the possible combinations of activities that could go on, we will never get anything written. At
this point, we are looking only for the three to seven major steps. Focus only on performing
the typical process that the use case represents.
Th e fourth step is to ensure that the steps listed in the normal fl ow of events are not too
complex or too long. Each step should be about the same size as the others. For example, if
we were writing steps for preparing a meal, steps such as take fork out of drawer and put fork
on table are much smaller than prepare cake using mix. If we end up with more than seven
steps or steps that vary greatly in size, we should go back and review each step carefully and
possibly rewrite the steps.
One good approach to produce the steps for a use case is to have the users visualize
themselves actually performing the use case and to have them write down the steps as if
they were writing a recipe for a cookbook. In most cases, the users will be able to quickly
defi ne what they do in the as-is model. Defi ning the steps for to-be use cases might take a bit
more coaching. In our experience, the descriptions of the steps change greatly as users work
through a use case. Our advice is to use a blackboard or whiteboard (or paper with pencil)
that can be easily erased to develop the list of steps, and then write the list on the use-case
form. It should be written on the use-case form only aft er the set of steps is fairly well defi ned.
19 Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis and Design.
20 C. Larman, Agile & Iterative Development: A Manager’s Guide (Boston, MA: Addison-Wesley, 2004).
2. Create Overview
Description
3. Describe the Normal
Flow of Events
4. Check the Normal
Flow of Events
148 Chapter 4 Business Process and Functional Modeling
Th e fi fth step focuses on identifying and writing the alternative or exceptional fl ows.
Alternative or exceptional fl ows are fl ows of success that represent optional or exceptional
behavior. Th ey tend to occur infrequently or as a result of a normal fl ow failing. Th ey should
be labeled so that there is no doubt as to which normal fl ow of events it is related. For example
in Figure 4-13, alternative/exceptional fl ow S-1, 2a1 executes when step 2 of subfl ow S-1 fails
(i.e., the requested appointment time was not available). Like the normal fl ows and subfl ows,
alternative or exceptional fl ows should be written in the SVDPI form whenever possible.
Th e sixth step is to carefully review the use-case description and confi rm that the use case is
correct as written, which means reviewing the use case with the users to make sure each step
is correct.21 Th e review should look for opportunities to simplify a use case by decomposing it
into a set of smaller use cases, merging it with others, looking for common aspects in both the
semantics and syntax of the use cases, and identifying new use cases. Th is is also the time to
look into adding the include, extend, and/or generalization relationships between use cases. Th e
most powerful way to confi rm a use case is to ask the user to role-play, or execute the process
using the written steps in the use case. Th e analyst hands the user pieces of paper labeled with
the major inputs to the use case and has the user follow the written steps like a recipe to make
sure that those steps really can produce the outputs defi ned for the use case using its inputs.
Th e seventh and fi nal step is to iterate the entire set of steps again. Users oft en change their
minds about what is a use case and what it includes. It is very easy to get trapped in the details
at this point, so remember that the goal is to just address the major use cases. Th erefore,
the analyst should continue to iterate these steps until he or she and the users believe that a
suffi cient number of use cases have been documented to begin identifying candidate classes
for the structural model (see Chapter 5). As candidate classes are identifi ed, it is likely that
additional use cases will be uncovered.
Campus Housing Example Th e fi rst step in documenting a use case with a use-case descrip-
tion is to choose a use case. For instructional purposes, we use the same use case used earlier
with the activity diagrams; the Maintain Available Rental Unit Information, which is associ-
ated with the apartment owners. Th e next step is to create an Overview Description of the use
case. In this case, the primary actor is the apartment owner. Given that the use-case descrip-
tion documents the detailed logic steps for the use case, the type of use case is Detailed and
Essential. Th e Stakeholders include the apartment owners and the campus housing service.
Th eir respective interests are to advertise their available apartments and to provide a service
that enables the apartment owners to rent their available apartments. Th e Brief Description
for this use case is “Th is use case describes how the campus housing service can maintain an
up-to-date listing of available apartments.” Th e trigger for the use case is when an apartment
owner wants to add or delete an available apartment. As such, the trigger is “fi red” from
outside of the system. In this case, the apartment owner’s action triggers this use case. Th ere
is only one Association relationship between this use case and its Primary Actor: Apartment
Owner. Figure 4-14 documents this information.
Next, we document and verify the logical steps necessary to successfully execute this use
case. Th at is, we document the normal fl ow of events, check the normal fl ow of events (pos-
sibly identifying subfl ows), identify any alternative or exceptional fl ows, and then carefully
review the description to be sure that it is complete. If you recall, in this specifi c example, the
apartment owners provided information to add a rental unit to the available rental units or
provided information that identifi ed an available unit that was no longer available and needed
21 Th is process is related to role-playing, which is discussed in Chapter 5.
5. Identify Alternative
or Exceptional Flows
6. Review the Use-
Case Description
7. Repeat Until Done
Business Process Documentation with Use Cases and Use-Case Descriptions 149
to be deleted from the list of available rental units. Th ese two processes were treated as two
subprocesses of the Maintain Available Rental Unit Information use case. Now that we have to
determine which of these subprocesses is to be treated as the Normal Flow of Events and which
is to be treated as an Alternative or Exceptional Flow. However, upon further refl ection, the
question as to whether these should be separated into two independent use cases or whether
they should remained together should be investigated. Th is is a great example where moving
from one representation (activity diagram) to another representation (use case description)
in an iterative and incremental manner raises issues that were not readily apparent. In this
example, it is probably better to replace the Maintain Available Rental Unit Information use
case with two simpler use cases: one for adding a rental unit and one for deleting a rental unit.
Consequently, we now have to change the use-case diagram (see Figure 4-15) and create two
activity diagrams to replace the earlier ones (see Figure 4-16). And, we must create two use-
case descriptions to replace the one that we just begun (see Figures 4-17 and 4-18). We will
return to this example in the next chapter when we begin to create a structural model for the
campus housing service. However, next we return to the university library problem.
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Type: External
Maintain Available Rental Unit Information 1 High
Apartment Owner
Apartment Owner—wants to advertise available apartment
Campus Housing Service—provides a service that enables the apartment owners to rent their available apartments
Use Case Type: Detail, Essential
Apartment Owner wants to add or delete an available apartment
Apartment Owner
This use case describes how the campus housing service can maintain an up-to-date listing of
available apartments.
Relationships:
Association:
Include:
Extend:
Generalization:
Campus Housing System
Apartment
Owner
Add Apartment
Delete Apartment
* *
*
*
*
*
Student
Search Available
Rental Units
FIGURE 4-14 Campus Housing Maintain Available Rental Unit Information Overview
Use-Case Description
FIGURE 4-15
Campus Housing
Use-Case Diagram
150 Chapter 4 Business Process and Functional Modeling
Capture Location
Capture Number of
Bedrooms
Capture Monthly
Rent
Add Apartment
Capture Apartment
Identifier
Delete Apartment
FIGURE 4-16
Campus Add and
Delete Apartment
Activity Diagrams
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Add Apartment 1 High
Apartment Owner
Apartment Owner—wants to advertise available apartment
Campus Housing Service—provides a service that enables the apartment owners to rent their available apartments
Use Case Type: Detail, Essential
Apartment Owner wants to add an available apartment
Apartment Owner
1. Capture the location of the apartment.
2. Capture the number of bedrooms in the apartment.
3. Capture the monthly rent of the apartment.
4. Add the apartment to the listing of available apartments.
This use case describes how the campus housing service can maintain an up-to-date listing of
available apartments.
Relationships:
Association:
Include:
Extend:
Generalization:
FIGURE 4-17 Campus Housing Service Add an Apartment Use-Case Description
Business Process Documentation with Use Cases and Use-Case Descriptions 151
FIGURE 4-18 Campus Housing Service Delete an Apartment Use-Case Description
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Delete Apartment 2 High
Apartment Owner
Apartment Owner—wants to delist apartment
Campus Housing Service—provides a service that enables the apartment owners to rent their available apartments
Use Case Type: Detail, Essential
Apartment Owner wants to delete an available apartment
Apartment Owner
1. Capture the apartment identifier.
2. Delete the apartment from the listing of available apartments.
This use case describes how the campus housing service can maintain an up-to-date listing of
available apartments.
Relationships:
Association:
Include:
Extend:
Generalization:
Library Example As with the Campus Housing example, the first step to document busi-
ness processes with use-case descriptions is to choose a use case. Because we previously
chose the Borrow Books use case in the Library Collection Management System example,
we will stay with it. Next, we need to create the overview description. In this case, we have
to go back and look at the use case diagram (see Figure 4-6) that describes the external
behavior of the Library Collection Management System and the activity diagram (see Fig-
ure 4-12) that describes the functionality of the Borrow Books use case. It also is a good
idea to refer back, once again, to the functional requirements that drove the creation of
the Borrow Books use case. Here they are:
Th e borrowing activities are built around checking books out and returning books by
borrowers. Th ere are three types of borrowers: students, faculty or staff , and guests.
Regardless of the type of borrower, the borrower must have a valid ID card. If the
borrower is a student, having the system check with the registrar’s student database
validates the ID card. If the borrower is a faculty or staff member, having the system
check with the personnel offi ce’s employee database validates the ID card. If the bor-
rower is a guest, the ID card is checked against the library’s own borrower database.
If the ID card is valid, the system must also check to determine whether the borrower
has any overdue books or unpaid fi nes. If the ID card is invalid, the borrower has
overdue books, or the borrower has unpaid fi nes, the system must reject the borrow-
er’s request to check out a book, otherwise the borrower’s request should be honored.
Based on these three critical pieces of information and using the use-case description tem-
plate (see Figure 4-13), we can create the overview description of the Borrow Books use case
(see Figure 4-19).
152 Chapter 4 Business Process and Functional Modeling
By carefully reviewing the functional requirements (above) and the activity diagram
(Figure 4-12), we can easily identify the Normal Flow of Events for the Borrow Books use case.
Furthermore, it is possible to decide whether any of the events contained in the Normal Flow
of Events list should be decomposed using Subfl ows or other use cases that would need to be
included. In the latter case, we would have to modify the Relationships section of the overview
description and modify the use-case diagram to refl ect this addition. Also, based on the logic
structure of the activity diagram, it is possible to identify the alternative exceptional fl ows to
the normal fl ow of events for the Borrow Books use case. Based on the overall simplicity of
the Borrow Books use case, we decided not to decompose the process using either subfl ows or
included use cases. However, due to the logic structure laid out in the activity diagram, there
were two alternate/exceptional fl ows identifi ed. Figure 4-20 depicts the Normal Flow of Events,
Subfl ows, and Alternative/Exceptional Flows sections of the Borrow Books use-case description.
Association: Borrower, Personnel Office, Registrar’s Office
Include:
Extend:
Generalization:
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
Trigger:
Type: External
Use Case Type:
Relationships:
Borrow Books ID: 2 Importance Level: High
Borrower
Borrower brings books to check out desk.
Detail, Essential
This use case describes how books are checked out of the library.
Borrower—wants to check outbooks
Librarian—wants to ensure borrower only gets books deserved
FIGURE 4-19 Overview Description for the Borrow Books Use Case
FIGURE 4-20 Flow Descriptions for the Borrow Books Use Case
Normal Flow of Events:
SubFlows:
1. The Borrower brings books to the Librarian at the check out desk.
2. The Borrower provides Librarian their ID card.
3. The Librarian checks the validity of the ID Card.
If the Borrower is a Student Borrower, Validate ID Card against Registrar’s Database.
If the Borrower is a Faculty/Staff Borrower, Validate ID Card against Personnel Database.
If the Borrower is a Guest Borrower, Validate ID Card against Library’s Guest Database.
4. The Librarian checks whether the Borrower has any overdue books and/or fines
5. The Borrower checks out the books
Alternate/Exceptional Flows:
4a The ID Card is invalid, the book request is rejected.
5a The Borrower either has overdue books, fines, or both, the book request is rejected.
Verifying and Validating the Business Processes and Functional Models 153
VERIFYING AND VALIDATING THE BUSINESS PROCESSES
AND FUNCTIONAL MODELS22
Before we move on to structural (Chapter 5) and behavioral (Chapter 6) modeling, we need to
verify and validate the current set of functional models to ensure that they faithfully represent
the business processes under consideration. Th is includes testing the fi delity of each model;
for example, we must be sure that the activity diagram(s), use-case descriptions, and use-case
diagrams all describe the same functional requirements. Before we describe the specifi c tests
to consider, we describe walkthroughs, a manual approach that supports verifying and vali-
dating the evolving models.23
Verifi cation and Validation through Walkthroughs
A walkthrough is essentially a peer review of a product. In the case of the functional
models, a walkthrough is a review of the diff erent models and diagrams created during
functional modeling. Th is review typically is completed by a team whose members come
from the development team and the client. Th e purpose of a walkthrough is to thoroughly
test the fi delity of the functional models to the functional requirements and to ensure that
the models are consistent. Th at is, a walkthrough uncovers errors or faults in the evolving
specifi cation. However, a walkthrough does not correct errors—it simply identifi es them.
Error correction is to be accomplished by the team aft er the walkthrough is completed.
Walkthroughs are very interactive. As the presenter walks through the representation,
members of the walkthrough team should ask questions regarding the representation. For
example, if the presenter is walking through an activity diagram, another member of the team
could ask why certain activities or objects were not included. Th e actual process of simply
presenting the representation to a new set of eyes can uncover obvious misunderstandings
and omissions. In many cases, the representation creator can get lost in the proverbial trees
and not see the forest.24 In fact, many times the act of walking through the representation
causes a presenter to see the error himself or herself. For psychological reasons, hearing the
representation helps the analyst to see the representation more completely.25 Th erefore, the
representation creators should regularly do a walkthrough of the models themselves by read-
ing the representations out loud to themselves, regardless of how they think it might make
them look.
Th ere are specifi ed roles that diff erent members of the walkthrough team can play. Th e
fi rst is the presenter role. Th is should be played by the person who is primarily responsible
for the specifi c representation being reviewed. Th is individual presents the representation
to the walkthrough team. Th e second role is recorder, or scribe. Th e recorder should be a
member of the analysis team. Th is individual carefully takes the minutes of the meeting
by recording all signifi cant events that occur during the walkthrough. In particular, all
errors that are uncovered must be documented so that the analysis team can address them.
Another important role is to have someone who raises issues regarding maintenance of
22 Th e material in this section has been adapted from E. Yourdon, Modern Structured Analysis (Englewood Cliff s, NJ:
Prentice Hall, 1989). Verifying and validating are types of testing.
23 Even though many modern CASE tools can automate much of the verifying and validating of the analysis models, we
feel that it is paramount that systems analysts understand the principles of verifi cation and validation. Furthermore,
some tools, such as Visio, that support UML diagramming are only diagramming tools. Regardless, the analyst is
expected to perform all diagramming correctly.
24 In fact, all joking aside, in many cases the developer is down at the knothole level and can’t even see the tree, let
alone the forest.
25 Th is has to do with using diff erent senses. Because our haptic senses are the most sensitive, touching the representa-
tion would be best. However, it is not clear how one can touch a use case or a class.
154 Chapter 4 Business Process and Functional Modeling
the representation. Yourdon refers to this individual as a maintenance oracle.26 Owing to
the emphasis on reusability in object-oriented development, this role becomes particularly
crucial. Finally, someone must be responsible for calling, setting up, and running the walk-
through meetings.
For a walkthrough to be successful, the members of the walkthrough team must be fully
prepared. All materials to be reviewed must be distributed with suffi cient time for the team
members to review them before the actual meeting. All team members should be expected to
mark up the representations so that during the walkthrough meeting, all relevant issues can
be discussed. Otherwise, the walkthrough will be ineffi cient and ineff ective. During the actual
meeting, as the presenter is walking through the representation, the team members should point
out any potential errors or misunderstandings. In many cases, the errors and misunderstandings
are caused by invalid assumptions that would not be uncovered without the walkthrough.
One potential danger of walkthroughs is when management decides the results of
uncovering errors in the representation are a refl ection of an analyst’s capability. Th is must
be avoided at all costs. Otherwise, the underlying purpose of the walkthrough—to improve
the fi delity of the representation—will be thwarted. Depending on the organization, it may
be necessary to omit management from the walkthrough process. If not, the walkthrough
process could break down into a slugfest to make some team members to look good by
destroying the presenter. To say the least, this is obviously counterproductive.
Functional Model Verifi cation and Validation
We have suggested three different representations for the functional model: activity dia-
grams, use-case descriptions, and use-case diagrams. In this section, we describe a set of
rules to ensure that these three representations are consistent among themselves.
First, when comparing an activity diagram to a use-case description, there should be at
least one event recorded in the normal fl ow of events, subfl ows, or alternative/exceptional
fl ows of the use-case description for each activity or action that is included on an activity dia-
gram, and each event should be associated with an activity or action. For example, in Figure
4-4, there is an activity labeled Get Patient Information that is associated with the fi rst two
events contained in the normal fl ow of events of the use-case description shown in Figure 4-13.
Second, all objects portrayed as an object node in an activity diagram must be mentioned in
an event in the normal fl ow of events, subfl ows, or alternative/exceptional fl ows of the use-case
description. For example, the activity diagram in Figure 4-4 portrays an Appt object, and the use-
case description refers to a new appointment and changing or canceling an existing appointment.
Th ird, sequential order of the events in a use-case description should occur in the same
sequential order of the activities contained in an activity diagram. For example in Figures 4-4
and 4-13, the events associated with the Get Patient Information activity (events 1 and 2) should
occur before the events associated with the Make Payment Arrangements activity (event 4).
Fourth, when comparing a use-case description to a use-case diagram, there must be one
and only one use-case description for each use case, and vice versa. For example, Figure 4-13
portrays the use-case description of the Make Old Patient Appt use case. However, the
use-case diagram shown in Figure 4-4, the activity diagram shown in Figure 4-8, and the use-
case description given in Figure 4-13 are inconsistent with each other. In this case, the use-case
diagram implies that the Make Payment Arrangements use case is optional regardless of
whether the patient is a new or old patient. However, when we review the activity diagram,
we see that it is an optional activity for old patients, but a required activity for a new patient.
Th erefore, only one of the diagrams is correct. In this instance, the use-case diagram needs to
be corrected. Th e new corrected use-case diagram is shown in Figure 4-21.
26 See Appendix D of Yourdon, Modern Structured Analysis.
Verifying and Validating the Business Processes and Functional Models 155
Fift h, all actors listed in a use-case description must be portrayed on the use-case dia-
gram. Each actor must have an association link that connects it to the use case and must be
listed with the association relationships in the use-case description. For example, the Old
Patient actor is listed in the use-case description of the Make Old Patient Appt use case (see
Figure 4-13), it is listed with the association relationships in the Make Old Patient Appt use-
case description, and it is connected to the use case in the use-case diagram (see Figure 4-21).
Sixth, in some organizations, we should also include the stakeholders listed in the use-
case description as actors in the use-case diagram. For example, there could have been an
association between the Doctor actor and the Make Old Patient Appt use case (see Figures
4-13 and 4-21). However, in this case it was decided not to include this association because
the Doctor never participates in the Make Old Patient Appt use case.27
Appointment System
Patient
New Patient
Old Patient
Produce Schedules
Update Patient
Information
Make Old
Patient Appt
Make New
Patient Appt
Make Payment
Arrangements
Create New
Patient
Manage
Appointments
Management
Doctor
Record
Availability Manage
Schedule
<
>
<>
<
>
<>
<>
*
*
*
*
**
**
<>
FIGURE 4-21 Modifi ed Use-Case Diagram for the Appointment System
27 Another possibility could have been to include a Receptionist actor. However, we had previously decided that the
Receptionist was in fact part of the Appointment System and not simply a user of the system. If UML supported the
idea of internal actors, or actor-to-actor associations, this implicit association could easily be made explicit by having
the Patient actor communicate with the Receptionist actor directly, regardless of whether the Receptionist actor was
part of the system or not. See footnote 4.
156 Chapter 4 Business Process and Functional Modeling
Seventh, all other relationships listed in a use-case description (include, extend, and
generalization) must be portrayed on a use-case diagram. For example, in Figure 4-13,
there is an extend relationship listed with the Update Patient Information use case, and
in Figure 4-21, we see that it appears on the diagram between the two use cases.
Finally, there are many diagram-specifi c requirements that must be enforced. For exam-
ple, in an activity diagram a decision node can be connected to activity or action nodes only
with a control fl ow, and for every decision node there should be a matching merge node.
Every type of node and fl ow has diff erent restrictions. However, the complete restrictions for
all the UML diagrams are beyond the scope of this text.28 Th e concept map in Figure 4-22
portrays the associations among the functional models.
28 A good reference for these types of restrictions is S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, UK:
Cambridge University Press, 2005).
Use Cases
Scenarios
Activity Diagram
Object Nodes
Object Flows
Activities/Actions
Stakeholders
Relationships
Control Flows
Events
Actors
Flows
Including
Contains
HasKinds
Contains
Contains
Have
AssociatedWith
AssociatedWith
AssociatedWith
AssociatedWith
AssociatedWith
AssociatedWith
AssociatedWith
Use-Case Diagram
Functional Models
Use-Case Descriptions
FIGURE 4-22 Interrelationships among Functional Models
Key Terms 157
KEY TERMS
Action
Activity
Activity diagram
Actor
Alternative fl ows
Association relationship
Black-hole activities
Brief description
Control fl ow
Control node
Decision node
Detail use case
Errors
Essential use case
Exceptional fl ows
Extend relationship
External trigger
Faults
Final-activity node
Final-fl ow node
Flow of events
Fork node
Functional decomposition
Generalization relationship
Guard condition
Importance level
Include relationship
Inheritance
Initial node
Iterate
Join node
Logical model
Maintenance oracle
Merge node
Miracle activity
Normal fl ow of events
Object fl ow
Object node
Overview use cases
Packages
Physical model
Presenter
Primary actor
Process models
Real use case
Recorder
Relationships
Role
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
In this chapter, you learned about business processes and functional models. Object-
oriented systems are developed in an incremental and iterative manner. Th is is especially
true when the phased approach is used as in the Patterson Superstore case. Th e team fi rst
developed a use-case diagram for the entire Integrated Health Clinic Delivery System.
Next, the team moved into modeling the processes of Version 1 of the system by creating
an activity diagram and use-case description for Schedule Appointment. You will also
see these models revisited and developed in further iterations as more information is
uncovered. Th e three versions of the Integrated Health Clinic Delivery System will each
go through individual process and functional modeling as well as structural and behavior
modeling with iteration across all of these tasks.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Explain the purpose of a use case in business process and functional modeling.
Describe the diff erent elements of a use-case diagram.
Create use-case diagrams that portray how business information systems interact with their environment.
Explain how to model a specifi c use case with an activity diagram.
Describe the diff erent elements of an activity diagram.
Create an activity diagram that represents a specifi c use case.
Document a business process with a use-case description.
Describe the diff erent types of use cases.
Describe the diff erent elements of a use-case description.
Create a use-case description that represents a specifi c use case.
Verify and validate the evolving functional model using walkthroughs.
Verify and validate the functional model by ensuring the consistency of the three functional representations: use-
case diagrams, activity diagrams, and use-case descriptions.
158 Chapter 4 Business Process and Functional Modeling
Scenario
Scribe
Specialized actor
Stakeholders
Subfl ows
Subject boundary
SVDPI
Swim lanes
Temporal trigger
Test
Trigger
Use case
Use-case description
Use-case diagram
Use-case ID number
Use-case name
Use-case type
Validation
Verifi cation
Walkthrough
QUESTIONS
1. Why is business process modeling important?
2. How do you create use cases?
3. Why do we strive to have about three to nine major
use cases in a business process?
4. How do you create use-case diagrams?
5. How is use-case diagramming related to functional
modeling?
6. Explain the following terms: actor, use case, system
boundary, relationship. Use layperson’s language, as
though you were describing them to a user.
7. Every association must be connected to at least one
_______ and one _________. Why?
8. What are some heuristics for creating a use-case diagram?
9. Why is iteration important in creating use cases?
10. What is the purpose of an activity diagram?
11. What is the diff erence between an activity and an
action?
12. What is the purpose of a fork node?
13. What are the diff erent types of control nodes?
14. What is the diff erence between a control fl ow and an
object fl ow?
15. What is an object node?
16. Explain how a detail use case diff ers from an overview
use case. When are each used?
17. How does an essential use case diff er from a real use case?
18. What are the major elements of an overview use case?
19. What are the major elements of a detail use case?
20. What is the viewpoint of a use case, and why is it
important?
21. What are some guidelines for designing a set of use
cases? Give two examples of the extend associations
on a use-case diagram. Give two examples for the
include associations.
22. Which of the following could be an actor found on a
use-case diagram? Why?
Ms. Mary Smith
Supplier
Customer
Internet customer
Mr. John Seals
Data entry clerk
Database administrator
23. What is CRUD? Why is it useful?
24. What is a walkthrough? How does it relate to verifi ca-
tion and validation?
25. What are the diff erent roles played during a walk-
through? What are their purposes?
26. How are the diff erent functional models related, and
how does this aff ect the verifi cation and validation of
the models?
EXERCISES
A. Investigate the UML website at the Object Manage-
ment Group (www.uml.org). Write a paragraph news
brief on the current state of UML (e.g., the cur-
rent version and when it will be released, future
improvements).
B. Investigate the Object Management Group. Write a
brief memo describing what it is, its purpose, and
its infl uence on UML and the object approach to
systems development. (Hint: A good resource is
www.omg.org.)
C. Draw a use-case diagram and a set of activity dia-
grams for the process of buying glasses from the
viewpoint of the patient. Th e fi rst step is to see an eye
doctor who will give you a prescription. Once you
have a prescription, you go to an optical dispensary,
where you select your frames and place the order for
your glasses. Once the glasses have been made, you
return to the store for a fi tting and pay for the glasses.
D. Create a set of detailed use-case descriptions for the
process of buying glasses in exercise C.
Exercises 159
E. Draw a use-case diagram and a set of activity dia-
grams for the following doctor’s offi ce system. When-
ever new patients are seen for the fi rst time, they
complete a patient information form that asks their
name, address, phone number, and brief medical
history, which are stored in the patient information
fi le. When a patient calls to schedule a new appoint-
ment or change an existing appointment, the recep-
tionist checks the appointment fi le for an available
time. Once a good time is found for the patient, the
appointment is scheduled. If the patient is a new
patient, an incomplete entry is made in the patient’s
fi le; the full information will be collected when the
patient arrives for the appointment. Because appoint-
ments are oft en made far in advance, the receptionist
usually mails a reminder postcard to each patient two
weeks before the appointment.
F. Create a set of detail use-case descriptions for the
dentist’s offi ce system in exercise E.
G. Draw a use-case diagram and a set of activity dia-
grams for an online university registration system.
Th e system should enable the staff of each academic
department to examine the courses off ered by their
department, add and remove courses, and change the
information about them (e.g., the maximum number
of students permitted). It should permit students to
examine currently available courses, add and drop
courses to and from their schedules, and examine the
courses for which they are enrolled. Department staff
should be able to print a variety of reports about the
courses and the students enrolled in them. Th e system
should ensure that no student takes too many courses
and that students who have any unpaid fees are not
permitted to register (assume that fees data are main-
tained by the university’s fi nancial offi ce, which the
registration system accesses but does not change).
H. Create a set of detailed use-case descriptions for the
online university registration system in exercise G.
I. Draw a use-case diagram and a set of activity dia-
grams for the following system. A Real Estate Inc.
(AREI) sells houses. People who want to sell their
houses sign a contract with AREI and provide infor-
mation on their house. Th is information is kept in a
database by AREI, and a subset of this information is
sent to the citywide multiple-listing service used by
all real estate agents. AREI works with two types of
potential buyers. Some buyers have an interest in one
specifi c house. In this case, AREI prints information
from its database, which the real estate agent uses to
help show the house to the buyer (a process beyond
the scope of the system to be modeled). Other buyers
seek AREI’s advice in fi nding a house that meets their
needs. In this case, the buyer completes a buyer infor-
mation form that is entered into a buyer database, and
AREI real estate agents use its information to search
AREI’s database and the multiple-listing service for
houses that meet their needs. Th e results of these
searches are printed and used to help the real estate
agent show houses to the buyer.
J. Create a set of detailed use-case descriptions for the
real estate system in exercise I.
K. Perform a verifi cation and validation walkthrough
of the functional models of the real estate system
described in exercises I and J.
L. Draw a use-case diagram and a set of activity dia-
grams for the following system. A Video Store (AVS)
runs a series of fairly standard video stores. Before
a video can be put on the shelf, it must be cataloged
and entered into the video database. Every customer
must have a valid AVS customer card in order to rent
a video. Customers rent videos for three days at a
time. Every time a customer rents a video, the system
must ensure that he or she does not have any overdue
videos. If so, the overdue videos must be returned and
an overdue fee paid before customer can rent more
videos. Likewise, if the customer has returned overdue
videos but has not paid the overdue fee, the fee must
be paid before new videos can be rented. Every morn-
ing, the store manager prints a report that lists over-
due videos. If a video is two or more days overdue,
the manager calls the customer to remind him or her
to return the video. If a video is returned in damaged
condition, the manager removes it from the video
database and may sometimes charge the customer.
M. Create a set of detailed use-case descriptions for the
video system in exercise L.
N. Perform a verifi cation and validation walkthrough
of the functional models of the video store system
described in exercises L and M.
O. Draw a use-case diagram and a set of activity dia-
grams for a gym membership system. When mem-
bers join the gym, they pay a fee for a certain length
of time. Most memberships are for one year, but
memberships as short as two months are available.
Th roughout the year, the gym off ers a variety of
discounts on their regular membership prices (e.g.,
two memberships for the price of one for Valentine’s
day). It is common for members to pay diff erent
160 Chapter 4 Business Process and Functional Modeling
amounts for the same length of membership. Th e
gym wants to mail out reminder letters to members
asking them to renew their memberships one month
before their memberships expire. Some members
have become angry when asked to renew at a much
higher rate than their original membership contract,
so the club wants to track the prices paid so that the
manager can override the regular prices with spe-
cial prices when members are asked to renew. Th e
system must track these new prices so that renewals
can be processed accurately. One of the problems in
the industry is the high turnover rate of members.
Although some members remain active for many
years, about half of the members do not renew their
memberships. Th is is a major problem, because the
gym spends a lot in advertising to attract each new
member. Th e manager wants the system to track each
time a member comes into the gym. Th e system will
then identify the heavy users and generate a report so
that the manager can ask them to renew their mem-
berships early, perhaps off ering them a reduced rate
for early renewal. Likewise, the system should identify
members who have not visited the gym in more than
a month, so the manager can call them and attempt to
reinterest them in the gym.
P. Create a set of detailed use-case descriptions for the
system in exercise O.
Q. Perform a verifi cation and validation walkthrough of
the functional models of the gym membership system
described in exercises O and P.
R. Draw a use-case diagram and a set of activity dia-
grams for the following system. Picnics R Us (PRU)
is a small catering fi rm with fi ve employees. During
a typical summer weekend, PRU caters fi ft een pic-
nics with twenty to fi ft y people each. Th e business
has grown rapidly over the past year, and the owner
wants to install a new computer system for man-
aging the ordering and buying process. PRU has
a set of ten standard menus. When potential cus-
tomers call, the receptionist describes the menus to
them. If the customer decides to book a picnic, the
receptionist records the customer information (e.g.,
name, address, phone number) and the information
about the picnic (e.g., place, date, time, which one of
the standard menus, total price) on a contract. Th e
customer is then faxed a copy of the contract and
must sign and return it along with a deposit (oft en
a credit card or by debit card) before the picnic is
offi cially booked. Th e remaining money is collected
when the picnic is delivered. Sometimes, the cus-
tomer wants something special (e.g., birthday cake).
In this case, the receptionist takes the information
and gives it to the owner, who determines the cost;
the receptionist then calls the customer back with the
price information. Sometimes the customer accepts
the price; other times, the customer requests some
changes that have to go back to the owner for a new
cost estimate. Each week, the owner looks through
the picnics scheduled for that weekend and orders the
supplies (e.g., plates) and food (e.g., bread, chicken)
needed to make them. Th e owner would like to use
the system for marketing as well. It should be able to
track how customers learned about PRU and identify
repeat customers, so that PRU can mail special off ers
to them. Th e owner also wants to track the picnics for
which PRU sent a contract, but the customer never
signed the contract and actually booked a picnic.
S. Create a set of detailed use-case descriptions for the
system in exercise R.
T. Perform a verifi cation and validation walkthrough
of the functional models of the catering system
described in exercises R and S.
U. Draw a use-case diagram and a set of activity dia-
grams for the following system. Of-the-Month Club
(OTMC) is an innovative young fi rm that sells
memberships to people who have an interest in
certain products. People pay membership fees for
one year and each month receive a product by mail.
For example, OTMC has a coff ee-of-the-month club
that sends members one pound of special coff ee
each month. OTMC currently has six memberships
(coff ee, wine, beer, cigars, fl owers, and computer
games), each of which costs a diff erent amount. Cus-
tomers usually belong to just one, but some belong
to two or more. When people join OTMC, the tele-
phone operator records the name, mailing address,
phone number, e-mail address, credit-card infor-
mation, start date, and membership service(s) (e.g.,
coff ee). Some customers request a double or triple
membership (e.g., two pounds of coff ee, three cases
of beer). Th e computer game membership operates
a bit diff erently from the others. In this case, the
member must also select the type of game (action,
arcade, fantasy/science fi ction, educational, etc.) and
age level. OTMC is planning to greatly expand the
number of memberships it off ers (e.g., video games,
movies, toys, cheese, fruit, and vegetables), so the
system needs to accommodate this future expansion.
OTMC is also planning to off er three-month and
six-month memberships.
Minicases 161
V. Create a set of detailed use-case descriptions for the
system in exercise U.
W. Perform a verifi cation and validation walkthrough of
the functional models of the Of-the-Month Club sys-
tem described in exercises U and V.
MINICASES
1. Williams Specialty Company is a small printing and
engraving organization. When Pat Williams, the
owner, brought computers into the business offi ce fi ve
years ago, the business was very small and very simple.
Pat was able to use an inexpensive PC-based account-
ing system to handle the basic information-processing
needs of the fi rm. As time has gone on, however, the
business has grown and the work being performed
has become signifi cantly more complex. Th e simple
accounting soft ware still in use is no longer adequate
to keep track of many of the company’s sophisticated
deals and arrangements with its customers.
Pat has a staff of four people in the business offi ce
who are familiar with the intricacies of the company’s
record-keeping requirements. Pat recently met with
her staff to discuss her plan to hire an IS consulting
fi rm to evaluate the organization’s information sys-
tem needs and recommend a strategy for upgrading
its computer system. Th e staff are excited about the
prospect of a new system, because the current system
causes them much annoyance. No one on the staff has
ever done anything like this before, however, and they
are a little wary of the consultants who will be con-
ducting the project.
Assume that you are a systems analyst on the con-
sulting team assigned to the Williams Specialty Co.
engagement. At your fi rst meeting with the Williams
staff , you want to be sure that they understand the
work that your team will be performing and how they
will participate in that work.
a. Explain, in clear, nontechnical terms, the goals of
the analysis of the project.
b. Explain, in clear, nontechnical terms, how func-
tional models will be used by the project team to
model the identifi ed business processes. Explain
what these models are, what they represent in the
system, and how they will be used by the team.
2. Professional and Scientifi c Staff Management (PSSM)
is a unique type of temporary staffi ng agency. Many
organizations today hire highly skilled technical
employees on a short-term, temporary basis to assist
with special projects or to provide a needed technical
skill. PSSM negotiates contracts with its client com-
panies in which it agrees to provide temporary staff in
specifi c job categories for a specifi ed cost. For exam-
ple, PSSM has a contract with an oil and gas explora-
tion company in which it agrees to supply geologists
with at least a master’s degree for $5,000 per week.
PSSM has contracts with a wide range of companies
and can place almost any type of professional or sci-
entifi c staff members, from computer programmers to
geologists to astrophysicists.
When a PSSM client company determines that
it will need a temporary professional or scientifi c
employee, it issues a staffi ng request against the con-
tract it had previously negotiated with PSSM. When
PSSM’s contract manager receives a staffi ng request,
the contract number referenced on the staffi ng request
is entered into the contract database. Using informa-
tion from the database, the contract manager reviews
the terms and conditions of the contract and deter-
mines whether the staffi ng request is valid. Th e staff –
ing request is valid if the contract has not expired, the
type of professional or scientifi c employee requested
is listed on the original contract, and the requested
fee falls within the negotiated fee range. If the staffi ng
request is not valid, the contract manager sends the
staffi ng request back to the client with a letter stating
why the staffi ng request cannot be fi lled, and a copy
of the letter is fi led. If the staffi ng request is valid, the
contract manager enters the staffi ng request into the
staffi ng request database as an outstanding staffi ng
request. Th e staffi ng request is then sent to the PSSM
placement department.
In the placement department, the type of staff
member, experience, and qualifi cations requested
on the staffi ng request are checked against the data-
base of available professional and scientifi c staff . If
a qualifi ed individual is found, he or she is marked
“reserved” in the staff database. If a qualifi ed individ-
ual cannot be found in the database or is not imme-
diately available, the placement department creates a
memo that explains the inability to meet the staffi ng
request and attaches it to the staffi ng request. All
staffi ng requests are then sent to the arrangements
department.
162 Chapter 4 Business Process and Functional Modeling
In the arrangements department, the prospective
temporary employee is contacted and asked to agree to
the placement. Aft er the placement details have been
worked out and agreed to, the staff member is marked
“placed” in the staff database. A copy of the staffi ng
request and a bill for the placement fee is sent to the
client. Finally, the staffi ng request, the “unable-to-fi ll”
memo (if any), and a copy of the placement fee bill are
sent to the contract manager. If the staffi ng request
was fi lled, the contract manager closes the open staff –
ing request in the staffi ng request database. If the
staffi ng request could not be fi lled, the client is notifi ed.
Th e staffi ng request, placement fee bill, and unable-to-
fi ll memo are then fi led in the contract offi ce.
a. Create a use-case diagram for the system described
here.
b. Create a set of activity diagrams for the business
processes described here.
c. For each major use case identifi ed in the use-case
diagram, develop both an overview and a detail
use-case description.
d. Verify and validate the functional models.
163163
A structural, or conceptual, model describes the structure of the objects that support
the business processes in an organization. During analysis, the structural model presents
the logical organization of the objects without indicating how they are stored, created, or
manipulated so that analysts can focus on the business, without being distracted by tech-
nical details. Later during design, the structural model is updated to refl ect exactly how
the objects will be stored in databases and fi les. Th is chapter describes class–responsibility–
collaboration (CRC) cards, class diagrams, and object diagrams, which are used to create the
structural model.
OBJECTIVES
■ Understand the rules and style guidelines for creating CRC cards, class diagrams,
and object diagrams.
■ Understand the processes used to create CRC cards, class diagrams, and object diagrams.
■ Be able to create CRC cards, class diagrams, and object diagrams.
■ Understand the relationship among the structural models.
■ Understand the relationship between the structural and functional models.
INTRODUCTION
During analysis, analysts create business process and functional models to represent how the
business system will behave. At the same time, analysts need to understand the information
that is used and created by the business system (e.g., customer information, order informa-
tion). In this chapter, we discuss how the objects underlying the behavior modeled in the
business process and functional models are organized and presented.
As pointed out in Chapter 1, all object-oriented systems development approaches are
use-case driven, architecture-centric, and iterative and incremental. Use cases, described in
Chapter 4, form the foundation on which the business information system is created. From
an architecture-centric perspective, structural modeling supports the creation of an internal
structural or static view of a business information system in that it shows how the system is
structured to support the underlying business processes. Finally, as with business process and
functional modeling, you will fi nd that you will need to not only iterate across the structural
models (described in this chapter), but you will also have to iterate across all three architec-
tural views (functional, structural, and behavioral) to fully capture and represent the require-
ments for a business information system.
A structural model is a formal way of representing the objects that are used and created by
a business system. It illustrates people, places, or things about which information is captured
and how they are related to one another. Th e structural model is drawn using an iterative
process in which the model becomes more detailed and less conceptual over time. In analysis,
analysts draw a conceptual model, which shows the logical organization of the objects without
C H A P T E R 5
Structural Modeling
164 Chapter 5 Structural Modeling
indicating how the objects are stored, created, or manipulated. Because this model is free
from any implementation or technical details, the analysts can focus more easily on matching
the model to the real business requirements of the system.
In design, analysts evolve the conceptual structural model into a design model that
refl ects how the objects will be organized in databases and soft ware. At this point, the model is
checked for redundancy, and the analysts investigate ways to make the objects easy to retrieve.
Th e specifi cs of the design model are discussed in detail in the design chapters.
STRUCTURAL MODELS
Every time a systems analyst encounters a new problem to solve, the analyst must learn the
underlying problem domain. Th e goal of the analyst is to discover the key objects contained
in the problem domain and to build a structural model. Object-oriented modeling allows the
analyst to reduce the semantic gap between the underlying problem domain and the evolving
structural model. However, the real world and the world of soft ware are very diff erent. Th e
real world tends to be messy, whereas the world of soft ware must be neat and logical. Th us,
an exact mapping between the structural model and the problem domain may not be possible.
In fact, it might not even be desirable.
One of the primary purposes of the structural model is to create a vocabulary that can be
used by the analyst and the users. Structural models represent the things, ideas, or concepts
contained in the domain of the problem. Th ey also allow the representation of the relation-
ships among the things, ideas, or concepts. By creating a structural model of the problem
domain, the analyst creates the vocabulary necessary for the analyst and users to communi-
cate eff ectively.
It is important to remember that at this stage of development, the structural model does not
represent soft ware components or classes in an object-oriented programming language, even
though the structural model does contain analysis classes, attributes, operations, and the relation-
ships among the analysis classes. Th e refi nement of these initial classes into programming-level
objects comes later. Nonetheless, the structural model at this point should represent the respon-
sibilities of each class and the collaborations among the classes. Typically, structural models are
depicted using CRC cards, class diagrams, and, in some cases, object diagrams. However, before
describing CRC cards, class diagrams, and object diagrams, we describe the basic elements of
structural models: classes, attributes, operations, and relationships.
Classes, Attributes, and Operations
A class is a general template that we use to create specifi c instances, or objects, in the problem
domain. All objects of a given class are identical in structure and behavior but contain diff er-
ent data in their attributes. Th ere are two general kinds of classes of interest during analysis:
concrete and abstract. Normally, when an analyst describes the application domain classes,
he or she is referring to concrete classes; that is, concrete classes are used to create objects.
Abstract classes do not actually exist in the real world; they are simply useful abstractions.
For example, from an employee class and a customer class, we may identify a generalization
of the two classes and name the abstract class person. We might not actually instantiate the
person class in the system itself, instead creating and using only employees and customers.1
1 Because abstract classes are essentially not necessary and are not instantiated, arguments have been made that it
would be better not to include any of them in the description of the evolving system at this stage of development (see
J. Evermann and Y. Wand, “Towards Ontologically Based Semantics for UML Constructs,” in H. S. Junii, S. Jajodia,
and A. Solvberg (eds.) ER 2001, Lecture Notes in Computer Science 2224 (Berlin: Springer-Verlag, 2001): 354–367).
However, because abstract classes traditionally have been included at this stage of development, we also include them.
Structural Models 165
A second classifi cation of classes is the type of real-world thing that a class repre-
sents. Th ere are domain classes, user-interface classes, data structure classes, fi le structure
classes, operating environment classes, document classes, and various types of multimedia
classes. At this point in the development of our evolving system, we are interested only
in domain classes. Later in design and implementation, the other types of classes become
more relevant.
An attribute of an analysis class represents a piece of information that is relevant to the
description of the class within the application domain of the problem being investigated.
An attribute contains information the analyst or user feels the system should keep track of.
For example, a possible relevant attribute of an employee class is employee name, whereas
one that might not be as relevant is hair color. Both describe something about an employee,
but hair color is probably not all that useful for most business applications. Only attributes
that are important to the task should be included in the class. Finally, only attributes that
are primitive or atomic types (i.e., integers, strings, doubles, date, time, Boolean, etc.) should
be added. Most complex or compound attributes are really placeholders for relationships
between classes. Th erefore, they should be modeled as relationships, not as attributes (see
the next section).
Th e behavior of an analysis class is defi ned in an operation or service. In later phases, the
operations are converted to methods. However, because methods are more related to imple-
mentation, at this point in the development we use the term operation to describe the actions
to which the instances of the class are capable of responding. Like attributes, only problem
domain–specifi c operations that are relevant to the problem being investigated should be
considered. For example, it is normally required that classes provide means of creating
instances, deleting instances, accessing individual attribute values, setting individual attrib-
ute values, accessing individual relationship values, and removing individual relationship
values. However, at this point in the development of the evolving system, the analyst should
avoid cluttering up the defi nition of the class with these basic types of operations and focus
only on relevant problem domain–specifi c operations.
Relationships
Th ere are many diff erent types of relationships that can be defi ned, but all can be classifi ed
into three basic categories of data abstraction mechanisms: generalization relationships,
aggregation relationships, and association relationships. Th ese data-abstraction mecha-
nisms allow the analyst to focus on the important dimensions while ignoring nonessen-
tial dimensions. As with attributes, the analyst must be careful to include only relevant
relationships.
Generalization Relationships Th e generalization abstraction enables the analyst to create
classes that inherit attributes and operations of other classes. Th e analyst creates a super-
class that contains basic attributes and operations that will be used in several subclasses.
Th e subclasses inherit the attributes and operations of their superclass and can also contain
attributes and operations that are unique just to them. For example, a customer class and
an employee class can be generalized into a person class by extracting the attributes and
operations both have in common and placing them into the new superclass, person. In this
way, the analyst can reduce the redundancy in the class defi nitions so that the common
elements are defi ned once and then reused in the subclasses. Generalization is represented
with the a-kind-of relationship, so that we say that an employee is a-kind-of person.
Th e analyst also can use the opposite of generalization. Specialization uncovers
additional classes by allowing new subclasses to be created from an existing class. For
example, an employee class can be specialized into a secretary class and an engineer class.
166 Chapter 5 Structural Modeling
Furthermore, generalization relationships between classes can be combined to form gener-
alization hierarchies. Based on the previous examples, a secretary class and an engineer class
can be subclasses of an employee class, which in turn could be a subclass of a person class.
Th is would be read as a secretary and an engineer are a-kind-of employee and a customer
and an employee are a-kind-of person.
Th e generalization data abstraction is a very powerful mechanism that encourages the
analyst to focus on the properties that make each class unique by allowing the similarities
to be factored into superclasses. However, to ensure that the semantics of the subclasses are
maintained, the analyst should apply the principle of substitutability. By this we mean that
the subclass should be capable of substituting for the superclass anywhere that uses the super-
class (e.g., anywhere we use the employee superclass, we could also logically use its secretary
subclass). By focusing on the a-kind-of interpretation of the generalization relationship, the
principle of substitutability is applied.
Aggregation Relationships Generally speaking, all aggregation relationships relate parts to
wholes or assemblies. For our purposes, we use the a-part-of or has-parts semantic relationship
to represent the aggregation abstraction. For example, a door is a-part-of a car, an employee is
a-part-of a department, or a department is a-part-of an organization. Like the generalization
relationship, aggregation relationships can be combined into aggregation hierarchies. For
example, a piston is a-part-of an engine, and an engine is a-part-of a car.
Aggregation relationships are bidirectional. Th e fl ip side of aggregation is decomposition.
Th e analyst can use decomposition to uncover parts of a class that should be modeled sepa-
rately. For example, if a door and an engine are a-part-of a car, then a car has-parts door and
engine. Th e analyst can bounce around between the various parts to uncover new parts. For
example, the analyst can ask, What other parts are there to a car? or To which other assem-
blies can a door belong?
Association Relationships Th ere are other types of relationships that do not fi t neatly into
a generalization (a-kind-of) or aggregation (a-part-of) framework. Technically speaking,
these relationships are usually a weaker form of the aggregation relationship. For example, a
patient schedules an appointment. It could be argued that a patient is a-part-of an appoint-
ment. However, there is a clear semantic diff erence between this type of relationship and one
that models the relationship between doors and cars or even workers and unions. Th us, they
are simply considered to be associations between instances of classes.
OBJECT IDENTIFICATION
Diff erent approaches have been suggested to aid the analyst in identifying a set of candidate
objects for the structural model. Th e four most common approaches are textual analysis,
brainstorming, common object lists, and patterns. Most analysts use a combination of these
techniques to make sure that no important objects and object attributes, operations, and
relationships have been overlooked.
Textual Analysis
Th e analyst performs textual analysis by reviewing the use-case diagrams and examining
the text in the use-case descriptions to identify potential objects, attributes, operations, and
relationships. Th e nouns in the use case suggest possible classes, and the verbs suggest pos-
sible operations. Figure 5-1 presents a summary of useful guidelines. Th e textual analysis of
Object Identifi cation 167
use-case descriptions has been criticized as being too simple, but because its primary pur-
pose is to create an initial rough-cut structural model, its simplicity is a major advantage.
For example, if we applied these rules to the Make Old Patient Appt use case described in
Chapter 4 and replicated in Figure 5-2, we can easily identify potential objects for an old
patient, doctor, appointment, patient, offi ce, receptionist, name, address, patient informa-
tion, payment, date, and time. We also can easily identify potential operations that can be
associated with the identifi ed objects. For example, patient contacts offi ce, makes a new
appointment, cancels an existing appointment, changes an existing appointment, matches
requested appointment times and dates with requested times and dates, and fi nds current
appointment.
Brainstorming
Brainstorming is a discovery technique that has been used successfully in identifying candi-
date classes. Essentially, in this context, brainstorming is a process that a set of individuals
sitting around a table suggest potential classes that could be useful for the problem under
consideration. Typically, a brainstorming session is kicked off by a facilitator who asks the
set of individuals to address a specifi c question or statement that frames the session. For
example, using the appointment problem described previously, the facilitator could ask the
development team and users to think about their experiences of making appointments and
to identify candidate classes based on their past experiences. Notice that this approach does
not use the functional models developed earlier. It simply asks the participants to identify
the objects with which they have interacted. For example, a potential set of objects that
come to mind are doctors, nurses, receptionists, appointment, illness, treatment, prescrip-
tions, insurance card, and medical records. Once a suffi cient number of candidate objects
have been identifi ed, the participants should discuss and select which of the candidate
objects should be considered further. Once these have been identifi ed, further brainstorm-
ing can take place to identify potential attributes, operations, and relationships for each of
the identifi ed objects.
Bellin and Simone2 have suggested a set of useful principles to guide a brainstorming
session. First, all suggestions should be taken seriously. At this point in the development
• A common or improper noun implies a class of objects.
• A proper noun or direct reference implies an instance of a class.
• A collective noun implies a class of objects made up of groups of instances of another class.
• An adjective implies an attribute of an object.
• A doing verb implies an operation.
• A being verb implies a classifi cation relationship between an object and its class.
• A having verb implies an aggregation or association relationship.
• A transitive verb implies an operation.
• An intransitive verb implies an exception.
• A predicate or descriptive verb phrase implies an operation.
• An adverb implies an attribute of a relationship or an operation.
Adapted from: These guidelines are based on Russell J. Abbott, “Program Design by Informal English Descriptions,”
Communications of the ACM 26, no. 11 (1983): 882–894; Peter P-S Chen, “English Sentence Structure and
Entity-Relationship Diagrams,” Information Sciences: An International Journal 29, no. 2–3 (1983): 127–149;
Ian Graham, Migrating to Object Technology (Reading, MA: Addison Wesley Longman, 1995).
FIGURE 5-1
Textual Analysis
Guidelines
2 D. Bellin and S. S. Simone, Th e CRC Card Book (Reading, MA: Addison-Wesley, 1997).
168 Chapter 5 Structural Modeling
FIGURE 5-2 Use-Case Description (Figure 4-13)
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Make Old Patient Appt 2 Low
Old Patient
Old Patient – wants to make, change, or cancel an appointment
Doctor – wants to ensure patient’s needs are met in a timely manner
Use Case Type: Detail, Essential
Patient calls and asks for a new appointment or asks to cancel or change an existing appointment.
Old Patient
Update Patient Information
Manage Appointments
1. The Patient contacts the office regarding an appointment.
2. The Patient provides the Receptionist with his or her name and address.
3. If the Patient’s information has changed
Execute the Update Patient Information use case.
4. If the Patient’s payment arrangements has changed
Execute the Make Payments Arrangements use case.
5. The Receptionist asks Patient if he or she would like to make a new appointment, cancel an existing appointment, or change
an existing appointment.
S-1: New Appointment
1. The Receptionist asks the Patient for possible appointment times.
2. The Receptionist matches the Patient’s desired appointment times with available dates and
times and schedules the new appointment.
S-2: Cancel Appointment
1. The Receptionist asks the Patient for the old appointment time.
2. The Receptionist finds the current appointment in the appointment file and cancels it.
S-3: Change Appointment
1. The Receptionist performs the S-2: cancel appointment subflow.
2. The Receptionist performs the S-1: new appointment subflow.
S-1, 2a1: The Receptionist proposes some alternative appointment times based on what is available in the
appointment schedule.
S-1, 2a2: The Patient chooses one of the proposed times or decides not to make an appointment.
6. The Receptionist provides the results of the transaction to the Patient.
This use case describes how we make an appointment as well as changing or canceling
an appointment for a previously seen patient.
Relationships:
Association:
Include:
Extend:
Generalization:
If the patient wants to make a new appointment,
the S-1: new appointment subflow is performed.
If the patient wants to cancel an existing appointment,
the S-2: cancel appointment subflow is performed.
If the patient wants to change an existing appointment,
the S-3: change appointment subflow is performed.
TEMPLATE
can be found at
www.wiley.com
/college/dennis
Object Identifi cation 169
of the system, it is much better to have to delete something later than to accidentally leave
something critical out. Second, all participants should begin thinking fast and furiously. Aft er
all ideas are out on the proverbial table, then the participants can be encouraged to ponder
the candidate classes they have identifi ed. Th ird, the facilitator must manage the fast and
furious thinking process. Otherwise, the process will be chaotic. Furthermore, the facilitator
should ensure that all participants are involved and that a few participants do not dominate
the process. To get the most complete view of the problem, we suggest using a round-robin
approach wherein participants take turns suggesting candidate classes. Another approach is
to use an electronic brainstorming tool that supports anonymity.3 Fourth, the facilitator can
use humor to break the ice so that all participants can feel comfortable in making suggestions.
Common Object Lists
As its name implies, a common object list is simply a list of objects common to the business
domain of the system. Several categories of objects have been found to help the analyst
in creating the list, such as physical or tangible things, incidents, roles, and interactions.4
Analysts should fi rst look for physical, or tangible, things in the business domain. Th ese could
include books, desks, chairs, and offi ce equipment. Normally, these types of objects are the
easiest to identify. Incidents are events that occur in the business domain, such as meetings,
fl ights, performances, or accidents. Reviewing the use cases can readily identify the roles that
the people play in the problem, such as doctor, nurse, patient, or receptionist. Typically, an
interaction is a transaction that takes place in the business domain, such as a sales transac-
tion. Other types of objects that can be identifi ed include places, containers, organizations,
business records, catalogs, and policies. In rare cases, processes themselves may need infor-
mation stored about them. In these cases, processes may need an object, in addition to a use
case, to represent them. Finally, there are libraries of reusable objects that have been created
for diff erent business domains. For example, with regard to the appointment problem, the
Common Open Source Medical Objects5 could be useful to investigate for potential objects
that should be included.
Patterns
Th e idea of using patterns is a relatively new area in object-oriented systems development.6
Th ere have been many defi nitions of exactly what a pattern is. From our perspective, a pat-
tern is simply a useful group of collaborating classes that provide a solution to a commonly
occurring problem. Because patterns provide a solution to commonly occurring problems,
they are reusable.
3 A.R. Dennis, J.S. Valacich, T. Connolly, and B.E. Wynne, “Process Structuring in Electronic Brainstorming,”
Information Systems Research 7, no. 2 (June 1996): 268–277.
4 For example, see C. Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis and Design
(Englewood Cliff s, NJ: Prentice Hall, 1998); S. Shlaer and S. J. Mellor, Object-Oriented Systems Analysis: Modeling the
World in Data (Englewood Cliff s, NJ: Yourdon Press, 1988).
5 See Common Open Source Medical Objects, Sourceforge, sourceforge.net/projects/cosmos/.
6 Many books have been devoted to this topic. For example, see P. Coad, D. North, and M. Mayfi eld, Object Mod-
els: Strategies, Patterns, & Applications, 2nd Ed. (Englewood Cliff s, NJ: Prentice Hall, 1997); H.-E. Eriksson and
M. Penker, Business Modeling with UML: Business Patterns at Work (New York: Wiley, 2000); M. Fowler, Analysis
Patterns: Reusable Object Models (Reading, MA: Addison-Wesley, 1997); E. Gamma, R. Helm, R. Johnson, and
J. Vlissides, Design Patterns: Elements of Reusable Object-Oriented Soft ware (Reading, MA: Addison-Wesley, 1995);
David C. Hay, Data Model Patterns: Conventions of Th ought (New York: Dorset House, 1996); L. Silverston, Th e
Data Model Resource Book: A Library of Universal Data Models for All Enterprises, Volume 1, Revised Ed. (New
York, NY; Wiley, 2001).
170 Chapter 5 Structural Modeling
FIGURE 5-3 Sample Patterns
Place
Transaction Transaction Line Item Item
Participant
0..*
0..*
1..*
1..1
1..1
1..1 1..10..*
Transaction Account
2..21..1
1..10..*
OrganizationPerson
Party
ServiceGood
Product
Entry
An architect, Christopher Alexander, has inspired much of the work associated with
using patterns in object-oriented systems development. According to Alexander and his col-
leagues,7 it is possible to make very sophisticated buildings by stringing together commonly
found patterns, rather than creating entirely new concepts and designs. In a similar man-
ner, it is possible to put together commonly found object-oriented patterns to form elegant
object-oriented information systems. For example, many business transactions involve the
same types of objects and interactions. Virtually all transactions would require a transaction
class, a transaction line item class, an item class, a location class, and a participant class. By
reusing these existing patterns of classes, we can more quickly and more completely defi ne
the system than if we start with a blank piece of paper.
Many types of patterns have been proposed, ranging from high-level business-oriented
patterns to more low-level design patterns. For example, Figure 5-3 depicts a set of useful
analysis patterns.8 Figure 5-4 portrays a class diagram that we created by merging the patterns
contained in Figure 5-3 into a single reusable pattern. In this case, we merged the Transaction–
Entry–Account pattern (located at the bottom left of Figure 5-3) with the Place–Transaction–
Participant–Transaction Line Item–Item pattern (located at the top left of Figure 5-3) on the
7 C. Alexander, S. Ishikawa, M. Silverstein, M. Jacobson, I. Fiksdahl-King, and S. Angel, A Pattern Language (New
York: Oxford University Press, 1977).
8 Th e patterns are portrayed using UML Class Diagrams. We describe the syntax of the diagrams later in this chapter.
Th e specifi c patterns shown have been adapted from patterns described in P. Coad, D. North, and M. Mayfi eld,
Object Models: Strategies, Patterns, & Applications, 2nd Ed.; M. Fowler, Analysis Patterns: Reusable Object Models;
L. Silverston, Th e Data Model Resource Book: A Library of Universal Data Models for All Enterprises, Volume 1,
Revised Edition.
Object Identifi cation 171
FIGURE 5-4 Sample Integration of Sample Patterns
Place
Transaction Transaction Line Item Item
Participant
0..*
0..*
1..*
1..1
1..1
1..1 1..10..*
1..12..2
OrganizationPerson
ServiceGood
Account
0..*1..1
Entry
common Transaction class. Next, we merged the Party–Person–Organization (located at the
top right of Figure 5-3) by merging the Participant and Party classes. Finally, we extended
the Item class by merging the Item class with the Product class of the Product–Good–Service
pattern (located at the bottom right of Figure 5-3).
In this manner, using patterns from diff erent sources in enables the development team
to leverage knowledge beyond that of the immediate team members and allows the team to
develop more complete and robust models of the problem domain. For example, in the case
of the appointment problem, we can look at the objects previously identifi ed through textual
analysis, brainstorming, and/or common object lists and see if it makes sense to map any of
them into any predefi ned reusable patterns. In this specifi c case, we can look at an appointment
as a type of transaction in which a doctor’s offi ce participates. By looking at an appointment as a
type of transaction, we can apply the pattern we created in Figure 5-4 and discover a set of pre-
viously unidentifi ed objects, such as Place, Patient as a type of Participant, and Transaction Line
Items that are associated with diff erent types of Items (Goods and/or Services). Discovering
these specifi c additional objects could be useful in developing the billing side of the appoint-
ment system. Even though these additional objects could be applicable, they were not uncov-
ered using the other techniques.
Based on this simple example, it is obvious that using patterns to develop structural
models can be advantageous. Figure 5-5 lists some common business domains for which
patterns have been developed and their source. If we are developing a business information
system in one of these business domains, then the patterns developed for that domain may
be a very useful starting point in identifying needed classes and their attributes, operations,
and relationships.
172 Chapter 5 Structural Modeling
CRC CARDS
CRC (Class–Responsibility–Collaboration) cards are used to document the responsibilities and
collaborations of a class. In some object-oriented systems-development methodologies, CRC
cards are seen to be an alternative competitor to the Unifi ed Process employment of use cases
and class diagrams. However, we see them as a useful, low-tech approach that can compli-
ment a typical high-tech Unifi ed Process approach that uses CASE tools. We use an extended
form of the CRC card to capture all relevant information associated with a class.9 We describe
the elements of our CRC cards later, aft er we explain responsibilities and collaborations.
Responsibilities and Collaborations
Responsibilities of a class can be broken into two separate types: knowing and doing.
Knowing responsibilities are those things that an instance of a class must be capable of know-
ing. An instance of a class typically knows the values of its attributes and its relationships.
Doing responsibilities are those things that an instance of a class must be capable of doing. In
this case, an instance of a class can execute its operations or it can request a second instance,
which it knows about, to execute one of its operations on behalf of the fi rst instance.
Th e structural model describes the objects necessary to support the business processes
modeled by the use cases. Most use cases involve a set of several classes, not just one class.
9 Our CRC cards are based on the work of D. Bellin and S. S. Simone, Th e CRC Card Book (Reading, MA: Addison-
Wesley, 1997); I. Graham, Migrating to Object Technology (Wokingham, England: Addison-Wesley, 1995); B.
Henderson-Sellers and B. Unhelkar, OPEN modeling with UML (Harlow, England: Addison-Wesley, 2000).
Business Domains Sources of Patterns
Accounting 3, 4
Actor-Role 2
Assembly-Part 1
Container-Content 1
Contract 2, 4
Document 2, 4
Employment 2, 4
Financial Derivative Contracts 3
Geographic Location 2, 4
Group-Member 1
Interaction 1
Material Requirements Planning 4
Organization and Party 2, 3
Plan 1, 3
Process Manufacturing 4
Trading 3
Transactions 1, 4
1. Peter Coad, David North, and Mark Mayfi eld, Object Models: Strategies, Patterns, and Applications,
2nd Ed. (Englewood Cliffs, NJ: Prentice Hall, 1997).
2. Hans-Erik Eriksson and Magnus Penker, Business Modeling with UML: Business Patterns at Work
(New York: Wiley, 2000).
3. Martin Fowler, Analysis Patterns: Reusable Object Models (Reading, MA: Addison-Wesley, 1997).
4. David C. Hay, Data Model Patterns: Conventions of Thought (New York, NY, Dorset House, 1996).
FIGURE 5-5
Useful Patterns
CRC Cards 173
Th ese classes form collaborations. Collaborations allow the analyst to think in terms of clients,
servers, and contracts.10 A client object is an instance of a class that sends a request to an
instance of another class for an operation to be executed. A server object is the instance that
receives the request from the client object. A contract formalizes the interactions between the
client and server objects. Chapter 8 provides a more-detailed explanation of contracts and
examples of their use.
An analyst can use the idea of class responsibilities and client–server–contract collabo-
rations to help identify the classes, along with the attributes, operations, and relationships,
involved with a use case. One of the easiest ways to use CRC cards in developing a structural
model is through anthropomorphism—pretending that the classes have human characteris-
tics. Members of the development team can either ask questions of themselves or be asked
questions by other members of the team. Typically the questions asked are of the form:
Who or what are you?
What do you know?
What can you do?
Th e answers to the questions are then used to add detail to the evolving CRC cards. For
example, in the appointment problem, a member of the team can pretend that he or she is an
appointment. In this case, the appointment would answer that he or she knows about the doc-
tor and patient who participate in the appointment and they would know the date and time
of the appointment. Furthermore, an appointment would have to know how to create itself,
delete itself, and to possibly change diff erent aspects of itself. In some cases, this approach will
uncover additional objects that have to be added to the evolving structural model.
Elements of a CRC Card
Th e set of CRC cards contains all the information necessary to build a logical structural model
of the problem under investigation. Figure 5-6 shows a sample CRC card. Each CRC card cap-
tures and describes the essential elements of a class. Th e front of the card contains the class’s
name, ID, type, description, associated use cases, responsibilities, and collaborators. Th e name
of a class should be a noun (but not a proper noun, such as the name of a specifi c person or
thing). Just like the use cases, in later stages of development, it is important to be able to trace
back design decisions to specifi c requirements. In conjunction with the list of associated use
cases, the ID number for each class can be used to accomplish this. Th e description is simply
a brief statement that can be used as a textual defi nition for the class. Th e responsibilities of
the class tend to be the operations that the class must contain (i.e., the doing responsibilities).
Th e back of a CRC card contains the attributes and relationships of the class. Th e attributes
of the class represent the knowing responsibilities that each instance of the class has to meet.
Typically, the data type of each attribute is listed with the name of the attribute (e.g., the amount
attribute is double and the insurance carrier is text). Th ree types of relationships typically are
captured at this point: generalization, aggregation, and other associations. In Figure 5-6, we see
that a Patient is a-kind-of Person and that a Patient is associated with Appointments.
CRC cards are used to document the essential properties of a class. However, once the
cards are fi lled out, the analyst can use the cards and anthropomorphisms in role-playing
(described in the next section) to uncover missing properties by executing the diff erent
10 For more information, see K. Beck and W. Cunningham, “A Laboratory for Teaching Object-Oriented Th ink-
ing,” Proceedings of OOPSLA, SIGPLAN Notices, 24, no. 10 (1989): 1–6; B. Henderson-Sellers and B. Unhelkar,
OPEN Modeling with UML (Harlow, England: Addison-Wesley, 2000); C. Larman, Applying UML and Patterns: An
Introduction to Object-Oriented Analysis and Design (Englewood Cliff s, NJ: Prentice Hall, 1998); B. Meyer, Object-
Oriented Soft ware Construction (Englewood Cliff s, NJ: Prentice Hall, 1994); R. Wirfs-Brock, B. Wilkerson, and
L. Wiener, Designing Object-Oriented Soft ware (Englewood Cliff s, NJ, Prentice Hall, 1990).
174 Chapter 5 Structural Modeling
FIGURE 5-6
Sample CRC Card
Front:
Class Name: Old Patient ID: 3
Calculate last visit
Make appointment
Change status
Provide medical history
Responsibilities
Associated Use Cases: 2Description: An individual who needs to receive or has received
medical attention
Type: Concrete, Domain
Appointment
Medical history
Collaborators
Back:
Attributes:
Insurance carrier (text)
Amount (double)
Relationships:
Generalization (a-kind-of): Person
Aggregation (has-parts): Medical History
Other Associations: Appointment
scenarios associated with the use cases (see Chapter 4). Role-playing also can be used as a
basis to test the clarity and completeness of the evolving representation of the system.
Role-Playing CRC Cards with Use Cases11, 12
In addition to the object identifi cation approaches described earlier (textual analysis, brain-
storming, common object lists, and patterns), CRC cards can be used in a role-playing exercise
that has been shown to be useful in discovering additional objects, attributes, relationships,
and operations. Furthermore, in addition to walkthroughs, described later in this chapter,
role-playing is very useful in testing the fi delity of the evolving structural model. In general,
members of the team perform roles associated with the actors and objects previously identifi ed
11 Th is step is related to the verifi cation and validation of the analysis models (functional, structural, and behavioral).
Because this deals with verifi cation and validation that take place between the models, in this case functional and
structural, we will return to this topic in Chapter 7.
12 Our role-playing approach is based on the work of D. Bellin and S. S. Simone, Th e CRC Card Book (Reading, MA:
Addison-Wesley, 1997).
CRC Cards 175
with the diff erent use cases. Technically speaking, the members of the team perform the dif-
ferent steps associated with a specifi c scenario of a use case. Remember, a scenario is a single,
unique execution path through a use case. A useful place to look for the diff erent scenarios of a
use case is the activity diagrams (e.g., see Figures 4-8, 4-9, 4-10, and 4-12). A diff erent scenario
exists for each time a decision node causes a split in the execution path of the use case. Also,
scenarios can be identifi ed from the alternative/exceptional fl ows in a use-case description.
Considering the incremental and iterative nature and that activity diagrams and use-case
descriptions should contain the same information, reviewing both representations will ensure
that relevant scenarios are not missed.
Th e fi rst step is to review the use-case descriptions (see Figure 5-2). Th is allows the team to
pick a specifi c use case to role-play. Even though it is tempting to try to complete as many use
cases as possible in a short time, the team should not choose the easiest use cases fi rst. Instead,
at this point in the development of the system, the team should choose the use case that is the
most important, the most complex, or the least understood.
Th e second step is to identify the relevant roles that are to be played. Each role is associated
with either an actor or an object. To choose the relevant objects, the team reviews each of the
CRC cards and picks the ones that are associated with the chosen use case. For example, in
Figure 5-6, we see that the CRC card that represents the Old Patient class is associated with
Use Case number 2. So if we were going to role-play the Make Old Patient Appt use case (see
Figure 5-2), we would need to include the Old Patient CRC card. By reviewing the use-case
description, we can easily identify the Old Patient and Doctor actors (see Primary Actor and
Stakeholders section of the use case description in Figure 5-2). By reading the event section of
the use-case description, we identify the internal actor role of Receptionist. Aft er identifying
all of the relevant roles, we assign each one to a diff erent member of the team.
Th e third step is to role-play scenarios of the use case by having the team members per-
form each one. To do this, each team member must pretend that he or she is an instance
of the role assigned to him or her. For example, if a team member was assigned the role of
the Receptionist, then he or she would have to be able to perform the diff erent steps in the
scenario associated with the Receptionist. In the case of the change appointment scenario,
this would include steps 2, 5, 6, S-3, S-1, and S-2. However, when this scenario is performed
(role-played), it would be discovered that steps 1, 3, and 4 were incomplete. For example,
in Step 1, what actually occurs? Does the Patient make a phone call? If so, who answers
the phone? In other words, a lot of information contained in the use-case description is
only identifi ed in an implicit, not explicit, manner. When the information is not identifi ed
explicitly, there is a lot of room for interpretation, which requires the team members to make
assumptions. It is much better to remove the need to make an assumption by making each
step explicit. In this case, Step 1 of the Normal Flow of Events should be modifi ed. Once the
step has been fi xed, the scenario is tried again. Th is process is repeated until the scenario can
be executed to a successful conclusion. Once the scenario has successfully concluded, the
next scenario is performed. Th is is repeated until all of the scenarios of the use case can be
performed successfully. 13
Th e fourth step is to simply repeat steps 1 through 3 for the remaining use cases.
13 In some cases, some scenarios are only executed in very rare circumstances. So, from a practical perspective, each
scenario could be prioritized individually and only “important” scenarios would have to be implemented for the fi rst
release of the system. Only those scenarios would have to be tested at this point in the evolution of the system.
2. Identify Relevant
Actors and Objects
3. Role-Play
Scenarios
1. Review Use Cases
4. Repeat Steps
1 through 3
176 Chapter 5 Structural Modeling
CLASS DIAGRAMS
A class diagram is a static model that shows the classes and the relationships among classes
that remain constant in the system over time. Th e class diagram depicts classes, which include
both behaviors and states, with the relationships between the classes. Th e following sections
present the elements of the class diagram, diff erent approaches that can be used to simplify a
class diagram, and an alternative structure diagram: the object diagram.
Elements of a Class Diagram
Figure 5-7 shows a class diagram that was created to refl ect the classes and relationships
associated with the appointment system. Th is diagram is based on the classes uncov-
ered through the object identifi cation techniques and the role-playing of the CRC cards
described earlier.
Class Th e main building block of a class diagram is the class, which stores and manages
information in the system (see Figure 5-8). During analysis, classes refer to the people,
places, and things about which the system will capture information. Later, during design
and implementation, classes can refer to implementation-specifi c artifacts such as win-
dows, forms, and other objects used to build the system. Each class is drawn using a three-
part rectangle, with the class’s name at the top, attributes in the middle, and operations
at the bottom. We can see that the classes identifi ed earlier, such as Participant, Doctor,
Patient, Receptionist, Medical History, Appointment, and Symptom, are included in Figure
5-7. Th e attributes of a class and their values defi ne the state of each object created from the
class, and the behavior is represented by the operations.
Attributes are properties of the class about which we want to capture information
(see Figure 5-8). Notice that the Participant class in Figure 5-7 contains the attributes:
lastname, fi rstname, address, phone, and birthdate. At times, you might want to store
derived attributes, which are attributes that can be calculated or derived; these special
attributes are denoted by placing a slash (/) before the attribute’s name. Notice how the
person class contains a derived attribute called /age, which can be derived by subtracting
the patient’s birth date from the current date. It is also possible to show the visibility
of the attribute on the diagram. Visibility relates to the level of information hiding to be
enforced for the attribute. Th e visibility of an attribute can be public (+), protected (#),
or private (−). A public attribute is one that is not hidden from any other object. As such,
other objects can modify its value. A protected attribute is one that is hidden from all other
classes except its immediate subclasses. A private attribute is one that is hidden from all
other classes. Th e default visibility for an attribute is normally private.
Operations are actions or functions that a class can perform (see Figure 5-8). Th e func-
tions that are available to all classes (e.g., create a new instance, return a value for a particular
attribute, set a value for a particular attribute, delete an instance) are not explicitly shown
within the class rectangle. Instead, only operations unique to the class are included, such
as the cancel without notice operation in the Appointment class and the calculate last visit
operation in the Patient class in Figure 5-7. Notice that both the operations are followed by
parentheses, which contain the parameter(s) needed by the operation. If an operation has no
parameters, the parentheses are still shown but are empty. As with attributes, the visibility
of an operation can be designated public, protected, or private. Th e default visibility for an
operation is normally public.
Th ere are four kinds of operations that a class can contain: constructor, query, update,
and destructor. A constructor operation creates a new instance of a class. For example, the
patient class may have a method called insert (), which creates a new patient instance as
177
FI
G
U
R
E
5-
7
S
am
p
le
C
la
ss
D
ia
gr
am
A
cc
ou
nt
Em
pl
oy
ee0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
1.
.*
1.
.*
1.
.*
1.
.*
1.
.1
1.
.1
1.
.*
1.
.1
1.
.1
0.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
D
eb
it
C
re
di
t
En
tr
y
A
pp
oi
nt
m
en
t
-t
im
e
-d
at
e
-r
ea
so
n
+
ca
nc
el
w
ith
ou
t n
ot
ic
e(
)
Pa
ti
en
t
-i
ns
ur
an
ce
c
ar
ri
er
+
m
ak
e
ap
po
in
tm
en
t()
+
ca
lc
ul
at
e
la
st
v
is
it(
)
+
ch
an
ge
s
ta
tu
s(
)
+
pr
ov
id
es
m
ed
ic
al
h
is
to
ry
()
Pa
rt
ic
ip
an
t
-l
as
tn
am
e
-fi
rs
tn
am
e
-a
dd
re
ss
-p
ho
ne
-b
ir
th
da
te
-/
ag
e
It
em
Tr
an
sa
ct
io
n
Li
ne
I
te
m
ha
s
ha
s
co
nt
ai
ns
A
ss
ig
ne
dT
o
+
p
ri
m
ar
y
in
su
ra
nc
e
ca
rr
ie
r
D
oc
to
r
R
ec
ep
ti
on
is
t
N
ur
se
M
ed
ic
al
H
is
to
ry
-h
ea
rt
d
is
ea
se
-h
ig
h
bl
oo
d
pr
es
su
re
-d
ia
be
te
s
-a
lle
rg
ie
s
pr
ov
id
es
su
ffe
rs
schedules
lo
ca
te
dA
t
G
oo
d
Se
rv
ic
e
Pr
es
cr
ip
ti
on
B
ra
ce
Ph
ys
ic
al
C
he
ck
up
Sy
m
pt
om
Il
ln
es
s
Pl
ac
e
Tr
ea
tm
en
t
m
ed
ic
at
io
n
in
st
ru
ct
io
ns
sy
m
pt
om
s
ev
er
ity
-a
m
ou
nt
-n
am
e
-d
es
cr
ip
tio
n
178 Chapter 5 Structural Modeling
FIGURE 5-8
Class Diagram Syntax
A class:
• Has a name typed in bold and centered in its top
compartment.
• Has a list of attributes in its middle compartment.
• Represents a kind of person, place, or thing about
which the system will need to capture and store
information.
• Has a list of operations in its bottom compartment.
attribute name
/derived attribute name
operation name ()
• Does not explicitly show operations that are
available to all classes.
An attribute:
• Represents properties that describe the state of an
object.
• Can be derived from other attributes, shown by
placing a slash before the attribute’s name.
An operation:
• Represents the actions or functions that a class
can perform.
• Can be classified as a constructor, query, or
update operation.
• Includes parentheses that may contain parameters
or information needed to perform the operation.
An association:
• Represents a relationship between multiple
classes or a class and itself.
A generalization:
• Is labeled using a verb phrase or a role name,
whichever better represents the relationship.
An aggregation:
• Represents a logical a-part-of relationship
between multiple classes or a class and itself.
• Is a special form of an association.
A composition:
• Represents a physical a-part-of relationship
between multiple classes or a class and itself
• Is a special form of an association.
• Can exist between one or more classes.
• Represents a-kind-of relationship between
multiple classes.
• Contains multiplicity symbols, which represent
the minimum and maximum times a class
instance can be associated with the related class
instance.
Class1
-Attribute-1
+Operation-1()
AssociatedWith
0..* 1
0..* 1IsPartOf
1..* 1IsPartOf
patients are entered into the system. As we just mentioned, if an operation implements one
of the basic functions (e.g., create a new instance), it is normally not explicitly shown on the
class diagram, so typically we do not see constructor methods explicitly on the class diagram.
A query operation makes information about the state of an object available to other
objects, but it does not alter the object in any way. For instance, the calculate last visit ()
operation that determines when a patient last visited the doctor’s offi ce will result in the
object’s being accessed by the system, but it will not make any change to its information. If a
query method merely asks for information from attributes in the class (e.g., a patient’s name,
address, phone), then it is not shown on the diagram because we assume that all objects have
operations that produce the values of their attributes.
Class Diagrams 179
An update operation changes the value of some or all the object’s attributes, which may
result in a change in the object’s state. Consider changing the status of a patient from new
to current with a method called change status() or associating a patient with a particular
appointment with make appointment (appointment). If the result of the operation can
change the state of the object, then the operation must be explicitly included on the class
diagram. On the other hand, if the update operation is a simple assignment operation, it can
be omitted from the diagram.
A destructor operation simply deletes or removes the object from the system. For exam-
ple, if an employee object no longer represents an actual employee associated with the fi rm,
the employee could need to be removed from the employee database, and a destructor opera-
tion would be used to implement this behavior. However, deleting an object is one of the basic
functions and therefore would not be included on the class diagram.
Relationships A primary purpose of a class diagram is to show the relationships, or asso-
ciations, that classes have with one another. Th ese are depicted on the diagram by drawing
lines between classes (see Figure 5-8). When multiple classes share a relationship (or a
class shares a relationship with itself), a line is drawn and labeled with either the name of
the relationship or the roles that the classes play in the relationship. For example, in Figure
5-7 the two classes patient and appointment are associated with one another whenever
a patient schedules an appointment. Th us, a line labeled schedules connects patient and
appointment, representing exactly how the two classes are related to each other. Also, notice
that there is a small solid triangle beside the name of the relationship. Th e triangle allows
a direction to be associated with the name of the relationship. In Figure 5-7, the schedules
relationship includes a triangle, indicating that the relationship is to be read as “patient
schedules appointment.” Inclusion of the triangle simply increases the readability of the
diagram. In Figure 5-9, three additional examples of associations are portrayed: An Invoice
is AssociatedWith a Purchase Order (and vice versa), a Pilot Flies an Aircraft , and a Spare
Tire IsLocatedIn a Trunk.
Sometimes a class is related to itself, as in the case of a patient being the primary insur-
ance carrier for other patients (e.g., spouse, children). In Figure 5-7, notice that a line was
drawn between the patient class and itself and called primary insurance carrier to depict
the role that the class plays in the relationship. Notice that a plus (+) sign is placed before
the label to communicate that it is a role as opposed to the name of the relationship. When
labeling an association, we use either a relationship name or a role name (not both), which-
ever communicates a more thorough understanding of the model.
FIGURE 5-9
Sample Association
Purchase Order
Aircraft
TrunkSpare Tire
Pilot
Invoice
0..* 1
0..* 0..*
Flies
0..1 0..1
IsLocatedIn
AssociatedWith
180 Chapter 5 Structural Modeling
Relationships also have multiplicity, which documents how an instance of an object can
be associated with other instances. Numbers are placed on the association path to denote the
minimum and maximum instances that can be related through the association in the format
minimum number.. maximum number (see Figure 5-10). Th e numbers specify the relation-
ship from the class at the far end of the relationship line to the end with the number. For exam-
ple, in Figure 5-7, there is a 0..* on the appointment end of the patient schedules appointment
relationship. Th is means that a patient can be associated with zero through many diff erent
appointments. At the patient end of this same relationship, there is a 1..1, meaning that an
appointment must be associated with one and only one patient. In Figure 5-9, we see that
an instance of the Invoice class must be AssociatedWith one instance of the Purchase Order
class and that an instance of the Purchase Order class may be AssociatedWith zero or more
instances of the Invoice class, that an instance of the Pilot class Flies zero or more instances
of the Aircraft class, and that an instance of the Aircraft class may be fl own by zero or more
instances of the Pilot class. Finally, we see that an instance the Spare Tire class IsLocatedIn
zero or one instance of the Trunk class, whereas an instance of the Trunk class can contain
zero or one instance of the Spare Tire class.
Th ere are times when a relationship itself has associated properties, especially when its
classes share a many-to-many relationship. In these cases, a class called an association class
is formed, which has its own attributes and operations.14 It is shown as a rectangle attached
1
0..*
1..*
0..1
2..4
1..3,5
Exactly one
A department has
one and only one
boss.
Zero or more
One or more
Zero or one
Specified range
Multiple, disjoint
ranges
An employee has
zero to many
children.
A boss is responsible
for one or more
employees.
An employee can
be married to zero
or one spouse.
An employee can
take from two to
four vacations each
year.
An employee is a
member of one to
three or five
committees.
Department Boss
1
Employee Child
0..*
Boss Employee
1..*
Employee Spouse
0..1
Employee Vacation
2..4
Employee Committee
1..3,5
FIGURE 5-10
Multiplicity
14 For those familiar with data modeling, associative classes serve a purpose similar to the one the associative entity
serves in ER diagramming.
Class Diagrams 181
by a dashed line to the association path, and the rectangle’s name matches the label of the
association. Th ink about the case of capturing information about illnesses and symptoms.
An illness (e.g., the fl u) can be associated with many symptoms (e.g., sore throat, fever),
and a symptom (e.g., sore throat) can be associated with many illnesses (e.g., the fl u, strep
throat, the common cold). Figure 5-7 shows how an association class can capture informa-
tion about remedies that change depending on the various combinations. For example, a
sore throat caused by strep throat requires antibiotics, whereas treatment for a sore throat
from the fl u or a cold could be throat lozenges or hot tea. Another way to decide when to
use an association class is when attributes that belong to the intersection of the two classes
involved in the association must be captured. We can visually think about an association
class as a Venn diagram. For example, in Figure 5-11, the Grade idea is really an intersection
of the Student and Course classes, because a grade exists only at the intersection of these
two ideas. Another example shown in Figure 5-11 is that a job may be viewed as the inter-
section between a Person and a Company. Most oft en, classes are related through a normal
association; however, there are two special cases of an association that you will see appear
quite oft en: generalization and aggregation.
Generalization and Aggregation Associations A generalization association shows that
one class (subclass) inherits from another class (superclass), meaning that the properties
and operations of the superclass are also valid for objects of the subclass. Th e generalization
path is shown with a solid line from the subclass to the superclass and a hollow arrow point-
ing at the superclass (see Figure 5-8). For example, Figure 5-7 communicates that doctors,
nurses, and receptionists are all kinds of employees and those employees and patients are
kinds of participants. Remember that the generalization relationship occurs when you need
to use words like “is a kind of” to describe the relationship. Some additional examples of
generalization are given in Figure 5-12. For example, Cardinal is a-kind-of Bird, which is
a-kind-of Animal; a General Practitioner is a-kind-of Physician, which is a-kind-of Person;
and a Truck is a-kind-of Land Vehicle, which is a-kind-of Vehicle.
An aggregation association is used when classes actually comprise other classes. For
example, think about a doctor’s offi ce that has decided to create health care teams that
include doctors, nurses, and administrative personnel. As patients enter the offi ce, they are
assigned to a health care team, which cares for their needs during their visits. We could
FIGURE 5-11 Sample Association Classes
CourseStudent
0..* 0..*
Grade
Student CourseGrade
CompanyPerson
0..* 0..*
Job
Person CompanyJob
182 Chapter 5 Structural Modeling
include this new knowledge in Figure 5-7 by adding two new classes (Administrative
Personnel and Health Team) and aggregation relationships from the Doctor, the Nurse,
and the new Administrative Personnel classes to the new Health Team class. A diamond
is placed nearest the class representing the aggregation (health care team), and lines are
drawn from the diamond to connect the classes that serve as its parts (doctors, nurses, and
administrative personnel). Typically, you can identify these kinds of associations when you
need to use words like “is a part of” or “is made up of” to describe the relationship. However,
from a UML perspective, there are two types of aggregation associations: aggregation and
composition (see Figure 5-8).
Aggregation is used to portray logical a-part-of relationships and is depicted on a UML
class diagram by a hollow or white diamond. For example in Figure 5-13, three logical
FIGURE 5-12 Sample Generalizations
TroutCardinal
FishBird
Animal
CarTruck
Land
HelicopterPlane
Air
SubmarineShip
Sea
PatientPhysician
SpecialistGeneral Practitioner
Person
Vehicle
Class Diagrams 183
aggregations are shown. Logical implies that it is possible for a part to be associated with
multiple wholes or that is relatively simple for the part to be removed from the whole. For
example, an instance of the Employee class IsPartOf an instance of at least one instance of the
Department class, an instance of the Wheel class IsPartOf an instance of the Vehicle class,
and an instance of the Desk class IsPartOf an instance of the Offi ce class. Obviously, in many
cases an employee can be associated with more than one department, and it is relatively easy
to remove a wheel from a vehicle or move a desk from an offi ce.
Composition is used to portray a physical part of relationships and is shown by a black
diamond. Physical implies that the part can be associated with only a single whole. For exam-
ple in Figure 5-14, three physical compositions are illustrated: an instance of a door can be a
part of only a single instance of a car, an instance of a room can be a part of an instance only of
a single building, and an instance of a button can be a part of only a single mouse. However, in
many cases, the distinction that you can achieve by including aggregation (white diamonds)
and composition (black diamonds) in a class diagram might not be worth the price of add-
ing additional graphical notation for the client to learn. Th erefore, many UML experts view
the inclusion of aggregation and composition notation to the UML class diagram as simply
“syntactic sugar” and not necessary because the same information can always be portrayed by
simply using the association syntax.
FIGURE 5-13
Sample Aggregation
Associations
Department
Vehicle
OfficeDesk
Wheel
Employee
1..* 1..*
1..* 1
IsPartOf
IsPartOf
IsPartOf0..* 1
FIGURE 5-14
Sample Composition
Associations
Car
Building
MouseButton
Room
Door
1..* 1
1..* 1
IsPartOf
IsPartOf
IsPartOf1..* 1
184 Chapter 5 Structural Modeling
15 See footnote 1.
16 For those familiar with structured analysis and design, packages serve a purpose similar to the leveling and bal-
ancing processes used in data fl ow diagramming. Packages and package diagrams are described in more detail in
Chapter 7.
Simplifying Class Diagrams
When a class diagram is fully populated with all the classes and relationships for a real-
world system, the class diagram can become very diffi cult to interpret (i.e., can be very
complex). When this occurs, it is sometimes necessary to simplify the diagram. One way
to simplify the class diagram is to show only concrete classes.15 However, depending on the
number of associations that are connected to abstract classes—and thus inherited down to
the concrete classes—this particular suggestion could make the diagram more diffi cult to
comprehend.
A second way to simplify the class diagram is through the use of a view mechanism.
Views were developed originally with relational database management systems to show only
a subset of the information contained in the database. In this case, the view would be a useful
subset of the class diagram, such as a use-case view that shows only the classes and relation-
ships relevant to a particular use case. A second view could be to show only a particular type
of relationship: aggregation, association, or generalization. A third type of view is to restrict
the information shown with each class, for example, show only the name of the class, the
name and attributes, or the name and operations. Th ese view mechanisms can be combined
to further simplify the diagram.
A third approach to simplifying a class diagram is through the use of packages (i.e.,
logical groups of classes). To make the diagrams easier to read and keep the models at a
reasonable level of complexity, the classes can be grouped together into packages. Packages
are general constructs that can be applied to any of the elements in UML models. In
Chapter 4, we introduced the package idea to simplify use-case diagrams. In the case of
class diagrams, it is simple to sort the classes into groups based on the relationships that
they share.16
Object Diagrams
Although class diagrams are necessary to document the structure of the classes, a second
type of static structure diagram, called an object diagram, can be useful in revealing addi-
tional information. An object diagram is essentially an instantiation of all or part of a class
diagram. Instantiation means to create an instance of the class with a set of appropriate
attribute values.
Object diagrams can be very useful when trying to uncover details of a class. Generally
speaking, it is easier to think in terms of concrete objects (instances) rather than abstrac-
tions of objects (classes). For example in Figure 5-15, a portion of the class diagram in
Figure 5-7 has been copied and instantiated. Th e top part of the fi gure simply is a copy
of a small view of the overall class diagram. Th e lower portion is the object diagram that
instantiates that subset of classes. By reviewing the actual instances involved, John Doe,
Appt1, Symptom1, and Dr. Smith, we may discover additional relevant attributes, relation-
ships, and/or operations or possibly misplaced attributes, relationships, and/or operations.
For example, an appointment has a reason attribute. Upon closer examination, the rea-
son attribute might have been better modeled as an association with the Symptom class.
Currently, the Symptom class is associated with the Patient class. Aft er reviewing the object
diagram, this seems to be in error. Th erefore, we should modify the class diagram to refl ect
this new understanding of the problem.
Creating Structural Models Using CRC Cards And Class Diagrams 185
FIGURE 5-15 Sample Object Diagram
Patient
-amount
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provide medical history()
0..*
0..*
0..*
0..*
1..*
1..*
1..1
1..1
Appointment
-time
-date
-reason
+cancel without notice()
+primary
insurance
carrier
Doctor
Symptom
-name
suffers
schedules
assignedTo
Participant
-lastname
-firstname
-address
-phone
-birthdate
-/age
Symptom1: Symptom
name = Muscle Pain
John Doe: Patient
lastname = Doe
firstname = John
address = 1000 Main St
phone = 555-555-5555
birthdate = 01/01/72
/ age = 40
amount = 0.00
insurance carrier = JD Health Ins
time = 3:00
date = 7/7/2012
reason = Pain in Neck
Appt1: Appointment
lastname = Smith
firstname = Jane
address = Doctor’s Clinic
phone = 999-999-9999
birthdate : 12/12/64
/ age = 48
Dr. Smith: Doctor
CREATING STRUCTURAL MODELS USING CRC CARDS
AND CLASS DIAGRAMS
Creating a structural model is an incremental and iterative process whereby the analyst
makes a rough cut of the model and then refi nes it over time. Structural models can become
quite complex—in fact, there are systems that have hundreds of classes. It is important to
remember that CRC cards and class diagrams can be used to describe both the as-is and
to-be structural models of the evolving system, but they are most oft en used for the to-be
model. Th ere are many diff erent ways to identify a set of candidate objects and to create
CRC cards and class diagrams. Today most object identifi cation begins with the use cases
186 Chapter 5 Structural Modeling
identifi ed for the problem (see Chapter 4). In this section, we describe a use-case–driven
process that can be used to create the structural model of a problem domain.
We could begin creating the structural model with a class diagram instead of CRC cards.
However, owing to the low-tech nature and the ease of role-playing use-case scenarios
with CRC cards, we prefer to create the CRC cards fi rst and then transfer the informa-
tion from the CRC cards into a class diagram later. As a result, the fi rst step of our rec-
ommended process is to create CRC cards. Performing textual analysis on the use-case
descriptions does this. If you recall, the normal fl ow of events, subfl ows, and alternative/
exceptional fl ows of the use-case description were written in a special form called Subject–
Verb–Direct-Object–Preposition–Indirect object (SVDPI). By writing the use-case events
in this form, it is easier to use the guidelines for textual analysis in Figure 5-1 to identify
the objects. Reviewing the primary actors, stakeholders and interests, and brief descriptions
of each use case allows additional candidate objects to be identifi ed. It is useful to go back
and review the original requirements to look for information that was not included in the
text of the use cases. Record all the uncovered information for each candidate object on a
CRC card.
Th e second step is to review the CRC cards to determine if additional candidate objects,
attributes, operations, and relationships are missing. In conjunction with this review, using
the brainstorming and common object list approaches described earlier can aid the team in
identifying missing classes, attributes, operations, and relationships. For example, the team
could start a brainstorming session with a set of questions such as:
■ What are the tangible things associated with the problem?
■ What are the roles played by the people in the problem domain?
■ What incidents and interactions take place in the problem domain?
As you can readily see, by beginning with the use-case descriptions, many of these questions
already have partial answers. For example, the primary actors and stakeholders are the roles that
are played by the people in the problem domain. However, it is possible to uncover additional
roles not thought of previously. Th is obviously would cause the use-case descriptions, and possi-
bly the use-case diagram, to be modifi ed and possibly expanded. As in the previous step, be sure
to record all the uncovered information onto the CRC cards. Th is includes any modifi cations
uncovered for any previously identifi ed candidate objects and any information regarding any
new candidate objects identifi ed.
Th e third step is to role-play each use-case scenario using the CRC cards. Each CRC card
should be assigned to an individual who will perform the operations for the class on the CRC
card. As the performers act out their roles, the system tends to break down. When this occurs,
additional objects, attributes, operations, or relationships will be identifi ed. Again, as in the
previous steps, any time any new information is discovered, new CRC cards are created or
modifi cations to existing CRC cards are made.
Th e fourth step is to create the class diagram based on the CRC cards. Information con-
tained on the CRC cards is transferred to the class diagrams. Th e responsibilities are trans-
ferred as operations; the attributes are drawn as attributes; and the relationships are drawn
as generalization, aggregation, or association relationships. However, the class diagram
also requires that the visibility of the attributes and operations be known. As a general
rule, attributes are private and operations are public. Th erefore, unless the analyst has a
4. Create Class
Diagram
2. Review CRC Cards
3. Role-Play the CRC
Cards
1. Create CRC Cards
Creating Structural Models Using CRC Cards And Class Diagrams 187
5. Review Class
Diagram
6. Incorporate
Patterns
7. Review the
Model
good reason to change the default visibility of these properties, then the defaults should
be accepted. Finally, the analyst should examine the model for additional opportunities to
use aggregation or generalization relationships. Th ese types of relationships can simplify
the individual class descriptions. As in the previous steps, all changes must be recorded
on the CRC cards.
Th e fi fth step is to review the structural model for missing and/or unnecessary classes, attrib-
utes, operations, and relationships. Until this step, the focus of the process has been on add-
ing information to the evolving model. At this point, the focus begins to switch from simply
adding information to also challenging the reasons for including the information contained
in the model. One very useful approach here is to play devil’s advocate, where a team member,
just for the sake of being a pain in the neck, challenges the reasoning for including all aspects
of the model.
Th e sixth step is to incorporate useful patterns into the evolving structural model. A useful
pattern is one that would allow the analyst to more fully describe the underlying domain of
the problem being investigated. Looking at the collection of patterns available (Figure 5-5)
and comparing the classes contained in the patterns with those in the evolving class dia-
gram enable this. Aft er identifying the useful patterns, the analyst incorporates the iden-
tifi ed patterns into the class diagram and modifi es the aff ected CRC cards. Th is includes
adding and removing classes, attributes, operations, and/or relationships.
Th e seventh and fi nal step is to validate the structural model, including both the CRC cards
and the class diagram. We discuss this content in the next section of the chapter and in
Chapter 7.
Campus Housing Example
In the previous chapter, we identifi ed a set of use cases (Add an Apartment, Delete an
Apartment, and Search Available Rental Units) for the campus housing service that helps
students fi nd apartments. By reviewing the use cases, we can easily determine that the cam-
pus housing service must keep track of information for each available apartment and its
owner. Th e information to be captured for each apartment is the location of the apartment,
the number of bedrooms in the apartment, the monthly rent, and how far the apartment is
from the campus. Regarding the owner of the apartment, we need to capture the owner’s
contact information (e.g., name, address, phone number, e-mail address). Since students
are simply users of the system, there is no need to capture any information regarding them;
that is, in this case, students are simply actors. Finally, with regards to relationships among
the classes, there is a single, optional, one to many association relationship that links the
two classes together. Th e Apartment Owner CRC card is portrayed in Figure 5-16, and the
class diagram for this situation is shown in Figure 5-17.
Library Example
As with the Campus Housing example, the fi rst step is to create the CRC cards that repre-
sent the classes in the structural model. In the previous chapter, we used the Library Book
Collection Management System example to describe the process of creating the functional
models (use-case and activity diagrams and use-case descriptions). In this chapter, we fol-
low the same familiar example. Because we are following a use-case-driven approach to
object-oriented systems development, we fi rst review the events described in the use-case
descriptions (see Figure 5-18).
188 Chapter 5 Structural Modeling
Front:
Class Name: Apartment Owner ID: 1
Delete apartment
Add apartment
Responsibilities
Associated Use Cases: 2Description: An apartment owner who has apartments for rent
Type: Concrete, Domain
Apartment
Apartment
Collaborators
Back:
Attributes:
Address (address)
Phone number (PhoneNumber)
Email (EmailAddress)
Name (string)
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: ApartmentFIGURE 5-16
Campus Housing
Apartment Owner
CRC Card
Next, we perform textual analysis on the events by applying the textual analysis rules
described in Figure 5-1. In this case, we can quickly identify the need to include classes
for Borrower, Books, Librarian, Check Out Desk, ID Card, Student Borrower, Faculty/
Staff Borrower, Guest Borrower, Registrar’s Database, Personnel Database, Library’s
Guest Database, Overdue Books, Fines, Book Request. We also can easily identify opera-
tions to “check the validity” of a book request, to “check out” the books, and to “reject” a
book request. Furthermore, the events suggest a “brings” relationship between Borrower
and Books and a “provides” relationship between Borrower and Librarian. Th is step
FIGURE 5-17
Campus Housing
Class Diagram
Apartment Owner Apartment
0..1 0..*
Creating Structural Models Using CRC Cards And Class Diagrams 189
also suggests that we should review the overview section of the use-case description (see
Figure 5-19). In this case, the only additional information gleaned from the use-case
description is the possible inclusion of classes for Personnel Offi ce and Registrar’s Offi ce.
Th is same process would also be completed for the remaining use cases contained in the
functional model: Process Overdue Books, Maintain Book Collection, Search Collection,
and Return Books (see Figure 4-6). Since we did not discuss these use cases in the previous
chapter, we will review the problem description as a basis for beginning the next step (see
Figure 5-20).
Normal Flow of Events:
SubFlows:
1. The Borrower brings books to the Librarian at the check out desk.
2. The Borrower provides Librarian his or her ID card.
3. The Librarian checks the validity of the ID Card.
If the Borrower is a Student Borrower, Validate ID Card against Registrar’s Database.
If the Borrower is a Faculty/Staff Borrower, Validate ID Card against Personnel Database.
If the Borrower is a Guest Borrower, Validate ID Card against Library’s Guest Database.
4. The Librarian checks whether the Borrower has any overdue books and/or fines.
5. The Borrower checks out the books.
Alternate/Exceptional Flows:
4a. The ID Card is invalid, the book request is rejected.
5a. The Borrower either has overdue books fines, or both, the book request is rejected.
FIGURE 5-18 Flow Descriptions for the Borrow Books Use Case (Figure 4-21)
Association: Borrower, Personnel Office, Registrar’s Office
Include:
Extend:
Generalization :
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Type: External
Use Case Type:
Relationships:
Borrow Books 2 High
Borrower
Borrower brings books to check out desk.
Detail, Essential
This use case describes how books are checked out of the library.
Borrower—wants to check out books
Librarian—wants to ensure borrower only gets books deserved
FIGURE 5-19 Overview Description for the Borrow Books Use Case (Figure 4-20)
190 Chapter 5 Structural Modeling
Th e second step is to review the CRC cards to determine if there is any informa-
tion missing. In the case of the library system, because we only used the Borrow Books
use-case description, some information is obviously missing. By reviewing Figure 5-20,
we see that we need to include the ability to search the book collection by title, author,
keywords, and ISBN. Th is obviously implies a Book Collection class with four diff erent
search operations: Search By Title, Search By Author, Search By Keywords, and Search By
ISBN. Interestingly, the description also implies either a set of subclasses or states for the
Book class: Checked Out, Overdue, Requested, Available, and Damaged. We will return
to the issue of states versus subclasses in the next chapter. Th e description implies many
additional operations, including Returning Books, Requesting Books, Adding Books,
Removing Books, Repairing Books, Fining Borrowers, and Emailing Reminders.
Next, we should use our own library experience to brainstorm potential additional
classes, attributes, operations, and relationships that could be useful to include in the
Library Book Collection Management System. In our library, there is also the need to
Retrieve Books From Storage, Move Books to Storage, Request Books from the Interlibrary
Loan System, Return Books to the Interlibrary Loan System, and Deal with E-Books. You
also could include classes for Journals, DVDs, and other media. As you can see, many
classes, attributes, operations, and relationships can be identifi ed.
Th e third step, role-playing the CRC cards, requires us to apply the three role-playing
steps described earlier:
■ Review Use Cases
■ Identify Relevant Actors and Objects
■ Role Play Scenarios
FIGURE 5-20
Overview Description
of the Library
Book Collection
Management System
The functional requirements for an automated university library circulation system include the
need to support searching, borrowing, and book-maintenance activities. The system should
support searching by title, author, keywords, and ISBN. Searching the library’s collection database
should be available on terminals in the library and available to potential borrowers via the World
Wide Web. If the book of interest is currently checked out, a valid borrower should be allowed
to request the book to be returned. Once the book has been checked back in, the borrower
requesting the book should be notifi ed of the book’s availability.
The borrowing activities are built around checking books out and returning books
by borrowers. There are three types of borrowers: students, faculty and staff, and guests.
Regardless of the type of borrower, the borrower must have a valid ID card. If the borrower
is a student, having the system check with the registrar’s student database validates the ID
card. If the borrower is a faculty or staff member, having the system check with the personnel
offi ce’s employee database validates the ID card. If the borrower is a guest, the ID card is
checked against the library’s own borrower database. If the ID card is valid, the system must
also check to determine whether the borrower has any overdue books or unpaid fi nes. If the ID
card is invalid, the borrower has overdue books, or the borrower has unpaid fi nes, the system
must reject the borrower’s request to check out a book; otherwise the borrower’s request
should be honored. If a book is checked out, the system must update the library’s collection
database to refl ect the book’s new status.
The book-maintenance activities deal with adding and removing books from the library’s
book collection. This requires a library manager to both logically and physically add and
remove the book. Books being purchased by the library or books being returned in a damaged
state typically cause these activities. If a book is determined to be damaged when it is returned
and it needs to be removed from the collection, the last borrower will be assessed a fi ne.
However, if the book can be repaired, depending on the cost of the repair, the borrower might
not be assessed a fi ne. Finally, every Monday, the library sends reminder e-mails to borrowers
who have overdue books. If a book is overdue more than two weeks, the borrower is assessed
a fi ne. Depending on how long the book remains overdue, the borrower can be assessed
additional fi nes every Monday.
Creating Structural Models Using CRC Cards And Class Diagrams 191
For our purposes, we will use the Borrow Books use case to demonstrate. Th e relevant
actors include Student Borrowers, Faculty/Staff Borrowers, Guest Borrowers, Librarians,
Personnel Offi ce, and Registrar’s Offi ce. Th ese can be easily gleaned from the overview
section of the use-case description (see Figure 5-19) and the use-case diagram (see Figure
4-6). Th e relevant objects seem to include Books, Borrower, and ID Card. Finally, to
role-play the scenarios, we need to assign the roles to the diff erent members of the team
and try to perform each of the paths through the events of the use-case (see Figure 5-18).
Based on the Events of the use case and the use case’s activity diagram (see Figure 5-21),
we can quickly identify nine scenarios, three for each type of Borrower (Student, Faculty/
Staff , and Guest): Valid ID and No Overdue Books & No Fines, Valid ID only, and no Valid
ID. When role-playing these scenarios, one question arises: What happens to the books that
are requested when the request is rejected? Based on the current functional and structural
models, the books are left sitting on the check out desk. Th at doesn’t quite seem right. In
reality, the books are reshelved. In fact, the notion of reshelving books is also relevant to
when books are checked back in or aft er books have been repaired. Furthermore, the idea of
adding books to the collection should also include the operation of shelving the books. As
you should readily see, building structural models will also help uncover behavior that was
omitted when building the functional models. Remember, object-oriented systems develop-
ment is not only use-case driven but also is incremental and iterative.
Th e fourth step is to put everything together and to draw the class diagram. Figure 5-22
represents the fi rst cut at drawing the class diagram for the Library Book Collection Management
System. Th e classes identifi ed in the previous steps have been linked with other classes via asso-
ciation, aggregation, and generalization relationships. For simplicity purposes, we only show the
classes and their relationships; not their attributes, operations, or even the multiplicities on the
association relationships.
FIGURE 5-21
Activity Diagram
for the Borrow
Books Use Case
(Figure 4-12)
[Valid Card]
[No Overdue Books & No Fines]
Validate ID Card
Check Out Books
Check for Overdue Books and Fines
192
B
oo
k
C
ol
le
ct
io
n
B
or
ro
w
er
ID
St
ud
en
t
Li
br
ar
ia
n
St
ud
en
t
ID
R
eg
is
tr
ar
’s
D
B
Pe
rs
on
ne
l D
B
Fa
cu
lt
y/
St
af
f I
D
Li
br
ar
y
D
B
G
ue
st
I
D
Fa
cu
lt
y/
St
af
f
G
ue
st
Fi
ne
R
eq
ue
st
Li
br
ar
y
Sh
el
f
R
es
he
lv
in
g
C
he
ck
O
ut
D
es
k
St
or
ag
e
Pe
rs
O
ff
R
eg
is
tr
ar
O
ff
In
te
rL
ib
ra
ry
L
oa
n
Sy
st
em
B
oo
k
D
V
D
Lo
ca
ti
on
B
oo
k
Lo
ca
ti
on
Jo
ur
na
l
O
th
er
M
ed
ia
M
ed
ia
* 1
FI
G
U
R
E
5-
22
Fi
rs
t-
C
u
t
C
la
ss
D
ia
gr
am
f
o
r
th
e
Li
b
ra
ry
B
o
o
k
C
o
lle
ct
io
n
S
ys
te
m
Creating Structural Models Using CRC Cards And Class Diagrams 193
FIGURE 5-23 Second-Cut Class Diagram for the Library Book Collection System
Student
Borrower Book
Guest Storage
Book Location
LibraryInterLibrary Loan System
Book Collection
Faculty/Staff
Librarian
0..* 0..* 1..10..*
1
*
Th e fi fth step is to carefully review what has been created. Not only should you look
for any missing classes, attributes, operations, and/or relationships, but you should also
challenge every aspect of the current model. Specifi cally, are there classes, attributes,
operations, and/or relationships that should be removed from the model? If so, there may
be classes on the diagram that should have been modeled as attributes. For example, the
Student, Fac/Staff , and Guest IDs should have been attributes with their respective classes.
Furthermore, because this is a book collection management system, the inclusion of other
media seems to be inappropriate. Finally, the Personnel Offi ce and Registrar’s Offi ce are
actually only actors in the system, not objects. Based on all of these deletions, a new version
of the class diagram was drawn (see Figure 5-23). Th is diagram is much simpler and easier
to understand.
Th e sixth step, incorporating useful patterns, enables us to take advantage of knowledge
that was developed elsewhere. In this case, the pattern used in the library problem includes
too many ideas that are not relevant to the current problem. However, by looking back to
Figure 5-3, we see that one of the original patterns (the Place, Transaction, Participant,
Transaction Line Item, and Item pattern—see the top left of the fi gure) is relevant. We
incorporate that pattern into the class diagram by replacing Place by Check Out Desk,
Participant by Borrower, Transaction by Check Out Trans, and Item by Book (Figure 5-24).
Technically speaking, each of these replacements is simply a pattern customized to the
problem at hand. We also then add the Transaction Line Item class that we had missed in
the original structural model.
Th e seventh step is to review the current state of the structural model. Needless to say, the
CRC card version and the class diagram version are no longer in agreement with each other.
We return to this step in the next section of the chapter.
194 Chapter 5 Structural Modeling
VERIFYING AND VALIDATING THE STRUCTURAL MODEL17
Before we move on to creating behavioral models (see Chapter 6) of the problem domain,
we need to verify and validate the structural model. In the previous chapter, we introduced
the notion of walkthroughs as a way to verify and validate business processes and functional
models. In this chapter, we combine walkthroughs with the power of role-playing as a way
to more completely verify and validate the structural model that will underlie the business
processes and functional models. In fact, all of the object identifi cation approaches described
in this chapter can be viewed as a way to test the fi delity of the structural model. Because we
have already introduced the idea of role-playing the CRC cards and object identifi cation, in
this section we focus on performing walkthroughs.
In this case, the verifi cation and validation of the structural model are accomplished
during a formal review meeting using a walkthrough approach in which an analyst presents
the model to a team of developers and users. Th e analyst walks through the model, explaining
each part of the model and all the reasoning behind the decision to include each of the classes
in the structural model. Th is explanation includes justifi cations for the attributes, operations,
and relationships associated with the classes. Each class should be linked back to at least one
use case; otherwise, the purpose of including the class in the structural model will not be
17 Th e material in this section has been adapted from E. Yourdon, Modern Structured Analysis (Englewood Cliff s,
NJ: Prentice Hall, 1989).
FIGURE 5-24 Class Diagram with Incorporated Pattern for the Library Book Collection System
Check Out Trans Transaction Line Item Book
Check Out Desk
Storage
Book Location
LibraryInterLibrary Loan System
Book Collection
Student
Borrower
GuestFaculty/Staff
Librarian
1..1 0..* 1..10..*
1..1
0..*
0..*
1..1
1..11..*
1
*
Verifying and Validating the Structural Model 195
understood. Also including people outside the development team who produced the model
can bring a fresh perspective to the model and uncover missing objects.
Previously, we suggested three representations that could be used for structural mode-
ling: CRC cards, class diagrams, and object diagrams. Because an object diagram is simply
an instantiation of some part of a class diagram, we limit our discussion to CRC cards and
class diagrams. Similar to how we verifi ed and validated the business process and func-
tional models in the last chapter, we provide a set of rules that will test the consistency
within the structural models. For example purposes, we use the appointment problem
described in Chapter 4 and in this chapter. An example of the CRC card for the old patient
class is shown in Figure 5-6, and the associated class diagram is portrayed in Figure 5-7.
First, every CRC card should be associated with a class on the class diagram, and vice
versa. In the appointment example, the Old Patient class represented by the CRC card does
not seem to be included on the class diagram. However, there is a Patient class on the class
diagram (see Figures 5-6 and 5-7). Th e Old Patient CRC card most likely should be changed
to simply Patient.
Second, the responsibilities listed on the front of the CRC card must be included as
operations in a class on a class diagram, and vice versa. Th e make appointment responsibil-
ity on the new Patient CRC card also appears as the make appointment() operation in the
Patient class on the class diagram. Every responsibility and operation must be checked.
Th ird, collaborators on the front of the CRC card imply some type of relationship on
the back of the CRC card and some type of association that is connected to the associated
class on the class diagram. Th e appointment collaborator on the front of the CRC card also
appears as another association on the back of the CRC card and as an association on the
class diagram that connects the Patient class with the Appointment class.
Fourth, attributes listed on the back of the CRC card must be included as attributes in
a class on a class diagram, and vice versa. For example, the amount attribute on the new
Patient CRC card is included in the attribute list of the Patient class on the class diagram.
Fift h, the object type of the attributes listed on the back of the CRC card and with the
attributes in the attribute list of the class on a class diagram implies an association from the
class to the class of the object type. For example, technically speaking, the amount attribute
implies an association with the double type. However, simple types such as int and double
are never shown on a class diagram. Furthermore, depending on the problem domain, object
types such as Person, Address, or Date might not be explicitly shown either. However, if we
know that messages are being sent to instances of those object types, we probably should
include these implied associations as relationships.
Sixth, the relationships included on the back of the CRC card must be portrayed using
the appropriate notation on the class diagram. For example in Figure 5-6, instances of the
Patient class are a-kind-of Person, it has instances of the Medical History class as part of it,
and it has an association with instances of the Appointment class. Th us, the association from
the Patient class to the Person class should indicate that the Person class is a generalization of
its subclasses, including the Patient class; the association from the Patient class to the Medical
History class should be in the form of an aggregation association (a white diamond); and the
association between instances of the Patient class and instances of the Appointment class
should be a simple association. However, when we review the class diagram in Figure 5-7,
this is not what we fi nd. If you recall, we included in the class diagram the transaction pattern
portrayed in Figure 5-4. When we did this, many changes were made to the classes contained
in the class diagram. All of these changes should have been cascaded back through all of
the CRC cards. In this case, the CRC card for the Patient class should show that a Patient is
a-kind-of Participant (not Person) and that the relationship from Patient to Medical History
should be a simple association (see Figure 5-25).
196 Chapter 5 Structural Modeling
FIGURE 5-25
Patient CRC Card
Front:
Class Name: Patient ID: 3
Medical history
Make appointment Appointment
Change status
Calculate last visit
Provide medical history
Responsibilities
Associated Use Cases: 2Description: An individual who needs to receive or has received
medical attention
Type: Concrete, Domain
Collaborators
Back:
Attributes:
Insurance carrier (text)
Amount (double)
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: Appointment, Medical History
Participant
Seventh, an association class, such as the Treatment class in Figure 5-7, should be created
only if there is indeed some unique characteristic (attribute, operation, or relationship) about
the intersection of the connecting classes. If no unique characteristic exists, then the associ-
ation class should be removed and only an association between the two connecting classes
should be displayed.
Finally, as in the functional models, specifi c representation rules must be enforced. For
example, a class cannot be a subclass of itself. Th e Patient CRC card cannot list Patient with
the generalization relationships on the back of the CRC card, nor can a generalization rela-
tionship be drawn from the Patient class to itself. Again, all the detailed restrictions for each
representation are beyond the scope of this book.18 Figure 5-26 portrays the associations
among the structural models.
18 A good reference for these types of restrictions is S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, UK:
Cambridge University Press, 2005).
Verifying and Validating the Structural Model 197
Structural Models
Including
Contains Contains
Contains
Class Diagram
Responsibilities
Collaborators
Association Aggregation
Composition
Attributes
Classes
Type
CRC Cards
Objects
Object Diagram
Contains
Represents
Associations/
Relationships
Have
HasKinds
Generalization
Association Class
HasKindsHasKinds
AssociatedWith
AssociatedWith
AssociatedWith
AssociatedWith
Operations
InstanceOf
FIGURE 5-26 Interrelationships among Structural Models
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
In Chapter 4, you learned how the functional models were developed in an iterative
manner. Aft er creating the functional model for the Mobile Scheduling (Version 1) of
the Integrated Health Clinic Delivery System, the team had a good understanding of the
business processes. Now it is time to identify the key data and to develop the structural
model of the objects that support those business processes. Structural modeling for
Mobile Scheduling (Version 1) involves creating, verifying, and validating CRC cards,
class diagram, and object diagrams.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
198 Chapter 5 Structural Modeling
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the purpose of a structural model.
Describe the diff erent elements of a structural model.
Explain the diff erence between abstract and concrete classes.
Describe the three general types of relationships typically used in a structural model.
Create a structural model using textual analysis of use-case descriptions, brainstorming, common object lists,
and patterns.
Explain the purpose of a CRC card in structural modeling.
Create a structural model using CRC cards.
Describe the diff erent elements of a CRC card.
Describe how to role-play CRC cards using use-case scenarios.
Describe the diff erent elements of a class diagram.
Describe the four basic operations that can be represented on a class diagram.
Explain the diff erences between the types of relationships supported on a class diagram.
Create a class diagram that represents a structural model.
Describe the diff erent elements of an object diagram.
Create an object diagram that represents an instantiation of a portion of a class diagram.
Verify and validate the evolving structural model using role-playing and walkthroughs.
Verify and validate the functional model by ensuring the consistency of the three structural representations: CRC
cards, class diagrams, and object diagrams.
KEY TERMS
A-kind-of
A-part-of
Abstract class
Aggregation association
Assemblies
Association
Association class
Attribute
Brainstorming
Class
Class diagram
Client
Collaboration
Common object list
Conceptual model
Concrete class
Constructor operation
Contract
Class–Responsibility–
Collaboration (CRC)
CRC cards
Decomposition
Derived attribute
Doing responsibility
Destructor operation
Generalization
association
Has-parts
Incidents
Instance
Interactions
Knowing responsibility
Method
Multiplicity
Object
Object diagram
Operation
Package
Parts
Pattern
Private
Protected
Public
Query operation
Responsibility
Role-playing
Roles
Server
State
Static model
Static structure
diagram
Structural model
Subclass
Substitutability
Superclass
SVDPI
Tangible things
Textual analysis
Update operation
View
Visibility
Wholes
QUESTIONS
1. Describe to a businessperson the multiplicity of a rela-
tionship between two classes.
2. Why are assumptions important to a structural
model?
3. What is an association class?
4. Contrast the following sets of terms: object, class,
method, attribute, superclass, subclass, concrete class,
abstract class.
5. Give three examples of derived attributes that may
exist on a class diagram. How would they be denoted
on the class diagram?
Exercises 199
EXERCISES
A. Create a CRC card for each of the following classes:
Movie (title, producer, length, director, genre)
Ticket (price, adult or child, showtime, movie)
Patron (name, adult or child, age)
B. Create a class diagram based on the CRC cards you
created for exercise A.
C. Create a CRC card for each of the following classes.
Consider that the entities represent a system for a
patient billing system. Include only the attributes that
would be appropriate for this context.
Patient (age, name, hobbies, blood type, occupation,
insurance carrier, address, phone)
Insurance carrier (name, number of patients on plan,
address, contact name, phone)
Doctor (specialty, provider identifi cation number,
golf handicap, age, phone, name)
D. Create a class diagram based on the CRC cards you
created for exercise C.
E. Draw a class diagram for each of the following situations:
1. Whenever new patients are seen for the fi rst time,
they complete a patient information form that asks
their name, address, phone number, and insurance
carrier, which are stored in the patient information
fi le. Patients can be signed up with only one carrier,
but they must be signed up to be seen by the doctor.
Each time a patient visits the doctor, an insurance
claim is sent to the carrier for payment. Th e claim
must contain information about the visit, such as
the date, purpose, and cost. It would be possible for
a patient to submit two claims on the same day.
2. Th e state of Georgia is interested in designing a
system that will track its researchers. Information
of interest includes researcher name, title, position,
researcher’s university name, university location, uni-
versity enrollment, and researcher’s research inter-
ests. Researchers are associated with one institution,
and each researcher has several research interests.
3. A department store has a wedding registry. Th is
registry keeps information about the customer
(usually the bride), the products that the store
carries, and the products for which each customer
registers. Customers typically register for a large
number of products, and many customers register
for the same products.
4. Jim Smith’s dealership sells Fords, Hondas, and
Toyotas. In order to get in touch with these man-
ufacturers easily, the dealership keeps informa-
tion about each of them. Th e dealership keeps
information about the models of cars from each
6. What are the diff erent types of visibility? How would
they be denoted on a class diagram?
7. Draw the relationships that are described by the fol-
lowing business rules. Include the multiplicities for
each relationship.
A patient must be assigned to only one doctor, and a
doctor can have one or many patients.
An employee has one phone extension, and a unique
phone extension is assigned to an employee.
A movie theater shows at least one movie, and a movie
can be shown at up to four other movie theaters
around town.
A movie either has one star, two costars, or more than
ten people starring together. A star must be in at
least one movie.
8. How do you designate the reading direction of a rela-
tionship on a class diagram?
9. For what is an association class used in a class dia-
gram? Give an example of an association class that
may be found in a class diagram that captures students
and the courses that they have taken.
10. Give two examples of aggregation, generalization, and
association relationships. How is each type of associa-
tion depicted on a class diagram?
11. Identify the following operations as constructor,
query, or update. Which operations would not need
to be shown in the class rectangle?
Calculate employee raise (raise percent)
Calculate sick days ()
Increment number of employee vacation days ()
Locate employee name ()
Place request for vacation (vacation day)
Find employee address ()
Insert employee ()
Change employee address ()
Insert spouse ()
12. How are the diff erent structural models related, and how
does this aff ect verifi cation and validation of the model?
200 Chapter 5 Structural Modeling
manufacturer, including dealer price, model name,
and series (e.g., Honda, Civic, LX). Additionally,
the dealership also keeps all sales information,
including buyer’s name, address and phone number,
car purchased, and amount paid.
F. Create object diagrams based on the class diagrams
you drew for exercise F.
G. Examine the class diagrams that you created for exer-
cise F. How would the models change (if at all) based
on these new assumptions?
1. Two patients have the same fi rst and last names.
2. Researchers can be associated with more than one
institution.
3. Th e store would like to keep track of purchase
items.
4. Many buyers have purchased multiple cars from
Jim over time because he is such a good dealer.
H. Visit a website that allows customers to order a
product over the Web (e.g., Amazon.com). Create a
structural model (CRC cards and class diagram) that
the site must need to support its business process.
Include classes to show what they need information
about. Be sure to include the attributes and operations
to represent the type of information they use and cre-
ate. Finally, draw relationships, making assumptions
about how the classes are related.
I. Using the seven-step process described in this chapter,
create a structural model (CRC cards and class dia-
gram) for exercise C in Chapter 4.
J. Perform a verifi cation and validation walkthrough for
the structural model created for exercise J.
K. Using the seven-step process described in this
chapter, create a structural model for exercise E in
Chapter 4.
L. Perform a verifi cation and validation walkthrough for
the structural model created for exercise L.
M. Using the seven-step process described in this
chapter, create a structural model for exercise G in
Chapter 4.
N. Perform a verifi cation and validation walkthrough for
the structural model created for exercise N.
O. Using the seven-step process described in this chapter,
create a structural model for exercise I in Chapter 4.
P. Perform a verifi cation and validation walkthrough for
the structural model created for exercise P.
Q. Using the seven-step process described in this chapter,
create a structural model for exercise L in Chapter 4.
R. Perform a verifi cation and validation walkthrough for
the structural model created for exercise R.
S. Using the seven-step process described in this
chapter, create a structural model for exercise O in
Chapter 4.
T. Perform a verifi cation and validation walkthrough for
the structural model created for exercise T.
U. Using the seven-step process described in this chapter,
create a structural model for exercise R in Chapter 4.
V. Perform a verifi cation and validation walkthrough for
the structural model created for exercise V.
W. Using the seven-step process described in this
chapter, create a structural model for exercise U in
Chapter 4.
X. Perform a verifi cation and validation walkthrough for
the structural model created for exercise X.
MINICASES
1. West Star Marinas is a chain of twelve marinas that
off er lakeside service to boaters; service and repair of
boats, motors, and marine equipment; and sales of
boats, motors, and other marine accessories. Th e sys-
tems development project team at West Star Marinas
has been hard at work on a project that eventually
will link all the marina’s facilities into one unifi ed,
networked system.
Th e project team has developed a use-case diagram
of the current system. Th is model has been carefully
checked. Last week, the team invited a number of
system users to role-play the various use cases, and
the use cases were refi ned to the users’ satisfaction.
Right now, the project manager feels confi dent that
the as-is system has been adequately represented in
the use-case diagram.
Th e director of operations for West Star is the
sponsor of this project. He sat in on the role-playing
of the use cases and was very pleased by the thorough
job the team had done in developing the model. He
made it clear to you, the project manager, that he
was anxious to see your team begin work on the use
cases for the to-be system. He was a little skeptical
that it was necessary for your team to spend any
Minicases 201
time modeling the current system in the fi rst place
but grudgingly admitted that the team really seemed
to understand the business aft er going through that
work.
Th e methodology you are following, however,
specifi es that the team should now turn its attention
to developing the structural models for the as-is sys-
tem. When you stated this to the project sponsor, he
seemed confused and a little irritated. “You are going
to spend even more time looking at the current sys-
tem? I thought you were done with that! Why is this
necessary? I want to see some progress on the way
things will work in the future!”
What is your response to the director of operations?
Why do we perform structural modeling? Is there any
benefi t to developing a structural model of the current
system at all? How do the use cases and use-case dia-
gram help us develop the structural model?
2. Holiday Travel Vehicles sells new recreational vehi-
cles and travel trailers. When new vehicles arrive at
Holiday Travel Vehicles, a new vehicle record is cre-
ated. Included in the new vehicle record are a vehicle
serial number, name, model, year, manufacturer, and
base cost.
When a customer arrives at Holiday Travel Vehi-
cles, he or she works with a salesperson to negotiate a
vehicle purchase. When a purchase has been agreed
upon, a sales invoice is completed by the salesperson.
Th e invoice summarizes the purchase, including full
customer information, information on the trade-in
vehicle (if any), the trade-in allowance, and infor-
mation on the purchased vehicle. If the customer
requests dealer-installed options, they are listed on the
invoice as well. Th e invoice also summarizes the fi nal
negotiated price, plus any applicable taxes and license
fees. Th e transaction concludes with a customer signa-
ture on the sales invoice.
a. Identify the classes described in the preceding scenario
(you should fi nd six). Create CRC cards for each class.
Customers are assigned a customer ID when they
make their fi rst purchase from Holiday Travel Vehi-
cles. Name, address, and phone number are recorded
for the customer. Th e trade-in vehicle is described by a
serial number, make, model, and year. Dealer-installed
options are described by an option code, description,
and price.
b. Develop a list of attributes for each class. Place the
attributes onto the CRC cards.
Each invoice lists just one customer. A person does
not become a customer until he or she purchases a
vehicle. Over time, a customer may purchase a num-
ber of vehicles from Holiday Travel Vehicles.
Every invoice must be fi lled out by only one sales-
person. A new salesperson might not have sold any
vehicles, but experienced salespeople have probably
sold many vehicles.
Each invoice only lists one new vehicle. If a new
vehicle in inventory has not been sold, there will be no
invoice for it. Once the vehicle sells, there will be just
one invoice for it.
A customer may decide to have no options added
to the vehicle or may choose to add many options. An
option may be listed on no invoices, or it may be listed
on many invoices.
A customer may trade in no more than one vehicle
on a purchase of a new vehicle. Th e trade-in vehicle
may be sold to another customer who later trades it in
on another Holiday Travel vehicle.
c. Based on the preceding business rules in force at
Holiday Travel Vehicles and CRC cards, draw a
class diagram and document the relationships with
the appropriate multiplicities. Remember to update
the CRC cards.
202
Behavioral models describe the internal dynamic aspects of an information system that
supports the business processes in an organization. During analysis, behavioral models
describe what the internal logic of the processes is without specifying how the processes
are to be implemented. Later, in the design and implementation phases, the detailed design
of the operations contained in the object is fully specifi ed. In this chapter, we describe three
Unifi ed Modeling Language (UML) diagrams that are used in behavioral modeling (sequence
diagrams, communication diagrams, and behavioral state machines) and CRUDE (create,
read, update, delete, execute) matrices.
OBJECTIVES
■ Understand the rules and style guidelines for sequence and communication diagrams
and behavioral state machines.
■ Understand the processes used to create sequence and communication diagrams,
behavioral state machines, and CRUDE matrices.
■ Be able to create sequence and communication diagrams, behavioral state machines,
and CRUDE matrices.
■ Understand the relationship between the behavioral models and the structural and
functional models.
INTRODUCTION
Th e previous two chapters discussed how analysts create both business process and functional
models and structural models. Systems analysts use business process and functional models to
describe the functional or external behavioral view of an information system. And, they use
structural models to depict the internal structural or static view of an information system. In
this chapter, we discuss how analysts use behavioral models to represent the internal behavior
or dynamic view of an information system.
By supporting all three views (functional, structural, and behavioral), object-oriented
systems analysis and design supports an architecture-centric approach to developing infor-
mation systems. Furthermore, the behavioral view is driven by the original use cases uncov-
ered during business process and functional modeling. As such, behavioral modeling is also
use case driven. Finally, as with business process and functional modeling and structural
modeling, you will fi nd that you will need to not only iterate across the behavioral models
(described in this chapter), but you will also have to iterate across all three architectural
views (functional, structural, and behavioral) to capture and represent the requirements for
a business information system.
Th ere are two types of behavioral models. First, there are behavioral models used to rep-
resent the underlying details of a business process portrayed by a use-case model. In UML,
C H A P T E R 6
Behavioral Modeling
Behavioral Models 203
interaction diagrams (sequence and communication) are used for this type of behavioral
model. Practically speaking, interaction diagrams allow the analyst to model the distribution
of the behavior of the system over the actors and objects in the system. In this way, we can
easily see how actors and objects collaborate to provide the functionality defi ned in a use case.
Second, a behavioral model is used to represent the changes that occur in the underlying data.
UML uses behavioral state machines for this.
During analysis, analysts use behavioral models to capture a basic understanding of
the dynamic aspects of the underlying business process. Traditionally, behavioral models
have been used primarily during design, where analysts refi ne the behavioral models to
include implementation details (see Chapter 8). For now, our focus is on what the dynamic
view of the evolving system is and not on how the dynamic aspect of the system will be
implemented.
In this chapter, we concentrate on creating behavioral models of the underlying busi-
ness process. Using the interaction diagrams (sequence and communication diagrams)
and behavioral state machines, it is possible to give a complete view of the dynamic
aspects of the evolving business information system. We fi rst describe behavioral models
and their components. We then describe each of the diagrams, how they are created,
and how they are related to the functional and structural models described in Chapters
4 and 5. Finally, we describe CRUDE analysis and the process to verify and validate the
behavioral models.
BEHAVIORAL MODELS
When an analyst is attempting to understand the underlying application domain of a
problem, he or she must consider both structural and behavioral aspects of the problem.
Unlike other approaches to the development of information systems, object-oriented
approaches attempt to view the underlying application domain in a holistic manner. By
viewing the problem domain as a set of use cases that are supported by a set of collabo-
rating objects, object-oriented approaches allow an analyst to minimize the semantic gap
between the real-world set of objects and the evolving object-oriented model of the prob-
lem domain. However, as we pointed out in the previous chapter, the real world tends to
be messy; because soft ware must be logical to work, perfect modeling of the application
domain is nearly impossible.
One of the primary purposes of behavioral models is to show how the underlying
objects in a problem domain will work together to form a collaboration to support each
of the use cases. Whereas structural models represent the objects and the relationships
between them, behavioral models depict the internal view of the business process that a
use case describes. Th e process can be shown by the interaction that takes place between
the objects that collaborate to support a use case through the use of interaction (sequence
and communication) diagrams. It is also possible to show the eff ect that the set of use cases
that make up the system has on the objects in the system through the use of behavioral
state machines.
Creating behavioral models is an iterative process that iterates not only over the indi-
vidual behavioral models [e.g., interaction (sequence and communication) diagrams and
behavioral state machines] but also over the functional (see Chapter 4) and structural (see
Chapter 5) models. As the behavioral models are created, it is not unusual to make changes
to the functional and structural models. In this chapter, we describe interaction diagrams,
behavioral state machines, and CRUDE analysis and when to use each.
204 Chapter 6 Behavioral Modeling
INTERACTION DIAGRAMS
One of the primary diff erences between class diagrams and interaction diagrams, besides the
obvious diff erence that one describes structure and the other behavior, is that the modeling
focus on a class diagram is at the class level, whereas the interaction diagrams focus on the
object level. In this section, we review objects, operations, and messages and we cover the two
diff erent diagrams (sequence and communication) that can be used to model the interactions
that take place between the objects in an information system.
Objects, Operations, and Messages
An object is an instantiation of a class, i.e., an actual person, place, or thing about which
we want to capture information. If we were building an appointment system for a doctor’s
offi ce, classes might include doctor, patient, and appointment. Th e specifi c patients, such as
Jim Maloney, Mary Wilson, and Th eresa Marks, are considered objects—i.e., instances of the
patient class.
Each object has attributes that describe information about the object, such as a patient’s
name, birth date, address, and phone number. Each object also has behaviors. At this point
in the development of the evolving system, the behaviors are described by operations. An
operation is nothing more than an action that an object can perform. For example, an
appointment object can probably schedule a new appointment, delete an appointment,
and locate the next available appointment. Later on during the development of the evolving
system, the behaviors will be implemented as methods.
Each object also can send and receive messages. Messages are information sent to objects
to tell an object to execute one of its behaviors. Essentially, a message is a function or proce-
dure call from one object to another object. For example, if a patient is new to the doctor’s
offi ce, the system sends an insert message to the application. Th e patient object receives the
instruction (the message) and does what it needs to do to insert the new patient into the sys-
tem (the behavior).
Sequence Diagrams
Sequence diagrams are one of two types of interaction diagrams. Th ey illustrate the objects
that participate in a use case and the messages that pass between them over time for one use
case. A sequence diagram is a dynamic model that shows the explicit sequence of messages
that are passed between objects in a defi ned interaction. Because sequence diagrams empha-
size the time-based ordering of the activity that takes place among a set of objects, they are
very helpful for understanding real-time specifi cations and complex use cases.
Th e sequence diagram can be a generic sequence diagram that shows all possible scenar-
ios1 for a use case, but usually each analyst develops a set of instance sequence diagrams, each
of which depicts a single scenario within the use case. If you are interested in understanding
the fl ow of control of a scenario by time, you should use a sequence diagram to depict this
information. Th e diagrams are used throughout the analysis and design phases. However, the
design diagrams are very implementation specifi c, oft en including database objects or specifi c
user interface components as the objects.
Elements of a Sequence Diagram Figure 6-1 shows an instance sequence diagram that depicts
the objects and messages for the Make Old Patient Appt use case, which describes the process by
which an existing patient creates a new appointment or cancels or reschedules an appointment
1 Remember that a scenario is a single executable path through a use case.
Interaction Diagrams 205
RequestAppt(name, address)
NewCancelChangeAppt?()
ApptTimes?()
aPatient
LookUpPatient()
aReceptionist
[aPatient Exists] LookupBills()
MatchAppts()
CreateAppt()
aPatient:Patient :UnpaidBill :Appointment
sd Make Appt Use Case
FIGURE 6-1 Example Sequence Diagram
for the doctor’s offi ce appointment system. In this specifi c instance, the Make Old Patient Appt
process is portrayed.
Actors and objects that participate in the sequence are placed across the top of the dia-
gram using actor symbols from the use-case diagram and object symbols from the object
diagram (see Figure 6-2). Notice that the actors and objects in Figure 6-1 are aPatient, aRe-
ceptionist, aPatient, UnpaidBill, and Appointment.2 For each of the objects, the name of the
class of which they are an instance is given aft er the object’s name (e.g., aPatient means that
aPatient is an instance of the Patient class).
A dotted line runs vertically below each actor and object to denote the lifeline of the
actors and objects over time (see Figure 6-1).3 Sometimes an object creates a temporary
object; in this case, an X is placed at the end of the lifeline at the point where the object is
destroyed (not shown). For example, think about a shopping cart object for a Web com-
merce application. Th e shopping cart is used for temporarily capturing line items for an
order, but once the order is confi rmed, the shopping cart is no longer needed. In this case,
an X would be located at the point at which the shopping cart object is destroyed. When
objects continue to exist in the system aft er they are used in the sequence diagram, then
the lifeline continues to the bottom of the diagram (this is the case with all of the objects
in Figure 6-1).
2 In some versions of the sequence diagram, object symbols are used as surrogates for the actors. However, for clarity,
we recommend using actor symbols for actors instead.
3 Technically speaking, in UML 2.0 the lifeline actually refers to both the object (actor) and the dashed line drawn
vertically underneath the object (actor). However, we prefer to use the older terminology because it is more
descriptive of what is actually being represented.
206 Chapter 6 Behavioral Modeling
Context
An actor:
■ Is a person or system that derives benefit from and is external to the system.
■ Participates in a sequence by sending and/or receiving messages.
■ Is placed across the top of the diagram.
■ Is depicted either as a stick figure (default) or, if a nonhuman actor is involved, as
a rectangle with <> in it (alternative).
An object:
■ Participates in a sequence by sending and/or receiving messages.
■ Is placed across the top of the diagram.
A lifeline:
■ Denotes the life of an object during a sequence.
■ Contains an X at the point at which the class no longer interacts.
An execution occurrence:
■ Is a long narrow rectangle placed atop a lifeline.
■ Denotes when an object is sending or receiving messages.
A message:
■ Conveys information from one object to another one.
■ A operation call is labeled with the message being sent and a solid arrow, whereas
a return is labeled with the value being returned and shown as a dashed arrow.
A guard condition:
■ Represents a test that must be met for the message to be sent.
<>
anActor
anActor
aMessage()
[aGuardCondition]:aMessage()
ReturnValue
For object destruction:
■ An X is placed at the end of an object’s lifeline to show that it is going out
of existence.
A frame:
■ Indicates the context of the sequence diagram.
X
Term and Definition Symbol
anObject : aClass
FIGURE 6-2 Sequence Diagram Syntax
A thin rectangular box, called the execution occurrence, is overlaid onto the lifeline
to show when the classes are sending and receiving messages (see Figure 6-2). A message
is a communication between objects that conveys information with the expectation that
Interaction Diagrams 207
activity will ensue. Many diff erent types of messages can be portrayed on a sequence diagram.
However, in the case of using sequence diagrams to model use cases, two types of messages
are typically used: operation call and return. Operation call messages passed between objects
are shown using solid lines connecting two objects with an arrow on the line showing which
way the message is being passed. Argument values for the message are placed in parentheses
next to the message’s name. Th e order of messages goes from the top to the bottom of the
page, so messages located higher on the diagram represent messages that occur earlier on in
the sequence, versus the lower messages that occur later. A return message is depicted as a
dashed line with an arrow on the end of the line portraying the direction of the return. Th e
information being returned is used to label the arrow. However, because adding return mes-
sages tends to clutter the diagram, unless the return messages add a lot of information to the
diagram, they can be omitted. For example, in Figure 6-1, no return messages are depicted.4
In Figure 6-1, LookUpPatient() is a message sent from the actor aReceptionist to the object
aPatient to determine whether the aPatient actor is a current patient.
At times a message is sent only if a condition is met. In those cases, the condition
is placed between a set of brackets, [ ]—for example, [aPatient Exists] LookupBills().
Th e condition is placed in front of the message name. However, when using a sequence
diagram to model a specifi c scenario, conditions are typically not shown on any single
sequence diagram. Instead, conditions are implied only through the existence of diff erent
sequence diagrams.
An object can send a message to itself, e.g., Create Sandwich in Figure 6-3. Th is is known
as self-delegation. Sometimes, an object creates another object. Th is is shown by the message
being sent directly to an object instead of its lifeline.
Figure 6-3 portrays two additional examples of instance-specifi c sequence diagrams. Th e
fi rst one is related to the Make Lunch use case that was described in the activity diagram por-
trayed in Figure 4-10. Th e second one is related to the Place Order use case associated with
the activity diagram in Figure 4-9. In both examples, the diagrams simply represent a single
scenario. Notice in the Make Lunch sequence diagram there is a message being sent from
an actor to itself [CreateSandwich()]. Depending on the complexity of the scenario being
modeled, this particular message could have been eliminated. Obviously, both the process
of making a lunch and placing an order can be quite a bit more complex. However, from a
learning point of view, you should be able to see how the sequence diagrams and the activity
diagrams relate to one another.
Guidelines for Creating Sequence Diagrams Ambler5 provides a set of guidelines when
drawing sequence diagrams. In this section, we review six of them.
■ Try to have the messages not only in a top-to-bottom order but also, when possible, in
a left -to-right order. Given that Western cultures tend to read left to right and top to
bottom, a sequence diagram is much easier to interpret if the messages are ordered
as much as possible in the same way. To accomplish this, order the actors and objects
along the top of the diagram in the order that they participate in the scenario of the
use case.
■ If an actor and an object conceptually represent the same idea, one inside of the
soft ware and the other outside, label them with the same name. In fact, this implies
that they exist in both the use-case diagram (as an actor) and in the class diagram
4 However, some CASE tools require the return messages to be displayed. Obviously, when you are using these tools,
you have to include the return messages on the diagram.
5 S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, England: Cambridge University Press, 2005).
208 Chapter 6 Behavioral Modeling
(as a class). At fi rst glance, this might seem to lead to confusion. However, if they do
indeed represent the same idea, then they should have the same name. For example,
a customer actor interacts with the system and the system stores information about
the customer. In this case, they do indeed represent the same conceptual idea.
■ Th e initiator of the scenario—actor or object—should be the drawn as the farthest left
item in the diagram. Th is guideline is essentially a specialization of the fi rst guideline.
In this case, it relates specifi cally to the actor or object that triggers the scenario.
■ When there are multiple objects of the same type, be sure to include a name for
the object in addition to the class of the object. For example, in the making a lunch
example (see Figure 6-3) there are two objects of type Parent. As such, they should be
named. Otherwise, you can simply use the class name. Th is will simplify the diagram.
In this case, the Child object did not have to be named. We could have simply placed
a colon in front of the classname instead.
sd Make Lunch Use Case
sd Submit Order Use Case
MakeLunch
Lunch
aChild:Child firstParent:Parent secondParent:Parent
CreateLunch
Sandwich
Lunch
GetSandwich
Create Sandwich
SubmitOrderRequest()
OrderRejected
aCustomer:Customer aSalesPerson:SalesPerson
SubmitCreditRequest()
CreditDenied
aCustomer:Customer
FIGURE 6-3
Additional Sample
Instance-Specifi c
Sequence Diagrams
Interaction Diagrams 209
■ Show return values only when they are not obvious. Showing all of the returns tends
to make a sequence diagram more complex and potentially diffi cult to comprehend.
In many cases, less is more. Only show the returns that actually add information for
the reader of the diagram.
■ Justify message names and return values near the arrowhead of the message and
return arrows, respectively. Th is makes it much easier to interpret the messages and
their return values.
Creating a Sequence Diagram In this section, we describe a six-step process used to
create a sequence diagram.6 Th e fi rst step in the process is to determine the context of the
sequence diagram. Th e context of the diagram can be a system, a use case, or a scenario of a
use case. Th e context of the diagram is depicted as a labeled frame around the diagram (see
Figures 6-1, 6-2, and 6-3). Most commonly, it is one use-case scenario. Figure 6-1 portrays
the instance-specifi c sequence diagram for the scenario from the Make Old Patient Appt
use case given in Figure 4-13 for making a new appointment for an existing patient. For
each possible scenario for the Make Old Patient Appt use case, a separate instance-specifi c
sequence diagram would be created. On the surface, this seems to be a lot of potentially
redundant and useless work. However, at this point in the representation of a system, we are
still trying to completely understand the problem. Th is process of creating instance-specifi c
sequence diagrams for each scenario instead of creating a single generic sequence diagram
for the entire use case will enable the developers to attain a more complete understanding
of the problem being addressed. Each instance-specifi c sequence diagram is fairly simple to
interpret, whereas a generic sequence diagram can be very complex. Th e testing of a specifi c
use case is accomplished in a much easier manner by validating and verifying the complete-
ness of the set of instance-specifi c sequence diagrams instead of trying to work through a
single complex generic sequence diagram.
Th e second step is to identify the actors and objects that participate in the sequence being
modeled—i.e., the actors and objects that interact with each other during the use-case scenario.
Th e actors were identifi ed during the creation of the functional model, whereas the objects
are identifi ed during the development of the structural model. Th ese are the classes on which
the objects of the sequence diagram for this scenario will be based. One very useful approach
to identifying all of the scenarios associated with a use case is to role-play the CRC cards (see
Chapter 5). Th is can help you identify potentially missing operations that are necessary to
support the business process, which the use case is representing, in a complete manner. Also,
during role-playing, it is likely that new classes, and hence new objects, will be uncovered.7
Don’t worry too much about identifying all the objects perfectly; remember that the behavioral
modeling process is iterative. Usually, the sequence diagrams are revised multiple times during
the behavioral modeling processes.
Th e third step is to set the lifeline for each object. To do this, you need to draw a vertical dot-
ted line below each class to represent the class’s existence during the sequence. An X should
be placed below the object at the point on the lifeline where the object goes out of existence.
Th e fourth step is to add the messages to the diagram. Th is is done by drawing arrows to
represent the messages being passed from object to object, with the arrow pointing in the
message’s transmission direction. Th e arrows should be placed in order from the fi rst message
6 Th e approach described in this section is adapted from Grady Booch, James Rumbaugh, and Ivar Jacobson, Th e
Unifi ed Modeling Language User Guide (Reading, MA: Addison-Wesley, 1999).
7 Th is obviously will cause you to go back and modify the structural model (see Chapter 5).
1. Set Context
2. Identify Actors
and Objects
4. Add Messages
3. Set Lifeline
210 Chapter 6 Behavioral Modeling
(at the top) to the last (at the bottom) to show time sequence. Any parameters passed along
with the messages should be placed in parentheses next to the message’s name. If a message is
expected to be returned as a response to a message, then the return message is not explicitly
shown on the diagram.
Th e fi fth step is to place the execution occurrence on each object’s lifeline by drawing a narrow
rectangle box over the lifelines to represent when the classes are sending and receiving messages.
Th e sixth and fi nal step is to validate the sequence diagram. Th e purpose of this step is to
guarantee that the sequence diagram completely represents the underlying process. Th is is
done by guaranteeing that the diagram depicts all the steps in the process.8
Campus Housing Example In Chapters 4 and 5, we created a set of functional and structural
models for the campus housing service. In this section, we are going to use those models to
create a sequence diagram for the Add Apartment use case. As stated above, the fi rst thing we
should do is to set the context, which is in this case the Add Apartment use case. Second, we
must identify the actors and objects that will participate in the execution of the use case. To do
this, we should review the functional and structural models that were created for the campus
housing service problem in Chapters 4 and 5. Figure 6-4 replicates these representations.
5. Place Execution
Occurrence
6. Validate
Campus Housing System
Apartment
Owner
Add Apartment
Delete Apartment
* *
*
*
*
*
Student
Search Available
Rental Units
Campus Housing Use-Case
Diagram (Figure 4-15)
8 We describe validation in more detail later in this chapter.
Capture Location
Capture Number of
Bedrooms
Capture Monthly
Rent
Add Apartment
Capture Apartment
Identifier
Delete Apartment
Campus Add and Delete Apartment Activity
Diagrams (Figure 4-16)
FIGURE 6-4
Campus Housing
Service Functional
and Structural
Models
Interaction Diagrams 211
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Add Apartment 1 High
Apartment Owner
Apartment Owner – wants to advertise available apartment
Campus Housing Service – provides a service that enables the apartment owners to rent their available apartments
Use Case Type: Detail, Essential
Apartment Owner wants to add an available apartment
Apartment Owner
1. Capture the location of the apartment.
2. Capture the number of bedrooms in the apartment.
3. Capture the monthly rent of the apartment.
4. Add the apartment to the listing of available apartments.
This use case describes how the campus housing service can maintain an up-to-date listing of
available apartments.
Relationships:
Association:
Include:
Extend:
Generalization:
Campus Housing Service Add an Apartment Use Case Description (Figure 4-17)
Use Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Delete Apartment 2 High
Apartment Owner
Apartment Owner – wants to delist apartment
Campus Housing Service – provides a service that enables the apartment owners to rent their available apartments
Use Case Type: Detail, Essential
Apartment Owner wants to delete an available apartment
Apartment Owner
1. Capture the apartment identifier.
2. Delete the apartment from the listing of available apartments.
This use case describes how the campus housing service can maintain an up-to-date listing of
available apartments.
Relationships:
Association:
Include:
Extend:
Generalization:
Campus Housing Service Delete an Apartment Use Case Description (Figure 4-18)
FIGURE 6-4 Continued
212 Chapter 6 Behavioral Modeling
Based on the functional and structural representations, we see that the actors involved
in the use case are the Apartment Owner and the Campus Housing Service itself. By looking
through the Normal Flow of Events and the activity diagram, we see that the only object
that seems to be relevant is the instance of the Apartment class that is being added. Given
that there are no Alternate/Exceptional Flows or any decisions being made in the Normal
Flow of Events, nor are there any decisions in the activity diagram associated with the Add
Front:
Class Name: Apartment Owner ID: 1
Delete Apartment
Add Apartment
Responsibilities
Associated Use Cases: 2Description: An apartment owner who has apartments for rent
Type: Concrete, Domain
Apartment
Apartment
Collaborators
Back:
Attributes:
Address (address)
Phone number (phone number)
Email (Email address)
Name (string)
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: Apartment
Campus Housing Apartment Owner CRC Card (Figure 5-16)
Apartment Owner Apartment
0..1 0..*
Campus Housing Class Diagram (Figure 5-17)
FIGURE 6-4
Continued
Interaction Diagrams 213
Apartment use case, there is only one scenario to be portrayed. Consequently, there is only
one instance-specifi c sequence diagram to be created. Figure 6-5 depicts the sequence dia-
gram for this use case.
Library Example In the previous chapters, we have demonstrated the diagramming and
modeling processes using the Borrow Books use case of the Library Book Collection Man-
agement System. When considering instance-specifi c scenario diagrams, we need to draw one
sequence diagram per scenario. In the case of the Borrow Books use case in Chapter 4, there
are nine diff erent scenarios. Th erefore, for this one use case, there would be nine separate
diagrams. In this example, we are setting the context of the sequence diagram to only one
specifi c scenario of the Borrow Books use case: Students who have a valid ID and do not have
any overdue books or any fi nes. Th e other scenarios include Students without a valid ID,
Students with a valid ID but who owe fi nes or have overdue books, and the same three sce-
narios for the other two types of Borrowers: Faculty/Staff and Guest. In this example, we are
only drawing the one sequence diagram for the Students with a valid ID scenario. To begin with,
we should review the Flow of Events of the use-case description (see Figure 6-6), the activity
diagram (see Figure 6-7), and the use-case diagram (see Figure 6-8).
Add Apartment()
Apartment Information
Request Apartment Information()
anApartment
Apartment Owner Campus Housing Service
Create(ApartmentInformation)
anApartment
Apartment
FIGURE 6-5
Sequence Diagram
for the Add
Apartment Use Case
FIGURE 6-6 Flow of Events Section of the Use-Case Description of the Borrow Books
Use Case
Normal Flow of Events:
SubFlows:
1. The Borrower brings books to the Librarian at the check out desk.
2. The Borrower provides Librarian their ID card.
3. The Librarian checks the validity of the ID Card.
If the Borrower is a Student Borrower, Validate ID Card against Registrar’s Database.
If the Borrower is a Faculty/Staff Borrower, Validate ID Card against Personnel Database.
If the Borrower is a Guest Borrower, Validate ID Card against Library’s Guest Database.
4. The Librarian checks whether the Borrower has any overdue books and/or fines.
5. The Borrower checks out the books.
Alternate/Exceptional Flows:
4a. The ID Card is invalid, the book request is rejected.
5a. The Borrower either has overdue books, fines, or both, the book request is rejected.
214 Chapter 6 Behavioral Modeling
FIGURE 6-7
Activity Diagram of
the Borrow Books
Use Case (Figure 4-12)
Validate ID Card
Check for Overdue Books and Fines
Check Out Books
[Valid Card]
[No Overdue Books & No Fines]
FIGURE 6-8
Use-Case Diagram
for the Library
Book Collection
Management System
(Figure 4-6)
Process Overdue
Books
Library Book
Collection
Management
System
Maintain Book
Collection
Borrow Books
Search Collection
Return Books
*
* *
*
*
*
*
*
* *
*
*
<>
Personnel Office
<>
Registrar Office
Librarian
*
Borrower
Interaction Diagrams 215
Th e next step is to identify the actors and objects involved in the scenario. By study-
ing the fl ow of events and the use-case diagram, we identify students, librarians, and the
registrar’s database as actors and borrowers, the book collection, and books as the objects.
We place the actors and objects across the top of the diagram based on the ordering of
their appearance in the normal fl ow of events. Th e next step involves simply drawing
the lifelines beneath the actors and objects in the scenario. Th e fourth step is to add the
actual messages to the diagram. To do this, we again review the actual steps taken when
executing this scenario by reviewing the fl ow of events (see Figure 6-6) and the activity
diagram (see Figure 6-7). We also should review any results from the role-playing of the
CRC cards (see Chapter 5). Th is will help us to properly portray where the functionality
is located. For example, in Figure 6-9, the Librarian executes the CheckOutBooks() pro-
cedure (the Student sends the message CheckOutBooks () to ask the Librarian to execute
the CheckOutBooks () procedure) when the student hands the librarian the books to
check out. Th e Librarian in return asks the Student for the ID card. When the student
hands the ID Card to the Librarian, the Librarian asks the Registrar’s Database to exe-
cute the ValidID() procedure when the Librarian passes the student’s ID number over
to the database system to ask the database system to validate the student’s ID number.
Th is continues until the ID Card and Books are returned to the student. Once we have
decided from whom the messages are to be sent and to whom they are sent, we can place
the messages on the diagram. Th e fi fth step then is to add the execution occurrence to
the diagrams to show when each actor or object is in the process of executing one of its
operations. Next, we must validate the diagram. Finally, we should replicate this process
for the other eight scenarios.
FIGURE 6-9
Sequence Diagram
of the Borrow
Books Use Case
for Students with
a Valid ID and No
Overdue Books
or Fines
sd Borrow Books Use Case
:Student
Books
BookAccts
FindBook(BookID)
BookAcct
FindBooks(BookID)
No
ValidID(ID
Number)
Yes
ID
IDCard?()
CheckOutBooks(Books)
:Librarian
Registrar’s Database :Book:Fine Database :BookCollection
Overdue Books or Fines(ID)
216 Chapter 6 Behavioral Modeling
Communication Diagrams
Communication diagrams, like sequence diagrams, essentially provide a view of the
dynamic aspects of an object-oriented system. Th ey can show how the members of a set of
objects collaborate to implement a use case or a use-case scenario. Th ey can also be used to
model all the interactions among a set of collaborating objects, in other words, a collabora-
tion (see CRC cards in Chapter 5). In this case, a communication diagram can portray how
dependent the diff erent objects are on one another.9 A communication diagram is essen-
tially an object diagram that shows message-passing relationships instead of aggregation
or generalization associations. Communication diagrams are very useful to show process
patterns (i.e., patterns of activity that occur over a set of collaborating classes).
Communication diagrams are equivalent to sequence diagrams, but they emphasize the
fl ow of messages through a set of objects, whereas the sequence diagrams focus on the time
ordering of the messages being passed. Th erefore, to understand the fl ow of control over a
set of collaborating objects or to understand which objects collaborate to support business
processes, a communication diagram can be used. For time ordering of the messages, a
sequence diagram should be used. In some cases, both can be used to more fully understand
the dynamic activity of the system.
Elements of a Communication Diagram Figure 6-10 shows a communication diagram for
the Make Old Patient Appt use case. Like the sequence diagram in Figure 6-1, the Make Old
Patient Appt process is portrayed.
Actors and objects that collaborate to execute the use case are placed on the commu-
nication diagram in a manner to emphasize the message passing that takes place between
them. Notice that the actors and objects in Figure 6-10 are the same ones in Figure 6-1:
aPatient, aReceptionist, aPatient, UnpaidBill, and Appointment.10 Again, as with the
sequence diagram, for each of the objects, the name of the class of which they are an
instance is given after the object’s name (e.g., aPatient: Patient). (The communication
9 We return to this idea of dependency in Chapters 7 and 8.
10 In some versions of the communication diagram, object symbols are used as surrogates for the actors. However,
again we recommend using actor symbols for actors instead.
FIGURE 6-10 Sample Communication Diagram
sd Make Appt Use Case
aPatient
1: RequestAppt(name, address)
4: NewCancelChangeAppt?
5: ApptTimes?
aReceptionist
2: LookUpPatient()
3: [aPatient Exists] LookupBills()
7: CreateAppt()
6: MatchAppts()
:Appointment
aPatient:Patient
:UnpaidBill
Interaction Diagrams 217
diagram syntax is given in Figure 6-11.) Unlike the sequence diagram, the communica-
tion diagram does not have a means to explicitly show an object being deleted or created.
It is assumed that when a delete, destroy, or remove message is sent to an object, it will
go out of existence, and a create or new message will cause a new object to come into
existence. Another difference between the two interaction diagrams is that the communi-
cation diagram never shows returns from message sends, whereas the sequence diagram
can optionally show them.
An association is shown between actors and objects with an undirected line. For exam-
ple, an association is shown between the aPatient and aReceptionist actors. Messages are
shown as labels on the associations. Included with the labels are lines with arrows showing
the direction of the message being sent. For example, in Figure 6-10, the aPatient actor
sends the RequestAppt() message to the aReceptionist actor, and the aReceptionist actor
FIGURE 6-11 Communication Diagram Syntax
An actor:
■ Is a person or system that derives benefit from and is external to the system.
■ Participates in a collaboration by sending and/or receiving messages.
An object:
■ Participates in a collaboration by sending and/or receiving messages.
An association:
■ Shows an association between actors and/or objects.
■ Is used to send messages.
A message:
■ Conveys information from one object to another one.
■ Has direction shown using an arrowhead.
■ Has sequence shown by a sequence number.
A frame:
■ Indicates the context of the communication diagram.
A guard condition:
■ Represents a test that must be met for the message to be sent.
<>
anActor
anActor
Context
SeqNumber: aMessage
SeqNumber: [aGuardCondition]: aMessage
Term and Definition Symbol
■ Is depicted either as a stick figure (default) or, if a nonhuman actor is involved,
as a rectangle with <> in it (alternative).
anObject : aClass
218 Chapter 6 Behavioral Modeling
sends the NewCancelChangeAppt?() and the ApptTimes?() messages to the aPatient actor.
Th e sequence of the message sends is designated with a sequence number. In Figure 6-10, the
RequestAppt() message is the fi rst message sent, whereas the NewCancelChangeAppt?() and
the ApptTimes?() messages are the fourth and fi ft h message sent, respectively.
Like the sequence diagram, the communication diagram can represent conditional mes-
sages. For example, in Figure 6-10, the LookupBills() message is sent only if the [aPa-
tient exists] condition is met. If a message is repeatedly sent, an asterisk is placed aft er the
sequence number. Finally, an association that loops onto an object shows self-delegation.
Th e message is shown as the label of the association.
When a communication diagram is fully populated with all the objects, it can become
very complex and diffi cult to understand. When this occurs, it is necessary to simplify the
diagram. One approach to simplifying a communication diagram, like use-case diagrams
(see Chapter 4) and class diagrams (see Chapter 5), is through the use of packages (i.e.,
logical groups of classes). In the case of communication diagrams, its objects are grouped
together based on the messages sent to and received from the other objects.11
Figure 6-12 provides two additional examples of communication diagrams. Th ese dia-
grams are equivalent to the sequence diagrams contained in Figure 6-3. However, when
comparing the communication diagrams to the sequence diagrams in these fi gures, you see
that quite a bit of information is lost. For example, the CreateSandwich() message is nowhere
to be found. However, the primary purpose of the communication diagram is to show how
the diff erent actors and classes interact, and this is exactly the information that is included.
Guidelines for Creating Communication Diagrams Ambler12 provides a set of guidelines
when drawing communication diagrams. In this section, in addition to the fi rst four guide-
lines for drawing sequence diagrams, we consider two more.
■ Use the correct diagram for the information you are interested in communicating
with the user. Communication diagrams allow the team to easily identify a set of
objects that are intertwined. Do not use communication diagrams to model process
fl ow. Instead, you should use an activity diagram with swimlanes that represent
11 For those familiar with structured analysis and design, packages serve a purpose similar to the leveling and bal-
ancing processes used in data fl ow diagramming. Packages and package diagrams are described in Chapter 7.
12 S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, England: Cambridge University Press, 2005).
FIGURE 6-12
Additional Sample
Communication
Diagrams
sd Make Lunch Use Case
2: CreateLunch
3: GetSandwich1: MakeLunch
aChild:Child firstParent:Parent secondParent:Parent
sd Submit Order Use Case
aCustomer:Customer
1: SubmitOrderRequest() 2: SubmitCreditRequest()
aCustomer:Customer aSalesPerson:SalesPerson
Interaction Diagrams 219
objects (see Chapter 4). On the other hand, it would be very diffi cult to “see” how the
objects collaborated in an activity diagram.
■ When trying to understand the sequencing of messages, a sequence diagram should
be used instead of a communication diagram. As in the previous guideline, this guide-
line essentially suggests that you should use the diagram that was designed to deal
with the issue at hand. Even though communication diagrams can show sequencing of
messages, this was never meant to be their primary purpose.
Creating a Communication Diagram13 Remember that a communication diagram is
basically an object diagram that shows message-passing relationships instead of aggregation
or generalization associations. In this section, we describe a fi ve-step process used to build a
communication diagram. Th e fi rst step in the process is to determine the context of the com-
munication diagram. Like a sequence diagram, the context of the diagram can be a system, a
use case, or a scenario of a use case. Th e context of the diagram is depicted as a labeled frame
around the diagram (see Figures 6-10, 6-11, and 6-12).
Th e second step is to identify the objects (actors) and the associations that link the objects
(actors) that participate in the collaboration together. Remember, the objects that partici-
pate in the collaboration are instances of the classes identifi ed during the development of
the structural model (see Chapter 5). Like the sequence-diagramming process, it is likely
that additional objects, and hence classes, will be discovered. Again, this is normal because
the underlying development process is iterative and incremental. In addition to the com-
munication diagram being modifi ed, the sequence diagrams and structural model probably
also have to be modifi ed. Additional functional requirements might also be uncovered,
hence requiring the functional models to be modifi ed as well (see Chapter 4).
Th e third step is to lay out the objects (actors) and their associations on the communication
diagram by placing them together based on the associations that they have with the other
objects in the collaboration. By focusing on the associations between the objects (actors)
and minimizing the number of associations that cross over one another, we can increase the
understandability of the diagram.
Th e fourth step is to add the messages to the associations between the objects. We do this
by adding the name of the message(s) to the association link between the objects and an
arrow showing the direction of the message being sent. Each message has a sequence num-
ber associated with it to portray the time-based ordering of the message.14
Th e fi fth and fi nal step is to validate the communication diagram. Th e purpose of this step
is to guarantee that the communication diagram faithfully portrays the underlying pro-
cess(es). Th is is done by ensuring that all steps in the process are depicted on the diagram.
Campus Housing Example As with the sequence diagram example, we return to the Add
Apartment use case for the Campus Housing Service. To begin with, we again set the
context for the communication diagram (the Add Apartment use case). Next, we identify
the objects (Apartment), actors (Apartment Owner and Campus Housing Service), and
13 Th e approach described in this section is adapted from Booch, Rumbaugh, and Jacobson, Th e Unifi ed Modeling
Language User Guide.
14 However, remember the sequence diagram portrays the time-based ordering of the messages in a top-down
manner. If your focus is on the time-based ordering of the messages, we recommend that you also use the sequence
diagram.
1. Set Context
3. Lay Out Diagram
4. Add Messages
2. Identify Objects,
Actors, &
Associations
5. Validate
220 Chapter 6 Behavioral Modeling
associations (links between the Apartment Owner actor and the Campus Housing Service
actor and links between the Campus Housing Service actor and the Apartment object).
Using this information, we lay out the diagram showing the actors, objects, and associ-
ations between them. Finally, we label the associations with the appropriate messages.
Figure 6-13 depicts the communication diagram for this use case.
Library Example As with the sequence diagramming example, we return to the Borrow
Books use case of the Library Book Collection Management System. In this case, to set the
context of the diagram, we visit the Student without a valid ID and Student with a Valid ID
but owes fi nes or has overdue books scenarios. We create two communication diagrams, one
for each scenario. As with the sequence-diagramming process, we review the Flow of Events
of the use-case description (see Figure 6-6), the activity diagram (see Figure 6-7), and the use
case diagram (see Figure 6-8).
Th e next step is to identify the actor, objects, and associations involved in the scenario.
In both scenarios, the actors are Student, Librarian, and the Registrar’s Database. However,
because the process is aborted very early in the Student without a valid ID scenario, there are
no objects in the scenario. Th e Student with a Valid ID but owes fi nes or has overdue books
scenario does include one object: Borrower. Both scenarios have an association between the
Student and Librarian actors and the Librarian and Registrar’s Database actor. Th e Student
with a Valid ID but owes fi nes or has overdue books scenario also has an association between
the Librarian actor and the Borrower object.
Th e next step is to lay out the diagram. In both cases, because the student initiates the
process, we place the Student actor to the far left of the diagram. We then place the other
actors on the diagram in the order in which they participate in the process. We also place
the :Borrower object to the far bottom right of the diagram that represents the Student
with a Valid ID but owes fi nes or has overdue books scenario to refl ect the left -to-right and
top-to-bottom direction of reading for most Western cultures.
Now we place the relevant associations between the actors and objects that participate in
the scenarios. In this step, we add the messages to the associations. We again review the fl ow
of events (see Figure 6-6) of the use-case description to identify the directionality and con-
tent of the messages. Figures 6-14 and 6-15 portray the communication diagrams created.
FIGURE 6-13
Communication
Diagram for the
Add Apartment
Use Case
Apartment
1: Add Apartment
2: RequestApartmentinformation 3: Create(ApartmentInformation)
Apartment Owner Campus Housing Service
FIGURE 6-14
Communication
Diagram for the
Student without a
Valid ID
sd Borrow Books Use Case
:Student
1: Checkout Books (Books)
2: IDCard?()
:Librarian
3: ValidID(IDNumber)
Registrar’s Database
Behavioral State Machines 221
Th e last step is to validate the diagrams. As with sequence diagrams, because we are
drawing instance specifi c versions of the communication diagram, we must also draw the
remaining seven diagrams for the other scenarios.
BEHAVIORAL STATE MACHINES
Some of the classes in the class diagrams represent a set of objects that are quite dynamic in
that they pass through a variety of states over the course of their existence. For example, a
patient can change over time from being new to current to former based on his or her status
with the doctor’s offi ce. A behavioral state machine is a dynamic model that shows the diff er-
ent states through which a single object passes during its life in response to events, along with
its responses and actions. Typically, behavioral state machines are not used for all objects;
rather, behavioral state machines are used with complex objects to further defi ne them and to
help simplify the design of algorithms for their methods. Th e behavioral state machine shows
the diff erent states of the object and what events cause the object to change from one state to
another. Behavioral state machines should be used to help understand the dynamic aspects of
a single class and how its instances evolve over time15 unlike interaction diagrams that show
how a particular use case or use-case scenario is executed over a set of classes.
In this section, we describe states, events, transitions, actions, and activities. We also
explain how behavioral state machines model the state changes through which complex
objects pass. As with interaction diagrams, when we create a behavioral state machine for
an object, it is possible that we will uncover additional events that need to be included in
the functional model (see Chapter 4) and additional operations that need to be included in
the structural model (see Chapter 5), so our interaction diagrams might have to be modifi ed
again. Because object-oriented development is iterative and incremental, this continuous
modifi cation of the evolving models (functional, structural, and behavioral) of the system is
to be expected.
States, Events, Transitions, Actions, and Activities
Th e state of an object is defi ned by the value of its attributes and its relationships with other
objects at a particular point in time. For example, a patient might have a state of new, current,
or former. Th e attributes or properties of an object aff ect the state that it is in; however, not
15 Some authors refer to this as modeling an object’s life cycle.
FIGURE 6-15
Communication
Diagram for the
Student with a Valid
ID but Owes Fines or
Has Overdue Books
sd Borrow Books Use Case
:Student
1: Checkout Books (Books)
2: IDCard?()
:Librarian
3: ValidID(IDNumber)
4: Overdue Books or Fines()
:Borrower
Registrar’s Database
222 Chapter 6 Behavioral Modeling
all attributes or attribute changes will make a diff erence. For example, think about a patient’s
address. Th ose attributes make very little diff erence to changes in a patient’s state. However, if
states were based on a patient’s geographic location (e.g., in-town patients were treated diff er-
ently than out-of-town patients), changes to the patient’s address would infl uence state changes.
An event is something that takes place at a certain point in time and changes a value or
values that describe an object, which, in turn, changes the object’s state. It can be a des-
ignated condition becoming true, the receipt of the call for a method by an object, or the
passage of a designated period of time. Th e state of the object determines exactly what the
response will be.
A transition is a relationship that represents the movement of an object from one state
to another state. Some transitions have a guard condition. A guard condition is a Boolean
expression that includes attribute values, which allows a transition to occur only if the condi-
tion is true. An object typically moves from one state to another based on the outcome of an
action triggered by an event. An action is an atomic, nondecomposable process that cannot
be interrupted. From a practical perspective, actions take zero time, and they are associated
with a transition. In contrast, an activity is a nonatomic, decomposable process that can be
interrupted. Activities take a long period of time to complete, and they can be started and
stopped by an action.
Elements of a Behavioral State Machine
Figure 6-16 presents an example of a behavioral state machine representing the patient class
in the context of a hospital environment. From this diagram, we can tell that a patient enters a
hospital and is admitted aft er checking in. If a doctor fi nds the patient to be healthy, he or she
is released and is no longer considered a patient aft er two weeks elapse. If a patient is found
to be unhealthy, he or she remains under observation until the diagnosis changes.
A state is a set of values that describes an object at a specifi c point in time and represents
a point in an object’s life in which it satisfi es some condition, performs some action, or waits
for something to happen (see Figure 6-17). In Figure 6-16 states include entering, admitted,
released, and under observation. A state is depicted by a state symbol, which is a rectangle
with rounded corners with a descriptive label that communicates a particular state. Th ere are
two exceptions. An initial state is shown using a small, fi lled-in circle, and an object’s fi nal
state is shown as a circle surrounding a small, fi lled-in circle. Th ese exceptions depict when
an object begins and ceases to exist, respectively.
FIGURE 6-16 Sample Behavioral State Machine Diagram
Patient
Enters Hospital Checks In [Diagnosis = Healthy] [> 2 weeks]
Entering Admitted Released
Under Observation
[Diagnosis = Unhealthy]
[Diagnosis = Healthy]
Behavioral State Machines 223
FIGURE 6-17 Behavioral State Machine Diagram Syntax
A state:
■ Is shown as a rectangle with rounded corners.
■ Has a name that represents the state of an object.
An initial state:
■ Is shown as a small, filled-in circle.
■ Represents the point at which an object begins to exist.
A final state:
■ Is shown as a circle surrounding a small, filled-in circle (bull’s-eye).
■ Represents the completion of activity.
An event:
■ Is a noteworthy occurrence that triggers a change in state.
■ Can be a designated condition becoming true, the receipt of an explicit signal
from one object to another, or the passage of a designated period of time.
■ Is used to label a transition.
A transition:
■ Indicates that an object in the first state will enter the second state.
■ Is triggered by the occurrence of the event labeling the transition.
■ Is shown as a solid arrow from one state to another, labeled by the event name.
A frame:
■ Indicates the context of the behavioral state machine.
Context
anEvent
aState
Term and Definition Symbol
Arrows are used to connect the state symbols, representing the transitions between
states. Each arrow is labeled with the appropriate event name and any parameters or condi-
tions that may apply. For example, the two transitions from admitted to released and under
observation contain guard conditions. As in the other behavioral diagrams, in many cases it
is useful to explicitly show the context of the behavioral state machine using a frame.
Figure 6-18 depicts two additional behavioral state machines. Th e fi rst one is for the
lunch object that was associated with the Make Lunch use-case scenario of Figures 6-3
and 6-12. In this case, there is obviously additional information that has been captured
about the lunch object. For example, the scenario of Figures 6-3 and 6-12 did not include
information regarding the lunch being taken out of the box or being eaten. Th is implies
additional use cases and/or use-case scenarios that would have to be included in a system
dealing with lunch processing. Th e second behavioral state machine deals with the life cycle
of an order. Th e order object is associated with the submit order use-case scenario described
in Figures 6-3 and 6-12. As in the lunch example, there is quite a bit of additional informa-
tion contained in this behavioral state machine. For an order-processing system, additional
224
FI
G
U
R
E
6
-1
8
A
d
d
it
io
n
al
B
eh
av
io
ra
l S
ta
te
M
ac
h
in
e
D
ia
gr
am
s
O
rd
er
D
en
ie
d
O
rd
er
is
cr
ea
te
d
C
us
to
m
er
su
bm
its
o
rd
er
C
us
to
m
er
ed
its
o
rd
er
in
fo
rm
at
io
n
A
ut
ho
ri
za
tio
n
=
D
en
ie
d
O
rd
er
s
en
t
fo
r
cr
ed
it
au
th
or
iz
at
io
n
C
us
to
m
er
w
ith
dr
aw
s
or
de
r
re
qu
es
t
O
rd
er
s
en
t t
o
cu
st
om
er
C
us
to
m
er
ac
ce
pt
s
sh
ip
m
en
t
R
ec
ei
ve
d
A
ut
ho
ri
za
tio
n
=
A
pp
ro
ve
d
O
rd
er
is
cl
os
ed
In
P
ro
ce
ss
O
rd
er
ed
Pr
oc
es
si
ng
Pl
ac
ed
Sh
ip
pe
d
Lu
nc
h
[C
re
at
ed
]
[P
la
ce
dI
nB
ox
]
[T
ak
en
O
ut
O
fB
ox
]
[E
at
en
]
C
re
at
ed
Pa
ck
ed
B
ei
ng
E
at
en
Behavioral State Machines 225
sequence and communication diagrams would be necessary to completely represent all the
processing associated with an order object. Obviously, because behavioral state machines
can uncover additional processing requirements, they can be very useful in fi lling out the
complete description of an evolving system.
Sometimes, states and subclasses can be confused. For example, in Figure 6-19, are the
classes Freshman, Sophomore, Junior, and Senior subclasses of the class Undergraduate or
are they states that an instance of the Undergraduate class goes through during its lifetime?
In this case, the latter is the better answer. When trying to identify all potential classes
Graduate
Student
DoctoralMasters
Freshman
VS.
&
[Accepted]
SophomoreFreshman
Undergraduate
Undergraduate
SeniorJunior DoctoralMasters
Graduate
Student
Sophomore
[>30 Hours Earned]
Junior
[>60 Hours Earned]
Senior
[>90 Hours Earned]
[Graduate]
FIGURE 6-19 States versus Subclasses
226 Chapter 6 Behavioral Modeling
during structural modeling (see Chapter 5), you might actually identify states of the rele-
vant superclass instead of subclasses. Th is is another example of how tightly intertwined
the functional, structural, and behavioral models can be. From a modeling perspective,
although we eventually removed the Freshman, Sophomore, Junior, and Senior subclasses
from the structural model, capturing that information during structural modeling and
removing it based on discoveries made during behavioral modeling were preferable to
omitting it and taking a chance of missing a crucial piece of information about the problem
domain. Remember, object-oriented development is iterative and incremental. As we pro-
gress to a correct model of the problem domain, we will make many mistakes.
Guidelines for Creating Behavioral State Machines As with the sequence and communi-
cation diagrams, Amble suggests a set of guidelines when drawing behavior state machines.
In this case, we consider six of his recommendations.16
■ Create a behavioral state machine for objects whose behavior changes based on the state
of the object. In other words, do not create a behavioral state machine for an object
whose behavior is always the same regardless of its state. Th ese objects are too simple.
■ To adhere to the left -to-right and top-to-bottom reading conventions of Western
cultures, the initial state should be drawn in the top left corner of the diagram and
the fi nal state should be drawn in the bottom right of the diagram.
■ Make sure that the names of the states are simple, intuitively obvious, and descrip-
tive. For example in Figure 6-16, the state names of the patient object are Entering,
Admitted, Under Observation, and Released.
■ Question black hole and miracle states. These types of states are problematic
for the same reason black hole and miracle activities are a problem for activity
diagrams (see Chapter 4). Black hole states, states that an object goes into
and never comes out of, most likely are actually fi nal states. Miracle states, states
that an object comes out of but never went into, most likely are initial states.
■ Be sure that all guard conditions are mutually exclusive (not overlapping). For
example, in Figure 6-16, the guard condition [Diagnosis = Healthy] and the guard
condition [Diagnosis = Unhealthy] do not overlap. However, if you created a guard
condition of [x >= 0] and a second guard condition [x <= 0], the guard conditions
overlap when x = 0, and it is not clear to which state the object would transition.
Th is would obviously cause confusion.
■ All transitions should be associated with a message and operation. Otherwise,
the state of the object could never change. Even though this may be stating the obvi-
ous, there have been numerous times that analysts forget to go back and ensure that
this is indeed true.
Creating a Behavioral State Machine
Behavioral state machines are drawn to depict an instance of a single class from a class dia-
gram. Typically, the classes are very dynamic and complex, requiring a good understanding of
their states over time and events triggering changes. You should examine your class diagram
to identify which classes undergo a complex series of state changes and draw a diagram for
each of them. In this section, we describe a fi ve-step process used to build a behavioral state
machine.17 Like the other behavioral models, the fi rst step in the process is determining the
16 S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, England: Cambridge University Press, 2005).
17 Th e approach described in this section is adapted from Booch, Rumbaugh, and Jacobson, Th e Unifi ed Modeling
Language User Guide.
1. Set Context
Behavioral State Machines 227
context of the behavioral state machine, which is shown in the label of the frame of the dia-
gram. Th e context of a behavioral state machine is usually a class. However, it also could be a
set of classes, a subsystem, or an entire system.
Th e second step is to identify the various states that an object will have over its lifetime. Th is
includes establishing the boundaries of the existence of an object by identifying the initial
and fi nal states of an object. We also must identify the states of an object. Th e information
necessary to perform this is gleaned from reading the use-case descriptions, talking with
users, and relying on the requirements-gathering techniques that you learned about in
Chapter 3. An easy way to identify the states of an object is to write the steps of what hap-
pens to an object over time, from start to fi nish, similar to how the normal fl ow of events
section of a use-case description would be created.
Th e third step is to determine the sequence of the states that an object will pass through dur-
ing its lifetime. Using this sequence, the states are placed onto the behavioral state machine
in a left -to-right order.
Th e fourth step is to identify the transitions between the states of the objects and to add the
events, actions, and guard conditions associated with the transitions. Th e events are the trig-
gers that cause an object to move from one state to the next state. In other words, an event
causes an action to execute that changes the value(s) of an object’s attribute(s) in a signifi cant
manner. Th e actions are typically operations contained within the object. Also, guard condi-
tions can model a set of test conditions that must be met for the transition to occur. At this
point in the process, the transitions are drawn between the relevant states and labeled with
the event, action, or guard condition.
Th e fi fth step is to validate the behavioral state machine by making sure that each state is
reachable and that it is possible to leave all states except for fi nal states. Obviously, if an iden-
tifi ed state is not reachable, either a transition is missing or the state was identifi ed in error.
Only fi nal states can be a dead end from the perspective of an object’s life cycle.
Campus Housing Example Based on the functional and structural models for the campus
housing service (see Figure 6-4), the sequence diagram for the Add Apartment use case (see
Figure 6-5), and the communication diagram for the Add Apartment use case (see Figure
6-13), in this section, we are going to create a behavioral state machine for the Apartment
class. By reviewing all of the representations, it is obvious that the behavioral state machine
will be very simple. In this case, an apartment comes into existence when it is added and goes
out of existence when it is deleted. Its only state is For Rent. Figure 6-20 depicts the behavioral
state machine for this class.
2. Identify Object
States
3. Lay Out Diagram
4. Add Transitions
5. Validate
FIGURE 6-20
Behavioral State
Machine for the
Apartment Class
[Add] [Delete]
For Rent
228 Chapter 6 Behavioral Modeling
FIGURE 6-21 Class Diagram for the Library Book Collection Management System
Check Out Trans
0..*
1..1
Borrower
1..1
0..*
Check Out Desk
Transaction Line Item Book Book Location
0..* 1..11..1 1..* 0..* 1..1
*
1
GuestStudent Faculty/Staff
Librarian
StorageInterlibrary Loan System Library
Book Collection
Library Example Th e fi rst step in drawing a behavioral state machine is to set the context. For
our purposes, the context typically is an instance of a class that has multiple states and whose
behavior depends upon the state in which it currently resides. As suggested earlier, we should
review the class diagram (see Figure 6-21) to identify the “interesting” classes. In the case of the
Library Book Collection Management System, the obvious class to consider is the Book class.
Th e next step is to identify the diff erent states through which an instance of the Book class
can traverse during its lifetime. Good places to look for possible state changes are the use-case
descriptions (see Figure 6-6), the activity diagrams (see Figure 6-7), the sequence diagrams
(see Figure 6-9), and the communication diagrams (see Figures 6-14 and 6-15). In the case of a
book, even though the states may be similar, you must be careful in identifying the states asso-
ciated with an instance of the Book class and not the states associated with the physical book
itself. In Chapter 5, we observed that there were a number of implied states to consider. Th ese
included Checked Out, Overdue, Requested, Available, and Damaged. If the book is damaged,
the book could either be repaired and put back into circulation or it could be too damaged to
repair and be removed from circulation instead. Even though a Borrower could be fi ned for
an overdue or damaged book, being fi ned is not a state of a book, it is a state of a borrower.
Next, we lay out the diagram by ordering the states in a sequential manner based on
the life cycle of a book. For example, it probably makes no sense to have a book to go from
a repaired state to a damaged state. However, going from a damaged state to a repaired state
makes sense. Nor does it make sense for a book to go from an available state directly to an
overdue state. However, the converse makes sense. Th e states we identifi ed for a book object
include Available, Checked Out, Overdue, Requested, Damaged, and Being Repaired. Next
we added the transitions between the states and labeled them with the appropriate guard
Crude Analysis 229
conditions. Th e behavioral state machine for an instance of the Book class is portrayed in
Figure 6-22.
Finally, we validate the diagram by checking for missing states or transitions and ensur-
ing that there are no black hole or miracle states.
CRUDE ANALYSIS
One useful technique to identify how the underlying objects in the problem domain work
together to collaborate in support of the use cases is CRUDE analysis.18 CRUDE analysis uses
a CRUDE matrix , in which each interaction among objects is labeled with a letter for the
type of interaction: C for create, R for read or reference, U for update, D for delete, and E for
execute. In an object-oriented approach, a class/actor-by-class/actor matrix is used.19 Each
cell in the matrix represents the interaction between instances of the classes. For example,
in Figure 6-1, an instance of the Receptionist actor creates an instance of the Appointment
class. Assuming a Row:Column ordering, a C is placed in the cell Receptionist:Appointment.
Also, in Figure 6-1, an instance of the Receptionist actor references an instance of the
Appointments class. In this case, an R is placed in the Receptionist:Appointments cell.
Figure 6-23 shows the CRUDE matrix based on the Make Old Patient Appt use case.
Unlike the interaction diagrams and behavioral state machines, a CRUDE matrix is most
useful as a system-wide representation. Once a CRUDE matrix is completed for the entire
FIGURE 6-22 Behavioral State Machine for an Instance of the Book Class in the Library Book
Collection Management System
Book
Damaged
[Added To
Collection]
[Borrower Checks in Book]
[Borrower Checks in Book]
[Borrower Checks in Book] [Another Borrower Requests Book]
[Borrowing Time
Expires]
[Borrower Checks
Out Book]
[Borrower Returns Book Damaged]
[Book Sent to be Repaired]
[Book Too Damaged]
[Book Repaired]
[Book Too Damaged][Book Taken Out of Circulation]
[Borrower Returns Book Damaged]
[Borrower Returns Book Damaged]
[Another Borrower
Requests Book]
Checked Out Overdue
Being Repaired
Available Requested
18 CRUD analysis has typically been associated with structured analysis and design [see Alan Dennis, Barbara Haley
Wixom and Roberta M. Roth, Systems Analysis Design, 3nd ed. (New York: Wiley, 2006)] and information engi-
neering [see James Martin, Information Engineering, Book II Planning and Analysis (Englewood Cliff s, NJ: Prentice
Hall, 1990)]. In our case, we have simply adapted it to object-oriented systems development. In the case of object
orientation, we have added an E to allow us to document the execution of operations that do not create, read, update,
or delete but that instead simply are executed for possible side-eff ect purposes. Specifi c details on collaborations are
described in Chapter 7.
19 Another useful but more-detailed form of the CRUDE matrix is a Class/Actor:Operation-by-Class/Actor:Oper-
ation matrix. For validation and verifi cation purposes, this more-detailed matrix is more useful. However, for our
purposes at this point in our discussion, the Class/Actor-by-Class/Actor matrix is suffi cient.
230 Chapter 6 Behavioral Modeling
FIGURE 6-23 CRUDE Matrix for the Make Old Patient Apt Use Case
Receptionist PatientList Patient UnpaidBills Appointments Appointment
Receptionist RU CRUD R RU CRUD
PatientList R
Patient
UnpaidBills
Appointments R
Appointment
FIGURE 6-24
Campus Housing
Service CRUDE
Matrix
Apartment
Owner Actor
Student
Actor
Apartment
Owner Class
Apartment
Class
Apartment Owner Actor C,D
Student Actor R R
Apartment Owner Class
Apartment Class
system, the matrix can be scanned quickly to ensure that every class can be instantiated. Each
type of interaction can be validated for each class. For example, if a class represents only tem-
porary objects, then the column in the matrix should have a D in it somewhere. Otherwise,
the instances of the class will never be deleted. Because a data warehouse contains historical
data, objects that are to be stored in one should not have any U or D entries in their associated
columns. In this way, CRUDE analysis can be used as a way to partially validate the interac-
tions among the objects in an object-oriented system. Finally, the more interactions among a
set of classes, the more likely they should be clustered together in a collaboration. However,
the number and type of interactions are only an estimate at this point in the development
of the system. Care should be taken when using this technique to cluster classes to identify
collaborations. We return to this subject in the next chapter when we deal with partitions
and collaborations.
CRUDE analysis also can be used to identify complex objects. Th e more (C)reate, (U)pdate,
or (D)elete entries in the column associated with a class, the more likely the instances of the
class have a complex life cycle. As such, these objects are candidates for state modeling with
a behavioral state machine.
Campus Housing Example In Chapters 4 and 5, we created a set of functional and structural
models for the campus housing service. In this section, we are going to use those models as a
basis for performing a CRUDE analysis. Th e fi rst thing we need to do is to identify all of the
actors and the classes that are involved in the campus housing service example. In this case,
the actors are apartment owner and student, and the classes are apartment owner and apart-
ment. Given this, our CRUDE matrix is a 4×4 matrix. In this simple example, we only support
creating, reading, and deleting instances. Specifi cally, an apartment owner actor can create
and delete instances of apartment, while a student actor can read instances of apartment and
apartment owner. Figure 6-24 depicts the CRUDE matrix for the campus housing service.
Crude Analysis 231
FIGURE 6-25 Corrected Campus Housing Service CRUDE Matrix
Apartment
Owner Actor
Student
Actor
Staff Member
Actor
Apartment
Owner Class
Apartment
Class
Apartment Owner Actor C,D
Student Actor R R
Staff Member Actor C,D
Apartment Owner Class
Apartment Class
However, upon review of the matrix, even though instances of apartment owner are
read, they are never created or deleted. Unless the instances of apartment owner are cre-
ated with another system, this is an impossible situation. Th is is another example of why
we follow an iterative and incremental approach in object-oriented systems development.
In this case, by creating a CRUDE matrix, we discovered an additional requirement that
had previously been overlooked. Consequently, we need to go back and add additional use
cases that add and delete apartment owners that are associated with an additional campus
housing service staff member actor that executes them (see Figure 6-25). Obviously, at this
point in time we should modify the use-case diagram; add activity diagrams; add sequence
diagrams; add communication diagrams; and review the class diagrams, CRC cards, and
behavioral state machines to ensure that they are still correct. We will leave those modifi ca-
tions to you and move on next to the library problem that we have been using in this and
the previous chapters.
Library Example Th e best way to create a CRUDE matrix is to conceptually merge the
sequence and communication diagrams that model all of the scenarios of all of the use cases
in a system. Th e easiest way to accomplish this is simply to create an empty class/actor-
by-class/actor matrix. In the case of the Library Book Collection Management System, we
have six actors (Student, Faculty/Staff , Guest, Librarian, Personnel Offi ce, and Registrar’s
Offi ce) and eight classes (Book, Book Collection, Student, Faculty/Staff , Guest, Interlibrary
Loan System, Library, and Storage). Once this matrix has been laid out, role-playing the
scenarios will show which actors and classes interact with each other. Based on the type of
interaction, record a C, R, U, D, or E in the appropriate cell of the matrix. Do this repeat-
edly until all of the scenarios of all of the use cases have been executed. Th e CRUDE matrix
for the Library Book Collection Management System is shown in Figure 6-26. One of the
functions that the matrix can serve is to begin the validation process of the entire system.
In this case, by quickly reviewing the matrix we can see that absolutely nothing seems to
be interacting with the Library and Storage objects. Th is raises an important question as
to whether these objects should exist or not. If nothing calls or uses them and they don’t
call or use anything, then why are they part of this system? Either they should be removed
from the current representation of the system, or we have managed to miss some interac-
tion. Knowing this allows us to go back to the user, in this case the Librarian, and ask what
should be done.
St
ud
en
t
A
ct
or
Fa
cu
lt
y/
St
af
f
A
ct
or
G
ue
st
A
ct
or
L
ib
ra
ri
an
A
ct
or
Pe
rs
on
ne
l
O
ffi
c
e
A
ct
or
R
eg
is
tr
ar
’s
O
ffi
c
e
A
ct
or
B
oo
k
B
oo
k
C
ol
le
ct
io
n
St
ud
en
t
C
la
ss
Fa
cu
lt
y/
St
af
f
C
la
ss
G
ue
st
C
la
ss
In
te
rl
ib
ra
ry
Lo
an
S
ys
te
m
Li
br
ar
y
St
or
ag
e
St
ud
en
t A
ct
or
E
R
,E
R
E
Fa
cu
lty
/S
ta
ff
A
ct
or
E
R
,E
R
E
G
ue
st
A
ct
or
E
R
,E
R
E
Li
br
ar
ia
n
A
ct
or
E
E
E
R
,E
R
,E
C
,R
,U
,D
,E
R
,U
,E
R
,U
R
,U
C
,R
,U
,D
,E
R
,E
Pe
rs
on
ne
l
O
ffi
ce
A
ct
or
R
eg
is
tr
ar
’s
O
ffi
ce
A
ct
or
B
oo
k
B
oo
k
C
ol
le
ct
io
n
St
ud
en
t C
la
ss
Fa
cu
lty
/S
ta
ff
C
la
ss
G
ue
st
C
la
ss
In
te
rl
ib
ra
ry
Lo
an
S
ys
te
m
Li
br
ar
y
St
or
ag
e
FI
G
U
R
E
6
-2
6
C
R
U
D
E
M
at
ri
x
fo
r
th
e
Li
b
ra
ry
B
o
o
k
C
o
lle
ct
io
n
M
an
ag
em
en
t
Sy
st
em
232
Verifying and Validating the Behavioral Model 233
VERIFYING AND VALIDATING THE BEHAVIORAL MODEL20
In this chapter, we described three diff erent diagrams (sequence diagram, communication
diagram, and behavioral state machine) and CRUDE matrices that could be used to represent
the behavioral model. Th e sequence and communication diagrams modeled the interaction
among instances of classes that work together to support the business processes included in
a system, the behavioral state machine described the state changes through which an object
traverses during its lifetime, and the CRUDE matrix represented a system-level overview of
the interactions among the objects in the system. In this chapter, we combine walkthroughs
with CRUDE matrices to more completely verify and validate the behavioral models. Since
we covered CRUDE analysis and matrices in the previous section, we focus only on walk-
throughs in this section. We again use the appointment system and focus on Figures 6-1, 6-10,
6-16, and 6-23 to describe a set of rules that can be used to ensure that the behavioral model
is internally consistent.
First, every actor and object included on a sequence diagram must be included as an actor
and an object on a communication diagram, and vice versa. For example, in Figures 6-1 and 6-10,
the aReceptionist actor and the Patients object appear on both diagrams.
Second, if there is a message on the sequence diagram, there must be an association on
the communications diagram, and vice versa. For example, Figure 6-1 portrays a message
being sent from the aReceptionist actor to the Patient object, and a matching association
appears in the corresponding communication diagram (see Figure 6-10).
Th ird, every message that is included on a sequence diagram must appear as a message on
an association in the corresponding communication diagram, and vice versa. For example, the
LookUpPatient() message sent by the aReceptionist actor to the Patient object on the sequence
diagram (see Figure 6-1) appears as a message on the association between the aReceptionist actor
and the Patient object on the communication diagram (see Figure 6-10).
Fourth, if a guard condition appears on a message in the sequence diagram, there must be
an equivalent guard condition on the corresponding communication diagram, and vice versa.
For example, the message sent from the aReceptionist actor to the UnpaidBills object has a
guard condition of [aPatient Exists] (see Figure 6-1). Figure 6-10 shows the matching guard
condition included on the communication diagram.
Fift h, the sequence number included as part of a message label in a communications
diagram implies the sequential order in which the message will be sent. Th erefore, it must
correspond to the top-down ordering of the messages being sent on the sequence diagram.
For example, the LookUpPatient message sent from the aReceptionist actor to the Patient
object on the sequence diagram (see Figure 6-1) is the second from the top of the diagram.
Th e LookUpPatient message sent from the aReceptionist actor to the Patients object on the
communications diagram (see Figure 6-10) is labeled with the number 2.21
Sixth, all transitions contained in a behavior state machine must be associated with a
message being sent on a sequence and communication diagram, and it must be classifi ed as
a (C)reate, (U)pdate, or (D)elete message in a CRUDE matrix. For example, in Figure 6-16 the
Checks In transition must be associated with a message in the corresponding sequence and
communication diagrams. Furthermore, it should be associated with an (U)pdate entry in the
CRUDE matrix associated with the hospital patient system.
Seventh, all entries in a CRUDE matrix imply a message being sent from an actor or object
to another actor or object. If the entry is a (C)reate, (U)pdate, or (D)elete, then there must be an
20 Th e material in this section has been adapted from E. Yourdon, Modern Structured Analysis (Englewood Cliff s,
NJ: Prentice Hall, 1989).
21 Th ere are more complicated numbering schemes that could be used. However, for our purposes, a simple sequen-
tial number is suffi cient.
234 Chapter 6 Behavioral Modeling
FIGURE 6-27 Interrelationships among Behavioral Models
Including
HasKinds
Contains
HasKinds
Describe
Contains Contains
Contains
AssociatedWith
AssociatedWith
Interaction Diagram
Behavioral Models
Sequence DiagramCommunication Diagram
Objects Actors
Delete
Update
States
Create
Messages
Cell Entries Transitions
CRUD Matrix
Associations
Read
Behavioral State Machine
22 We have delayed the description of designing operations and methods until Chapter 8. Th erefore, the detailed infor-
mation required to understand a specifi c message has not been created yet. However, in many cases, enough information
will already have been created to validate many of the transitions in behavioral state machines and CRUDE matrices.
23 A good reference for these types of restrictions is S.W. Ambler, Th e Elements of UML 2.0 Style (Cambridge,
England: Cambridge University Press, 2005).
associated transition in a behavioral state machine that represents the instances of the receiving
class. For example in Figure 6-23 the R and U entries in the Receptionist row and Appointments
column imply that instances of the Receptionist actor will read and update instances of the
Appointments class. Th us, there should be read and update messages on the sequence and
communication diagrams corresponding with the appointments processes. Reviewing Figures
6-1 and 6-10, we see that there is a message, MatchAppts(), from the aReceptionist actor to the
Appointments object. However, based on this review, it is unclear whether the MatchAppts()
message represents a read, an update, or both. Th erefore, additional analysis is required.22
Because there is an (U)pdate message involved, there must be a transition on a behavioral state
machine that portrays the life cycle of an Appointments object.
Finally, many representation-specifi c rules have been proposed. However, as in the other
models, these rules are beyond the scope of this section on verifi cation and validation.23
Figure 6-27 portrays the associations among the behavioral models.
Key Terms 235
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Aft er developing the functional and structural models, the project manager, Ruby Ross,
tasked the team with developing the behavioral models for the Mobile Scheduling
(Version 1) of the Integrated Health Clinic Delivery System. Th e behavioral models include
the interactions diagram (sequence and communications) as well as the behavioral state
machine. In addition, the team created a CRUDE matrix to analyze the collaboration
between the objects identifi ed during structural modeling. While the structural model
depicted the static aspects of the system, behavioral models show the internal dynamic
aspects of the system. By modeling both the static (structural) and dynamic (behavioral)
aspects of a system, object-oriented systems analysis and design attempts to view the
underlying problem domain in a holistic way.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the purpose of the behavioral models.
Describe the purpose of the interaction diagrams.
Describe the diff erent elements of the sequence diagrams.
Create a behavioral model using a sequence diagram.
Describe the diff erent elements of the communication diagrams.
Create a behavioral model using a communication diagram.
Explain the purpose of a behavioral state machine.
Describe the diff erent elements of the behavioral state machines.
Create a behavioral model using a behavioral state machine.
Describe the purpose of CRUDE analysis.
Create a behavioral model using a CRUDE matrix.
Verify and validate the evolving behavioral model using CRUDE analysis and walkthroughs.
Verify and validate the functional model by ensuring the consistency of the four behavioral representations:
sequence diagrams, communication diagrams, behavioral state machines, and a CRUDE matrix.
KEY TERMS
Action
Activity
Actor
Association
Attributes
Behavior
Behavior models
Behavioral state machines
Black hole states
Class
Class diagram
Collaboration
Communication diagram
Condition
CRC cards
CRUDE analysis
CRUDE matrix
Dynamic model
Event
Execution occurrence
Final state
Frame
Generic sequence diagram
Guard condition
Initial state
Instance
Instance sequence
diagram
Lifeline
Message
Method
Miracle states
Object
Operation
Operation call message
Packages
Return message
Scenario
Self-delegation
Sequence
diagram
State
State symbol
Temporary object
Transition
Trigger
Use case
236 Chapter 6 Behavioral Modeling
1. How is behavioral modeling related to functional and
structural modeling?
2. How does a use case relate to a sequence diagram? A
communication diagram?
3. Contrast the following sets of terms: state, behavior,
class, object, action, and activity.
4. Why is iteration important when creating a behavioral
model?
5. What are the main building blocks for the sequence
diagram? How are they represented on the model?
6. How do you show that a temporary object is to go out
of existence on a sequence diagram?
7. Do lifelines always continue down the entire page of a
sequence diagram? Explain.
8. Describe the steps used to create a sequence diagram.
9. When drawing a sequence diagram, what guidelines
should you follow?
10. Describe the main building blocks for the commu-
nication diagram and how they are represented on
the model.
11. How do you show the sequence of messages on a
communication diagram?
12. How do you show the direction of a message on a
communication diagram?
13. Describe the steps used to create a communication
diagram.
14. When drawing a communication diagram, what
guidelines should you follow?
15. Are states always depicted using rounded rectangles
on a behavioral state machine? Explain.
16. What kinds of events can lead to state transitions on a
behavioral state machine?
17. What are the steps in building a behavioral state
machine?
18. When drawing a behavioral state machine, what
guidelines should you follow?
19. How are guard conditions shown on a behavioral state
machine?
20. Describe the type of class that is best represented by
a behavioral state machine. Give two examples of
classes that would be good candidates for a behavioral
state machine.
21. What is CRUDE analysis, and what is it used for?
22. Identify the models that contain each of the following
components: actor, association, class, extends, asso-
ciation, fi nal state, guard condition, initial state, links,
message, multiplicity, object, state, transition, and
update operation.
QUESTIONS
EXERCISES
A. Th ink about sending a fi rst-class letter to an interna-
tional pen pal. Describe the process that the letter goes
through to get from your initial creation of the letter
to being read by your friend, from the letter’s perspec-
tive. Draw a behavioral state machine that depicts the
states that the letter moves through.
B. Draw a behavioral state machine that describes the
various states that a travel authorization can have
through its approval process. A travel authorization
form is used in most companies to approve travel
expenses for employees. Typically, an employee fi lls
out a blank form and sends it to his or her boss for a
signature. If the amount is fairly small ($300), then the boss signs the form and
sends it to the CFO, along with a paragraph explain-
ing the purpose of the travel; the CFO signs the form
and passes it along to accounts payable. Of course,
the boss and the CFO can reject the travel authori-
zation form if they do not feel that the expenses are
reasonable. In this case, the employee can change the
form to include more explanation or decide to pay
the expenses.
C. Th ink about the system that handles student admissions
at your university. Th e primary function of the system
should be able to track a student from the request for
information through the admissions process until the
student is either admitted to the school or rejected.
1. Write a use-case description that can describe an
Admit Student use case.
Assume that applicants who are children of alumni
are handled diff erently from other applicants. Also,
assume that a generic Update Student Information use
case is available for your system to use.
Minicases 237
2. Create a use-case diagram that includes all of the
above use cases.
Assume that an admissions form includes the con-
tents of the form, SAT information, and references.
Additional information is captured about children of
alumni, such as their parent’s graduation year, contact
information, and college major.
3. Create a class diagram for the use cases identifi ed
with questions 1 and 2. Also, be sure to include the
above information.
Assume that a temporary student object is used by the
system to hold information about people before they
send in an admission form. Aft er the form is sent in,
these people are considered students.
4. Create sequence diagrams for the scenarios of the
above use cases.
5. Create a communication diagram for the scenarios
of the above use cases.
6. Create a behavioral state machine to depict a person
as he or she moves through the admissions process.
7. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
D. For the A Real Estate Inc. problem in Chapters 4
(exercises I, J, and K) and 5 (exercises P and Q):
1. Choose one use case and, for each scenario, create
a sequence diagram.
2. Create a communication diagram for each scenario
of the use case chosen in Question 1.
3. Create a behavioral state machine to depict one
of the classes on the class diagram you created for
Chapter 5, exercise P.
4. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
5. Perform a verifi cation and validation walkthrough
of the problem.
E. For the A Video Store problem in Chapters 4 (exer-
cises L, M, and N) and 5 (exercises R and S):
1. Choose one use case and, for each scenario, create
a sequence diagram.
2. Create a communication diagram for each scenario
of the use case chosen in Question 1.
3. Create a behavioral state machine to depict one
of the classes on the class diagram you created for
Chapter 5, exercise R.
4. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
5. Perform a verifi cation and validation walkthrough
of the problem.
F. For the gym membership problem in Chapters 4 (exer-
cises O, P, and Q) and 5 (exercises T and U):
1. Choose one use case and, for each scenario, create
a sequence diagram.
2. Create a communication diagram for each scenario
of the use case chosen in Question 1.
3. Create a behavioral state machine to depict one
of the classes on the class diagram you created for
Chapter 5, exercise T.
4. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
5. Perform a verifi cation and validation walkthrough
of the problem.
G. For the Picnics R Us problem in Chapters 4 (exercises
R, S, and T) and 5 (exercises V and W):
1. Choose one use case and, for each scenario, create
a sequence diagram.
2. Create a communication diagram for each scenario
of the use case chosen in Question 1.
3. Create a behavioral state machine to depict one
of the classes on the class diagram you created for
Chapter 5, exercise V.
4. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
5. Perform a verifi cation and validation walkthrough
of the problem.
H. For the Of-the-Month-Club problem in Chapters 4
(exercises U, V, and W) and 5 (exercises X and Y):
1. Choose one use case and, for each scenario, create
a sequence diagram.
2. Create a communication diagram for each scenario
of the use case chosen in Question 1.
3. Create a behavioral state machine to depict one
of the classes on the class diagram you created for
Chapter 5, exercise X.
4. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
5. Perform a verifi cation and validation walkthrough
of the problem.
1. Refer to the functional model (use-case diagram,
activity diagrams, and use-case descriptions) you pre-
pared for the Professional and Scientifi c Staff Manage-
ment (PSSM) Minicase in Chapter 4. Based on your
performance, PSSM was so satisfi ed that it wanted you
to develop both the structural and behavioral models
MINICASES
238 Chapter 6 Behavioral Modeling
so that it could more fully understand both the inter-
action that would take place between the users and the
system and the system itself in greater detail.
a. Create both CRC cards and a class diagram based
on the functional models created in Chapter 4.
b. Create a sequence and a communication diagram
for each scenario of each use case identifi ed in the
functional model.
c. Create a behavioral state machine for each of the
complex classes in the class diagram.
d. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
e. Perform a verifi cation and validation walkthrough
of each model: functional, structural, and behavioral.
2. Refer to the structural model (CRC cards and class dia-
gram) that you created for the Holiday Travel Vehicles
Minicase in Chapter 5. Based on your performance,
Holiday Travel Vehicles was so satisfi ed that it wanted
you to develop both the functional and behavioral
models so that it could more fully understand both the
interaction that would take place between the users
and the system and the system itself in greater detail.
a. Based on the structural model you created in Chap-
ter 5 and the problem description in Chapter 5, create
a functional model (use case diagram, activity dia-
grams, and use case descriptions) for the business
processes associated with the Holiday Travel Vehi-
cles sales system.
b. Create a sequence and a communication diagram
for each scenario of each use case identifi ed in the
functional model.
c. Create a behavioral state machine for each of the
complex classes in the class diagram.
d. Perform a CRUDE analysis to show the interactiv-
ity of the objects in the system.
e. Perform a verifi cation and validation walk-
through of each model: functional, structural, and
behavioral.
Storyboards
U
ser Interface
Prototypes
C
ontracts
Interface
Tem
plates
W
indow
s
N
avigation
D
iagram
s
Whereas analysis modeling concentrated on the functional
requirements of the evolving system, design modeling incorporates
the nonfunctional requirements. Th at is, design modeling focuses
on how the system will operate. First, the project team verifi es
and validates the analysis models (functional, structural, and
behavioral). Next, a set of factored and partitioned analysis models
are created. Th e class and method designs are illustrated using
the class specifi cations (using CRC cards and class diagrams),
contracts, and method specifi cations. Next, the data management
layer is addressed by designing the actual database or fi le structure
to be used for object persistence, and a set of classes that will map
the class specifi cations into the object persistence format chosen.
Concurrently, the team produces the user interface layer design
using use scenarios, windows navigation diagrams, real use cases,
interface templates, storyboards, windows layout diagrams, and
user interface prototypes. Th e physical architecture layer design
is created using deployment diagrams and hardware soft ware
specifi cations. Th is collection of deliverables represents the
system specifi cation that is handed to the programming team for
implementation.
CHAPTER 7
Moving on
to Design
CHAPTER 8
Class and
Method Design
CHAPTER 9
Data
Management
Layer Design
CHAPTER 10
Human
Computer
Interaction Layer
Design
CHAPTER 11
Physical
Architecture
Layer Design
Factored/
Partitioned
Functional M
odel
Factored/
Partitioned
B
ehavioral M
odel
Factored/
Partitioned
Structural M
odel
Package
D
iagram
sA
lternative
M
atrix
C
lass
Specifi cations
M
ethod
Specifi cations
O
bject
Persistence
D
esign
D
ata A
ccess &
M
anipulation
C
lass D
esign
U
se
Scenarios
R
eal U
se C
ases
P A R T T W O
Design Modeling
W
indow
s
Layout
D
iagram
s
D
eploym
ent
D
iagram
s
H
ardw
are/
Softw
are
Specifi cation
240
Object-oriented system development uses the requirements that were gathered during
analysis to create a blueprint for the future system. A successful object-oriented design
builds upon what was learned in earlier phases and leads to a smooth implementation by
creating a clear, accurate plan of what needs to be done. Th is chapter describes the initial
transition from analysis to design and presents three ways to approach the design for the
new system.
OBJECTIVES
■ Understand the verifi cation and validation of the analysis models.
■ Understand the transition from analysis to design.
■ Understand the use of factoring, partitions, and layers.
■ Be able to create package diagrams.
■ Be familiar with the custom, packaged, and outsource design alternatives.
■ Be able to create an alternative matrix.
INTRODUCTION
Th e purpose of analysis is to fi gure out what the business needs are. Th e purpose of design is
to decide how to build the system. Th e major activity that takes place during design is evolving
the set of analysis representations into design representations.
Th roughout design, the project team carefully considers the new system with respect
to the current environment and systems that exist within the organization as a whole.
Major considerations in determining how the system will work include environmental fac-
tors, such as integrating with existing systems, converting data from legacy systems, and
leveraging skills that exist in-house. Although the planning and analysis are undertaken
to develop a possible system, the goal of design is to create a blueprint for a system that
can be implemented.
An important initial part of design is to examine several design strategies and decide
which will be used to build the system. Systems can be built from scratch, purchased and
customized, or outsourced to others, and the project team needs to investigate the via-
bility of each alternative. Th is decision infl uences the tasks that are to be accomplished
during design.
At the same time, detailed design of the individual classes and methods that are used to
map out the nuts and bolts of the system and how they are to be stored must still be completed.
Techniques such as CRC cards, class diagrams, contract specifi cation, method specifi cation,
and database design provide the fi nal design details in preparation for the implementation
C H A P T E R 7
Moving on to Design
Introduction 241
phase, and they ensure that programmers have suffi cient information to build the right sys-
tem effi ciently. Th ese topics are covered in Chapters 8 and 9.
Design also includes activities such as designing the user interface, system inputs, and
system outputs, which involve the ways that the user interacts with the system. Chapter 10
describes these three activities in detail, along with techniques such as storyboarding and
prototyping, which help the project team design a system that meets the needs of its users
and is satisfying to use.
Finally, physical architecture decisions are made regarding the hardware and soft ware
that will be purchased to support the new system and the way that the processing of the
system will be organized. For example, the system can be organized so that its processing is
centralized at one location, distributed, or both centralized and distributed, and each solution
off ers unique benefi ts and challenges to the project team. Because global issues and security
infl uence the implementation plans that are made, they need to be considered along with the
system’s technical architecture. Physical architecture, security, and global issues are described
in Chapter 11.
Th e many steps of design are highly interrelated and, as with the steps in analysis, the
analysts oft en go back and forth among them. For example, prototyping in the interface
design step oft en uncovers additional information that is needed in the system. Alternatively,
a system that is being designed for an organization that has centralized systems might require
substantial hardware and soft ware investments if the project team decides to change to a sys-
tem in which all the processing is distributed.
Avoiding Classic Design
In Chapter 2, we discussed several classic mistakes and
how to avoid them. Here, we summarize four classic mis-
takes in design and discuss how to avoid them.
1. Reducing design time: If time is short, there is a
temptation to reduce the time spent in “unproduc-
tive” activities such as design so that the team can
jump into “productive” programming. This results
in missing important details that have to be investi-
gated later at a much higher time and cost (usually at
least ten times higher).
Solution: If time pressure is intense, use timeboxing
to eliminate functionality or move it into future
versions.
2. Feature creep: Even if you are successful at avoiding
scope creep, about 25 percent of system require-
ments will still change. And, changes—big and
small—can signifi cantly increase time and cost.
Solution: Ensure that all changes are vital and that
the users are aware of the impact on cost and time.
Try to move proposed changes into future versions.
3. Silver bullet syndrome: Analysts sometimes believe
the marketing claims for some design tools that claim
to solve all problems and magically reduce time and
costs. No one tool or technique can eliminate overall
time or costs by more than 25 percent (although some
can reduce individual steps by this much).
Solution: If a design tool has claims that appear too
good to be true, just say no.
4. Switching tools midproject: Sometimes analysts
switch to what appears to be a better tool during
design in the hopes of saving time or costs. Usually,
any benefi ts are outweighed by the need to learn the
new tool. This also applies even to minor upgrades
to current tools.
Solution: Don’t switch or upgrade unless there is a
compelling need for specifi c features in the new tool,
and then explicitly increase the schedule to include
learning time.
Based upon material from Steve McConnell, Rapid Development (Redmond,
WA: Microsoft Press, 1966).
PRACTICAL
TIP
242 Chapter 7 Moving on to Design
VERIFYING AND VALIDATING THE ANALYSIS MODELS1
Before we evolve our analysis representations into design representations, we need to
verify and validate the current set of analysis models to ensure that they faithfully represent
the problem domain under consideration. Th is includes testing the fi delity of each model;
for example, we must be sure that the activity diagram(s), use-case descriptions, and use-
case diagrams all describe the same functional requirements. It also involves testing the
fi delity between the models; for instance, transitions on a behavioral state machine are
associated with operations contained in a class diagram. In Chapters 4, 5, and 6, we focused
on verifying and validating the individual models: function, structural, and behavioral. In
this chapter, we center our attention on ensuring that the diff erent models are consistent.
Figure 7-1 portrays the fact that the object-oriented analysis models are highly interrelated.
For example, do the functional and structural models agree? What about the functional and
behavioral models? And fi nally, are the structural and behavioral models trustworthy? In this
section, we describe a set of rules that are useful to verify and validate the intersections of the
analysis models. Depending on the specifi c constructs of each actual model, diff erent inter-
relationships are relevant. Th e process of ensuring the consistency among them is known as
balancing the models.
Balancing Functional and Structural Models
To balance the functional and structural models, we must ensure that the two sets of models
are consistent with each other. Th at is, the activity diagrams, use-case descriptions, and
use-case diagrams must agree with the CRC cards and class diagrams that represent the
evolving model of the problem domain. Figure 7-2 shows the interrelationships between
the functional and structural models. By reviewing this fi gure, we uncover four sets of
associations between the models. Th is gives us a place to begin balancing the functional
and structural models.2
First, every class on a class diagram and every CRC card must be associated with at least
one use case, and vice versa. For example, the CRC card portrayed in Figure 7-3 and its related
class contained in the class diagram (see Figure 7-4) are associated with the Make Old Patient
Appt use case described in Figure 7-5.
Second, every activity or action contained in an activity diagram (see Figure 7-6) and
every event contained in a use-case description (see Figure 7-5) should be related to one
or more responsibilities on a CRC card and one or more operations in a class on a class
diagram and vice versa. For example, the Get Patient Information activity on the example
activity diagram (see Figure 7-6) and the fi rst two events on the use-case description (see
Figure 7-5) are associated with the make appointment responsibility on the CRC card
(see Figure 7-3) and the makeAppointment() operation in the Patient class on the class
diagram (see Figure 7-4).
Th ird, every object node on an activity diagram must be associated with an instance of
a class on a class diagram (i.e., an object) and a CRC card or an attribute contained in a class
and on a CRC card. However, in Figure 7-6, there is an object node, Appt Request Info, that
does not seem to be related to any class in the class diagram portrayed in Figure 7-4. Th us,
either the activity or class diagram is in error or the object node must represent an attrib-
ute. In this case, it does not seem to represent an attribute. We could add a class to the
1 Th e material in this section is based upon material from E. Yourdon, Modern Structured Analysis (Englewood
Cliff s, NJ: Prentice Hall, 1989). Verifying and validating are a type of testing. We also describe testing in Chapter 12.
2 Role-playing the CRC cards (see Chapter 5) also can be very useful in verifying and validating the relationships
among the functional and structural models.
Verifying and Validating the Analysis Models 243
class diagram that creates temporary objects associated with the object node on the activity
diagram. However, it is unclear what operations, if any, would be associated with these
temporary objects. Th erefore, a better solution would be to delete the Appt Request Info
object nodes from the activity diagram. In reality, this object node represented only a set of
bundled attribute values, i.e., data that would be used in the appointment system process
(see Figure 7-7).
Fourth, every attribute and association/aggregation relationships contained on a CRC
card (and connected to a class on a class diagram) should be related to the subject or object
of an event in a use-case description. For example, in Figure 7-5, the second event states: Th e
Patient provides the Receptionist with his or her name and address. By reviewing the CRC card
in Figure 7-3 and the class diagram in Figure 7-4, we see that the Patient class is a subclass of
the Participant class and hence inherits all the attributes, associations, and operations defi ned
with the Participant class, where name and address attributes are defi ned.
Balancing Functional and Behavioral Models
As in balancing the functional and structural models, we must ensure the consistency of the
two sets of models. In this case, the activity diagrams, use-case descriptions, and use-case
diagrams must agree with the sequence diagrams, communication diagrams, behavioral state
machines, and CRUDE matrix. Figure 7-8 portrays the relationships between the functional
and behavioral models. Based on these interrelationships, we see that there are four areas with
which we must be concerned.3
First, the sequence and communication diagrams must be associated with a use case
on the use-case diagram and a use-case description. For example, the sequence diagram in
Figure 7-9 and the communication diagram in Figure 7-10 are related to scenarios of the
Make Old Patient Appt use case that appears in the use-case description in Figure 7-5 and the
use-case diagram in Figure 7-11.
Second, actors on sequence diagrams, communication diagrams, and/or CRUDE
matrices must be associated with actors on the use-case diagram or referenced in the use-
case description, and vice versa. For example, the aPatient actor in the sequence diagram
in Figure 7-9, the communication diagram in Figure 7-10, and the Patient row and column
in the CRUDE matrix in Figure 7-12 appears in the use-case diagram in Figure 7-11 and
the use-case description in Figure 7-5. However, the aReceptionist does not appear in the
use-case diagram but is referenced in the events associated with the Make Old Patient Appt
use-case description. In this case, the aReceptionist actor is obviously an internal actor, which
cannot be portrayed on UML’s use-case diagram.
FIGURE 7-1
Object-Oriented
Analysis Models
Functional
Models
Structural
Models
Behavioral
Models
3 Performing CRUDE analysis (see Chapter 6) could also be useful in reviewing the intersections among the func-
tional and behavioral models.
244
FI
G
U
R
E
7
-2
R
el
at
io
n
sh
ip
s
am
o
n
g
Fu
n
ct
io
n
al
a
n
d
S
tr
u
ct
u
ra
l M
o
d
el
s
In
cl
ud
in
g
C
on
ta
in
s
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
In
st
an
ce
O
f
R
ep
re
se
nt
s
H
av
e
H
as
K
in
ds
H
as
K
in
ds
H
as
K
in
dsC
on
ta
in
s
In
cl
ud
in
g
C
on
ta
in
s
H
as
K
in
ds
C
on
ta
in
s
C
on
ta
in
s
H
av
e
C
on
ta
in
s
C
on
ta
in
s
St
ru
ct
ur
al
M
od
el
s
C
R
C
C
ar
ds
O
bj
ec
t
D
ia
gr
am
C
la
ss
D
ia
gr
am
C
ol
la
bo
ra
to
rs
R
es
po
ns
ib
ili
ti
es
Ty
pe
O
bj
ec
ts
C
la
ss
es
A
gg
re
ga
ti
on
A
ss
oc
ia
ti
on
G
en
er
al
iz
at
io
n
A
ss
oc
ia
ti
on
s/
R
el
at
io
ns
hi
ps
C
om
po
si
ti
on
A
ss
oc
ia
ti
on
C
la
ss
Fu
nc
ti
on
al
M
od
el
s
U
se
-C
as
e
D
ia
gr
am
A
ct
iv
it
y
D
ia
gr
am
A
ct
iv
it
ie
s/
A
ct
io
ns
U
se
C
as
e
D
es
cr
ip
ti
on
s
Fl
ow
s
Ev
en
ts
A
ct
or
s
O
bj
ec
t
N
od
es
St
ak
eh
ol
de
rs
R
el
at
io
ns
hi
ps
O
bj
ec
t
Fl
ow
s
C
on
tr
ol
F
lo
w
s
U
se
C
as
es
Sc
en
ar
io
s
A
tt
ri
bu
te
s
O
pe
ra
ti
on
s
Verifying and Validating the Analysis Models 245
Th ird, messages on sequence and communication diagrams, transitions on behavioral
state machines, and entries in a CRUDE matrix must be related to activities and actions
on an activity diagram and events listed in a use-case description, and vice versa. For
example, the CreateAppt() message on the sequence and communication diagrams (see
Figures 7-9 and 7-10) is related to the CreateAppointment activity (see Figure 7-7) and the
S-1: New Appointment subfl ow on the use-case description (see Figure 7-5). Th e C entry
in the Receptionist Appointment cell of the CRUDE matrix is also associated with these
messages, activity, and subfl ow.
Fourth, all complex objects represented by an object node in an activity diagram must
have a behavioral state machine that represents the object’s lifecycle, and vice versa. As stated
in Chapter 6, complex objects tend to be very dynamic and pass through a variety of states
during their lifetimes. However, in this case because we no longer have any object nodes in
the activity diagram (see Figure 7-7), there is no necessity for a behavioral state machine to
be created based on the activity diagram.
Front:
Class Name: Patient ID: 3
Change status
Medical history
Calculate last visit
Make appointment Appointment
Provide medical history
Responsibilities
Associated Use Cases: 2Description: An individual who needs to receive or has received
medical attention
Type: Concrete, Domain
Collaborators
Back:
Attributes:
Insurance carrier (text)
Amount (double)
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: Appointment, Medical History
Participant
FIGURE 7-3
Old Patient CRC
Card (Figure 5-25)
A
cc
ou
nt
Em
pl
oy
ee0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
1.
.*
1.
.*
1.
.*
1.
.*
1.
.*
1.
.1
v
1.
.1
1.
.1
1.
.1
1.
.1
0.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
D
eb
it
C
re
di
t
En
tr
y
A
pp
oi
nt
m
en
t
-t
im
e
-d
at
e
-r
ea
so
n
+
ca
nc
el
w
ith
ou
t n
ot
ic
e(
)
Pa
ti
en
t
-a
m
ou
nt
-i
ns
ur
an
ce
c
ar
ri
er
+
m
ak
e
ap
po
in
tm
en
t()
+
ca
lc
ul
at
e
la
st
v
is
it(
)
+
ch
an
ge
s
ta
tu
s(
)
+
pr
ov
id
es
m
ed
ic
al
h
is
to
ry
()
Pa
rt
ic
ip
an
t
-l
as
tn
am
e
-fi
rs
tn
am
e
-a
dd
re
ss
-p
ho
ne
-b
ir
th
da
te
-/
ag
e
It
em
Tr
an
sa
ct
io
n
Li
ne
I
te
m
ha
s
ha
s
co
nt
ai
ns
A
ss
ig
ne
dT
o
D
oc
to
r
R
ec
ep
ti
on
is
t
N
ur
se
M
ed
ic
al
H
is
to
ry
-h
ea
rt
d
is
ea
se
-h
ig
h
bl
oo
d
pr
es
su
re
-d
ia
be
te
s
-a
lle
rg
ie
s
pr
ov
id
es
su
ffe
rs
schedules
lo
ca
te
dA
t
G
oo
d
Se
rv
ic
e
Pr
es
cr
ip
ti
on
B
ra
ce
Ph
ys
ic
al
C
he
ck
up
Sy
m
pt
om
Il
ln
es
s
Pl
ac
e
Tr
ea
tm
en
t
m
ed
ic
at
io
n
in
st
ru
ct
io
ns
sy
m
pt
om
s
ev
er
ity
+
p
ri
m
ar
y
in
su
ra
nc
e
ca
rr
ie
r
-n
am
e
-d
es
cr
ip
tio
n
FI
G
U
R
E
7
-4
A
p
p
o
in
tm
en
t
Pr
o
b
le
m
C
la
ss
D
ia
gr
am
(
Fi
gu
re
5
-7
)
246
Verifying and Validating the Analysis Models 247
FIGURE 7-5 Use-Case Description for the Make Old Patient Appt Use Case (Figure 4-13)
Use-Case Name:
Primary Actor:
Stakeholders and Interests:
Brief Description:
ID: Importance Level:
Trigger:
Normal Flow of Events:
SubFlows:
Alternate/Exceptional Flows:
Type: External
Make Old Patient Appt 2 Low
Old Patient
Old patient – wants to make, change, or cancel an appointment
Doctor – wants to ensure patient’s needs are met in a timely manner
Use Case Type: Detail, Essential
Patient calls and asks for a new appointment or asks to cancel or change an existing appointment
Old Patient
Update Patient Information
Manage Appointments
1. The Patient contacts the office regarding an appointment.
2. The Patient provides the Receptionist with his or her name and address.
3. If the Patient’s information has changed
Execute the Update Patient Information use case.
4. If the Patient’s payment arrangements has changed
Execute the Make Payments Arrangements use case.
5. The Receptionist asks Patient if he or she would like to make a new appointment, cancel an existing appointment, or change
an existing appointment.
S-1: New Appointment
1. The Receptionist asks the Patient for possible appointment times.
2. The Receptionist matches the Patient’s desired appointment times with available dates and
times and schedules the new appointment.
S-2: Cancel Appointment
1. The Receptionist asks the Patient for the old appointment time.
2. The Receptionist finds the current appointment in the appointment file and cancels it.
S-3: Change Appointment
1. The Receptionist performs the S-2: cancel appointment subflow.
2. The Receptionist performs the S-1: new appointment subflow.
S-1, 2a1: The Receptionist proposes some alternative appointment times based on what is available in the
appointment schedule.
S-1, 2a2: The Patient chooses one of the proposed times or decides not to make an appointment.
6. The Receptionist provides the results of the transaction to the Patient.
This use case describes how we make an appointment as well as changing or canceling
an appointment for a previously seen patient.
Relationships:
Association:
Include:
Extend:
Generalization:
If the patient wants to make a new appointment,
the S-1: new appointment subflow is performed.
If the patient wants to cancel an existing appointment,
the S-2: cancel appointment subflow is performed.
If the patient wants to change an existing appointment,
the S-3: change appointment subflow is performed.
TEMPLATE
can be found at
www.wiley.com
/college/dennis
248 Chapter 7 Moving on to Design
Get Patient Information
Appt
Request Info
Appt
Request InfoCreate New Patient
Update Patient Information
[New Patient][Old Patient]
[Change]
Cancel Appointment Change AppointmentCreate Appointment
Make Payment Arrangements
Create Appointment
Make Payment Arrangements
[New Info]
[New Arrange]
[Cancel]
[Create]
FIGURE 7-6 Activity Diagram for the Manage Appointments Use Case (Figure 4-8)
Get Patient Information
Create New Patient
Update Patient Information
[New Patient][Old Patient]
[Change]
Cancel Appointment Change AppointmentCreate Appointment
Make Payment Arrangements
Create Appointment
Make Payment Arrangements
[New Info]
[New Arrange]
[Cancel]
[Create]
FIGURE 7-7 Corrected Activity Diagram for the Manage Appointments Use Case
Verifying and Validating the Analysis Models 249
In
cl
ud
in
g
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
C
on
ta
in
s
D
es
cr
ib
e
In
cl
ud
in
g
C
on
ta
in
s
C
on
ta
in
s
C
on
ta
in
s
H
as
K
in
ds
H
as
K
in
ds
C
on
ta
in
s
C
on
ta
in
s
H
av
e
H
as
K
in
ds
C
on
ta
in
s
B
eh
av
io
ra
l M
od
el
s
C
R
U
D
E
M
at
ri
x
C
om
m
un
ic
at
io
n
D
ia
gr
am
A
ss
oc
ia
ti
on
s
M
es
sa
ge
s
Se
qu
en
ce
D
ia
gr
am
C
el
l E
nt
ri
es
Tr
an
si
ti
on
s
St
at
es
A
ct
iv
it
y
D
ia
gr
am
U
se
-C
as
e
D
ia
gr
am
U
se
-C
as
e
D
es
cr
ip
ti
on
s
O
bj
ec
t
N
od
es
A
ct
iv
it
ie
s/
A
ct
io
ns
St
ak
eh
ol
de
rs
R
el
at
io
ns
hi
ps
U
pd
at
e
C
re
at
e
O
bj
ec
ts
U
se
C
as
es
C
on
tr
ol
F
lo
w
s
O
bj
ec
t
Fl
ow
s
Fl
ow
s
Ev
en
ts
A
ct
or
s
Fu
nc
ti
on
al
M
od
el
s
B
eh
av
io
ra
l S
ta
te
M
ac
hi
ne
In
te
ra
ct
io
n
D
ia
gr
am
R
ea
d
D
el
et
e
Sc
en
ar
io
s
A
ct
or
s
FI
G
U
R
E
7
-8
R
el
at
io
n
sh
ip
s
b
et
w
ee
n
F
u
n
ct
io
n
al
a
n
d
B
eh
av
io
ra
l M
o
d
el
s
250
Verifying and Validating the Analysis Models 251
sd Make Appt Use Case
RequestAppt(name, address)
NewCancelChangeAppt?()
ApptTimes?()
aPatient
LookUpPatient()
aReceptionist
[aPatient Exists] LookupBills()
MatchAppts()
CreateAppt()
aPatient:Patient :UnpaidBill :Appointment
FIGURE 7-9
Sequence Diagram
for a Scenario of the
Make Old Patient
Appt Use Case
(Figure 6-1)
sd Make Appt Use Case
aPatient
1: RequestAppt(name, address)
4: NewCancelChangeAppt?
5: ApptTimes?
aReceptionist
2: LookUpPatient()
3: [aPatient Exists] LookupBills()
7: CreateAppt
6: MatchAppts
:Appointment
aPatient:Patient
:UnpaidBill
FIGURE 7-10 Communication Diagram for a Scenario of the Make Old Patient Appt
Use Case (Figure 6-10)
Balancing Structural and Behavioral Models
To discover the relationships between the structural and behavioral models, we use the
concept map in Figure 7-13. In this case, there are fi ve areas in which we must ensure the
consistency between the models.4
4 Role-playing (see Chapter 5) and CRUDE analysis (see Chapter 6) also can be very useful in this undertaking.
252 Chapter 7 Moving on to Design
Appointment System
Patient
New Patient
Old Patient
Update Patient
Information
Make Old Patient
Appt
Make New Patient
Appt
Make Payment
Arrangements
Create New Patient
Manage
Appointments
<
>
<
>
*
*
*
*
<>
<>
Produce Schedules
Management
Doctor
Record
Availability
Manage
Schedule
<>
<>
*
*
*
*
FIGURE 7-11 Modifi ed Use-Case Diagram for the Appointment System (Figure 4-21)
FIGURE 7-12 CRUDE Matrix for the Make Old Patient Apt Use Case (Figure 6-23)
Receptionist RU CRUD R RU CRUD
PatientList R
Patient
UnpaidBills
Appointments R
Appointment
Receptionist PatientList Patient UnpaidBills Appointments Appointment
A
ss
oc
ia
te
dW
it
h
St
ru
ct
ur
al
M
od
el
s
C
R
C
C
ar
ds
In
cl
ud
in
g
C
on
ta
in
s
C
on
ta
in
s
In
st
an
ce
O
f
R
ep
re
se
nt
s
C
on
ta
in
s
A
ss
oc
ia
ti
on
s/
R
el
at
io
ns
hi
ps
A
tt
ri
bu
te
s
H
av
e
Ty
pe
O
pe
ra
ti
on
s
R
es
po
ns
ib
ili
ti
es
C
on
ta
in
s
In
cl
ud
in
g
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
G
en
er
al
iz
at
io
n
A
ct
or
s
H
as
K
in
ds
H
as
K
in
ds
H
as
K
in
ds
H
as
K
in
ds
H
as
K
in
ds
A
ss
oc
ia
ti
on
A
gg
re
ga
ti
on
C
on
ta
in
s
A
ss
oc
ia
ti
on
s
C
om
m
un
ic
at
io
n
D
ia
gr
am
Se
qu
en
ce
D
ia
gr
am
In
te
ra
ct
io
n
D
ia
gr
am
B
eh
av
io
ra
l M
od
el
s
C
R
U
D
E
M
at
ri
x
C
on
ta
in
s
D
es
cr
ib
e
O
bj
ec
ts
O
bj
ec
ts
M
es
sa
ge
s
C
on
ta
in
s
C
on
ta
in
s
D
el
et
e
Tr
an
si
ti
on
s
R
ea
d
C
re
at
e
C
el
l E
nt
ri
es
St
at
es
C
la
ss
D
ia
gr
am
O
bj
ec
t
D
ia
gr
am
C
ol
la
bo
ra
to
rs
A
ss
oc
ia
ti
on
C
la
ss
C
om
po
si
ti
on
U
pd
at
e
B
eh
av
io
ra
l S
ta
te
M
ac
hi
ne
C
la
ss
es
FI
G
U
R
E
7
-1
3
R
el
at
io
n
sh
ip
s
b
et
w
ee
n
S
tr
u
ct
u
ra
l a
n
d
B
eh
av
io
ra
l M
o
d
el
s
253
254 Chapter 7 Moving on to Design
First, objects that appear in a CRUDE matrix must be associated with classes that are
represented by CRC cards and appear on the class diagram, and vice versa. For example,
the Patient class in the CRUDE matrix in Figure 7-12 is associated with the CRC card in
Figure 7-3 and the Patient class in the class diagram in Figure 7-4.
Second, because behavioral state machines represent the life cycle of complex objects, they
must be associated with instances (objects) of classes on a class diagram and with a CRC card
that represents the class of the instance. For example, the behavioral state machine that describes
an instance of a Patient class in Figure 7-14 implies that a Patient class exists on a related class
diagram (see Figure 7-4) and that a CRC card exists for the related class (see Figure 7-3).
Th ird, communication and sequence diagrams contain objects that must be an instan-
tiation of a class that is represented by a CRC card and is located on a class diagram. For
example, Figure 7-9 and Figure 7-10 have an anAppt object that is an instantiation of the
Appointment class. Th erefore, the Appointment class must exist in the class diagram (see
Figure 7-4), and a CRC card should exist that describes it. However, there is an object
on the communication and sequence diagrams associated with a class that did not exist
on the class diagram: UnpaidBill. At this point, the analyst must decide to either modify
the class diagram by adding these classes or rethink the communication and sequence
diagrams. In this case, it is better to add the class to the class diagram (see Figure 7-15).
Fourth, messages contained on the sequence and communication diagrams, transitions
on behavioral state machines, and cell entries on a CRUDE matrix must be associated with
responsibilities and associations on CRC cards and operations in classes and asso ciations
connected to the classes on class diagrams. For example, the CreateAppt() message on the
sequence and communication diagrams (see Figures 7-9 and 7-10) relate to the makeAp-
pointment operation of the Patient class and the schedules association between the Patient
and Appointment classes on the class diagram (see Figure 7-15).
Fift h, the states in a behavioral state machine must be associated with diff erent values
of an attribute or set of attributes that describe an object. For example, the behavioral state
machine for the hospital patient object implies that there should be an attribute, possibly
current status, which needs to be included in the defi nition of the class.
Summary
Figure 7-16 portrays a concept map that is a complete picture of the interrelationships
among the diagrams covered in this section. It is obvious from the complexity of this
FIGURE 7-14 Behavioral State Machine for Hospital Patient (Figure 6-16)
Patient
Enters Hospital Checks In [Diagnosis = Healthy] [> 2 weeks]
Entering Admitted Released
Under Observation
[Diagnosis = Unhealthy]
[Diagnosis = Healthy]
A
cc
ou
nt
Em
pl
oy
ee0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
1.
.*
1.
.*
1.
.*
1.
.*
1.
.*
1.
.1
1.
.1
1.
.1
1.
.1
0.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
D
eb
it
C
re
di
t
En
tr
y
A
pp
oi
nt
m
en
t
-t
im
e
-d
at
e
-r
ea
so
n
+
ca
nc
el
w
ith
ou
t n
ot
ic
e(
)
Pa
ti
en
t
-a
m
ou
nt
-i
ns
ur
an
ce
c
ar
ri
er
+
m
ak
e
ap
po
in
tm
en
t()
+
ca
lc
ul
at
e
la
st
v
is
it(
)
+
ch
an
ge
s
ta
tu
s(
)
+
pr
ov
id
es
m
ed
ic
al
h
is
to
ry
()
Pa
rt
ic
ip
an
t
-l
as
tn
am
e
-fi
rs
tn
am
e
-a
dd
re
ss
-p
ho
ne
-b
ir
th
da
te
-/
ag
e
It
em
Tr
an
sa
ct
io
n
Li
ne
I
te
m
ha
s
ha
s
co
nt
ai
ns
A
ss
ig
ne
dT
o
+
p
ri
m
ar
y
in
su
ra
nc
e
ca
rr
ie
r
D
oc
to
r
R
ec
ep
ti
on
is
t
U
np
ai
d
B
ill
N
ur
se
M
ed
ic
al
H
is
to
ry
-h
ea
rt
d
is
ea
se
-h
ig
h
bl
oo
d
pr
es
su
re
-d
ia
be
te
s
-a
lle
rg
ie
s
pr
ov
id
es
su
ffe
rs
ow
esschedules
lo
ca
te
dA
t
G
oo
d
Se
rv
ic
e
Pr
es
cr
ip
ti
on
B
ra
ce
Ph
ys
ic
al
C
he
ck
up
Sy
m
pt
om
Il
ln
es
s
Pl
ac
e
Tr
ea
tm
en
t
m
ed
ic
at
io
n
in
st
ru
ct
io
ns
sy
m
pt
om
s
ev
er
ity
-n
am
e
-d
es
cr
ip
tio
n
FI
G
U
R
E
7
-1
5
C
o
rr
ec
te
d
A
p
p
o
in
tm
en
t
Sy
st
em
C
la
ss
D
ia
gr
am
255
In
cl
ud
in
g C
on
ta
in
s
C
on
ta
in
s
C
on
ta
in
s
R
ep
re
se
nt
s
C
on
ta
in
s
H
as
K
in
ds
H
as
K
in
ds
H
as
K
in
ds
H
av
e
H
as
K
in
ds
In
cl
ud
in
g
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
A
ss
oc
ia
te
dW
it
h
In
cl
ud
es
H
as
K
in
ds
In
cl
ud
in
g
C
on
ta
in
s
C
on
ta
in
s
C
on
ta
in
sC
on
ta
in
s
C
on
ta
in
s
C
on
ta
in
s
C
on
ta
in
s
H
as
K
in
ds
St
ru
ct
ur
al
M
od
el
s
C
R
C
C
ar
ds
O
bj
ec
t
D
ia
gr
am
C
la
ss
D
ia
gr
am
C
ol
la
bo
ra
to
rs A
tt
ri
bu
te
s
O
pe
ra
ti
on
s
Ty
pe
C
la
ss
es
A
ss
oc
ia
ti
on
s/
R
el
at
io
ns
hi
ps
A
gg
re
ga
ti
on
G
en
er
al
iz
at
io
n
C
om
po
si
ti
on
A
ss
oc
ia
ti
on
C
la
ss
B
eh
av
io
ra
l M
od
el
s
O
bj
ec
t-
O
ri
en
te
d
A
na
ly
si
s
M
od
el
s
In
te
ra
ct
io
n
D
ia
gr
am
B
eh
av
io
ra
l S
ta
te
M
ac
hi
ne
C
R
U
D
E
M
at
ri
x
C
om
m
un
ic
at
io
n
D
ia
gr
am
M
es
sa
ge
s
Se
qu
en
ce
D
ia
gr
am
C
el
l E
nt
ri
es
A
ct
iv
it
y
D
ia
gr
am
U
se
-C
as
e
D
ia
gr
am
U
se
C
as
e
D
es
cr
ip
ti
on
s
O
bj
ec
t
Fl
ow
s
C
on
tr
ol
F
lo
w
s
O
bj
ec
t
N
od
es
A
ct
iv
it
ie
s/
A
ct
io
ns
Fl
ow
s
St
ak
eh
ol
de
rs
R
el
at
io
ns
hi
ps
Ev
en
ts
D
el
et
e
R
ea
d
U
pd
at
e
C
re
at
e
Sc
en
ar
io
s
Fu
nc
ti
on
al
M
od
el
s
A
ct
or
s
O
bj
ec
ts
A
ss
oc
ia
te
dW
it
h
Tr
an
si
ti
on
s
St
at
es
A
ss
oc
ia
ti
on
H
av
e
U
se
C
as
es
R
es
po
ns
ib
ili
ti
es
In
st
an
ce
O
f
A
ss
oc
ia
te
d
W
it
h
FI
G
U
R
E
7
-1
6
In
te
rr
el
at
io
n
sh
ip
s
am
o
n
g
O
b
je
ct
-O
ri
en
te
d
A
n
al
ys
is
M
o
d
el
s
256
Evolving the Analysis Models into Design Models 257
fi gure that balancing all the functional, structural, and behavioral models is a very time-
consuming, tedious, and diffi cult task. However, without paying this level of attention to the
evolving models that represent the system, the models will not provide a sound foundation
on which to design and build the system.
EVOLVING THE ANALYSIS MODELS INTO DESIGN MODELS
Now that we have successfully verifi ed and validated our analysis models, we need to begin
evolving them into appropriate design models. Th e purpose of the analysis models was
to represent the underlying business problem domain as a set of collaborating objects. In
other words, the analysis activities defi ned the functional requirements. To achieve this, the
analysis activities ignored nonfunctional requirements such as performance and the system
environment issues (e.g., distributed or centralized processing, user-interface issues, and
database issues). In contrast, the primary purpose of the design models is to increase the
likelihood of successfully delivering a system that implements the functional requirements in
a manner that is aff ordable and easily maintainable. Th erefore, in systems design, we address
both the functional and nonfunctional requirements.
From an object-oriented perspective, system design models simply refi ne the system
analysis models by adding system environment (or solution domain) details to them and
refi ning the problem domain information already contained in the analysis models. When
evolving the analysis model into the design model, you should fi rst carefully review the use
cases and the current set of classes (their operations and attributes and the relationships
between them). Are all the classes necessary? Are there any missing classes? Are the classes fully
defi ned? Are any attributes or methods missing? Do the classes have any unnecessary attributes
and methods? Is the current representation of the evolving system optimal? Obviously, if we
have already verifi ed and validated the analysis models, quite a bit of this has already taken
place. Yet, object-oriented systems development is both incremental and iterative. Th erefore,
we must review the analysis models again. However, this time we begin looking at the models
of the problem domain through a design lens. In this step, we make modifi cations to the prob-
lem domain models that will enhance the effi ciency and eff ectiveness of the evolving system.
In the following sections, we introduce factoring, partitions and collaborations, and
layers as a way to evolve problem domain-oriented analysis models into optimal solu-
tion domain-oriented design models. From an enhanced Unifi ed Process perspective (see
Figure 1-16), we are moving from the analysis workfl ow to the design workfl ow, and we are
moving further into the Elaboration phase and partially into the Construction phase.
Factoring
Factoring is the process of separating out a module into a stand-alone module. Th e new
module can be a new class or a new method. For example, when reviewing a set of classes,
it may be discovered that they have a similar set of attributes and methods. Th us, it might
make sense to factor out the similarities into a separate class. Depending on whether the
new class should be in a superclass relationship to the existing classes or not, the new
class can be related to the existing classes through a generalization (a-kind-of) or possibly
through an aggregation (has-parts) relationship. Using the appointment system example,
if the Employee class had not been identified, we could possibly identify it at this stage by
factoring out the similar methods and attributes from the Nurse, Receptionist, and Doctor
classes. In this case, we would relate the new class (Employee) to the existing classes using
the generalization (a-kind-of) relationship. Obviously, by extension we also could have
created the Participant class if it had not been previously identifi ed.
Abstraction and refi nement are two processes closely related to factoring. Abstraction deals
with the creation of a higher-level idea from a set of ideas. Identifying the Employee class is an
example of abstracting from a set of lower classes to a higher one. In some cases, the abstraction
process identifi es abstract classes, whereas in other situations, it identifi es additional concrete
classes.5 Th e refi nement process is the opposite of the abstraction process. In the appointment
system example, we could identify additional subclasses of the Employee class, such as Secretary
and Bookkeeper. Of course we would add the new classes only if there were suffi cient diff erences
among them. Otherwise, the more general class, Employee, would suffi ce.
Partitions and Collaborations
Based on all the factoring, refi ning, and abstracting that can take place to the evolving system,
the sheer size of the system representation can overload the user and the developer. At this
point in the evolution of the system, it might make sense to split the representation into a
set of partitions. A partition is the object-oriented equivalent of a subsystem,6 where a sub-
system is a decomposition of a larger system into its component systems (e.g., an accounting
information system could be functionally decomposed into an accounts-payable system, an
accounts-receivable system, a payroll system, etc.). From an object- oriented perspective,
partitions are based on the pattern of activity (messages sent) among the objects in an object-
oriented system. We describe an easy approach to model partitions and collaborations later
in this chapter: packages and package diagrams.
A good place to look for potential partitions is the collaborations modeled in UML’s com-
munication diagrams (see Chapter 6). If you recall, one useful way to identify collaborations is
to create a communication diagram for each use case. However, because an individual class can
support multiple use cases, an individual class can participate in multiple use-case-based col-
laborations. In cases where classes are supporting multiple use cases, the collaborations should
be merged. Th e class diagram should be reviewed to see how the diff erent classes are related
to one another. For example, if attributes of a class have complex object types, such as Person,
Address, or Department, and these object types were not modeled as associations in the class
diagram, we need to recognize these implied associations. Creating a diagram that combines the
class diagram with the communication diagrams can be very useful to show to what degree the
classes are coupled.7 Th e greater the coupling between classes, the more likely the classes should
be grouped together in a collaboration or partition. By looking at a CRUDE matrix, we can use
CRUDE analysis (see Chapter 6) to identify potential classes on which to merge collaborations.
One of the easiest techniques to identify the classes that could be grouped to form a
collaboration is through the use of cluster analysis or multiple dimensional scaling. Th ese
statistical techniques enable the team to objectively group classes together based on their
affi nity for each other. Th e affi nity can be based on semantic relationships, diff erent types
of messages being sent between them (e.g., create, read, update, delete, or execute), or some
weighted combination of both. Th ere are many diff erent similarity measures and many
diff erent algorithms on which the clusters can be based, so one must be careful when using
these techniques. Always make sure that the collaborations identifi ed using these techniques
258 Chapter 7 Moving on to Design
5 See Chapter 5 for the diff erences between abstract and concrete classes.
6 Some authors refer to partitions as subsystems [e.g., see R. Wirfs-Brock, B. Wilkerson, and L. Weiner, Designing
Object-Oriented Soft ware (Englewood Cliff s, NJ: Prentice Hall, 1990)], whereas others refer to them as layers [e.g.,
see I. Graham, Migrating to Object Technology (Reading, MA: Addison-Wesley, 1994)]. However, we have chosen
to use the term partition [C. Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis
and Design (Englewood Cliff s, NJ: Prentice Hall, 1998)] to minimize confusion between subsystems in a traditional
systems development approach and layers associated with Rational’s Unifi ed Approach.
7 We describe the concept of coupling in Chapter 8.
Evolving the Analysis Models into Design Models 259
make sense from the problem domain perspective. Just because a mathematical algorithm
suggests that the classes belong together does not make it so. However, this is a good
approach to create a fi rst-cut set of collaborations.
Depending on the complexity of the merged collaboration, it may be useful in decompos-
ing the collaboration into multiple partitions. In this case, in addition to having collaborations
between objects, it is possible to have collaborations among partitions. Th e general rule is the
more messages sent between objects, the more likely the objects belong in the same partition. Th e
fewer messages sent, the less likely the two objects belong together.
Another useful approach to identifying potential partitions is to model each collab-
oration between objects in terms of clients, servers, and contracts. A client is an instance
of a class that sends a message to an instance of another class for a method to be executed;
a server is the instance of a class that receives the message; and a contract is the specifi ca-
tion that formalizes the interactions between the client and server objects (see Chapters 5
and 8). Th is approach allows the developer to build up potential partitions by looking at the
contracts that have been specifi ed between objects. In this case, the more contracts there are
between objects, the more likely that the objects belong in the same partition. Th e fewer con-
tracts, the less chance there is that the two classes belong in the same partition.
Remember, the primary purpose of identifying collaborations and partitions is to deter-
mine which classes should be grouped together in design.
Layers
Until this point in the development of our system, we have focused only on the problem
domain; we have totally ignored the system environment (data management, user interface,
and physical architecture). To successfully evolve the analysis model of the system into a
design model of the system, we must add the system environment information. One useful
way to do this, without overloading the developer, is to use layers. A layer represents an ele-
ment of the soft ware architecture of the evolving system. We have focused only on one layer
in the evolving soft ware architecture: the problem domain layer. Th ere should be a layer for
each of the diff erent elements of the system environment (e.g., data management, user inter-
face, physical architecture). Like partitions and collaborations, layers also can be portrayed
using packages and package diagrams (see the next section of this chapter).
Th e idea of separating the diff erent elements of the architecture into separate layers can
be traced back to the MVC architecture of Smalltalk.8 When Smalltalk was fi rst created,9
the authors decided to separate the application logic from the logic of the user interface.
In this manner, it was possible to easily develop diff erent user interfaces that worked with
the same application. To accomplish this, they created the Model–View–Controller (MVC)
architecture, where Models implemented the application logic (problem domain) and Views
and Controllers implemented the logic for the user interface. Views handled the output, and
Controllers handled the input. Because graphical user interfaces were fi rst developed in the
Smalltalk language, the MVC architecture served as the foundation for virtually all graphical
user interfaces that have been developed today (including the Mac interfaces, the Windows
family, and the various Unix-based GUI environments).
8 See S. Lewis, Th e Art and Science of Smalltalk: An Introduction to Object-Oriented Programming Using Visual-Works
(Englewood Cliff s, NJ: Prentice Hall, 1995).
9 Smalltalk was invented in the early 1970s by a soft ware-development research team at Xerox PARC. It introduced
many new ideas into the area of programming languages (e.g., object orientation, windows-based user interfaces,
reusable class library, and the development environment). In many ways, Smalltalk is the parent of all object-based
and object-oriented languages, such as Visual Basic, C++, and Java.
Based on Smalltalk’s innovative MVC architecture, many diff erent soft ware layers have
been proposed.10 We suggest the following layers on which to base soft ware architecture:
foundation, problem domain, data management, human–computer interaction, and phys-
ical architecture (see Figure 7-17). Each layer limits the types of classes that can exist on it
(e.g., only user interface classes may exist on the human–computer interaction layer).
Foundation Th e foundation layer is, in many ways, a very uninteresting layer. It contains
classes that are necessary for any object-oriented application to exist. Th ey include classes that
represent fundamental data types (e.g., integers, real numbers, characters, strings), classes that
represent fundamental data structures, sometimes referred to as container classes (e.g., lists,
trees, graphs, sets, stacks, queues), and classes that represent useful abstractions, sometimes
referred to as utility classes (e.g., date, time, money). Th ese classes are rarely, if ever, modi-
fi ed by a developer. Th ey are simply used. Today, the classes found on this layer are typically
included with the object-oriented development environments.
Problem Domain Th e problem-domain layer is what we have focused our attention on up
until now. At this stage in the development of our system, we need to further detail the classes
so that we can implement them in an eff ective and effi cient manner. Many issues need to be
addressed when designing classes, no matter on which layer they appear. For example, there
are issues related to factoring, cohesion and coupling, connascence, encapsulation, proper use
of inheritance and polymorphism, constraints, contract specifi cation, and detailed method
design. Th ese issues are discussed in Chapter 8.
Data Management Th e data management layer addresses the issues involving the persistence of
the objects contained in the system. Th e types of classes that appear in this layer deal with how
objects can be stored and retrieved. Th e classes contained in this layer are called the Data Access
and Manipulation (DAM) classes. Th e DAM classes allow the problem domain classes to be
independent of the storage used and, hence, increase the portability of the evolving system. Some
of the issues related to this layer include choice of the storage format and optimization. Th ere is
a plethora of diff erent options in which to choose to store objects. Th ese include sequential fi les,
random access fi les, relational databases, object/relational databases, object-oriented databases,
260 Chapter 7 Moving on to Design
Foundation Date, Enumeration 7, 8
Problem Domain Employee, Customer 4, 5, 6, 7, 8
Data Management DataInputStream, 8, 9
FileInputStream
Human–Computer Interaction Button, Panel 8, 10
Physical Architecture ServerSocket, URLConnection 8, 11
FIGURE 7-17
Layers and
Sample Classes
Layers Examples Relevant Chapters
10 For example, Problem Domain, Human Interaction, Task Management, and Data Management [P. Coad and
E. Yourdon, Object-Oriented Design (Englewood Cliff s, NJ: Yourdon Press, 1991)]; Domain, Application, and
Interface (I. Graham, Migrating to Object Technology [Reading, MA: Addison-Wesley, 1994)]; Domain, Service,
and Presentation [C. Larman, Applying UML and Patterns: An Introduction to Object-Oriented Analysis and
Design (Englewood Cliff s, NJ: Prentice Hall, 1998)]; Business, View, and Access [A. Bahrami, Object-Oriented
Systems Development using the Unifi ed Modeling Language (New York: McGraw-Hill, 1999)]; Application-
Specifi c, Application-General, Middleware, System-Soft ware [I. Jacobson, G. Booch, and J. Rumbaugh, Th e Uni-
fi ed Soft ware Development Process (Reading, MA: Addison-Wesley, 1999)]; Foundation, Architecture, Business,
and Application [M. Page-Jones, Fundamentals of Object-Oriented Design in UML (Reading, MA: Addison-
Wesley, 2000)].
and NoSQL data stores. Each of these options has been optimized to provide solutions for diff er-
ent access and storage problems. Today, from a practical perspective, there is no single solution
that optimally serves all applications. Th e correct solution is most likely some combination of
the diff erent storage options. A complete description of all the issues related to the data manage-
ment layer is well beyond the scope of this book.11 However, we do present the fundamentals
in Chapter 9.
Human–Computer Interaction Th e human–computer interaction layer contains classes asso-
ciated with the View and Controller idea from Smalltalk. Th e primary purpose of this layer is to
keep the specifi c user-interface implementation separate from the problem domain classes. Th is
increases the portability of the evolving system. Typical classes found on this layer include classes
that can be used to represent buttons, windows, text fi elds, scroll bars, check boxes, drop-down
lists, and many other classes that represent user-interface elements.
When designing the user interface for an application, many issues must be addressed: How
important is consistency across diff erent user interfaces? What about diff ering levels of user
experience? How is the user expected to be able to navigate through the system? What about help
systems and online manuals? What types of input elements should be included? What types of
output elements should be included? Other questions that must be addressed are related to the
platform on which the soft ware will be deployed. For example, is the application going to run
on a stand-alone computer, is it going to be distributed, or is the application going mobile? If it
is expected to run on mobile devices, what type of platform: notebooks, tablets, or phones? Will
it be deployed using Web technology, which runs on multiple devices, or will it be created using
apps that are based on Android from Google, iOS from Apple, or Windows from Microsoft ?
Depending on the answer to these questions, diff erent types of user interfaces are possible.
With the advent of social networking platforms, such as Facebook, Twitter, blogs,
YouTube, and LinkedIn, the implications for the user interface can be mind boggling.
Depending on the application, diff erent social networking platforms may be appropriate for
diff erent aspects of the application. Furthermore, each of the diff erent social networking plat-
forms enables (or prevents) consideration of diff erent types of user interfaces. Finally, with
the potential audience of your application being global, many diff erent cultural issues will
arise in the design and development of culturally aware user interfaces (such as multilingual
requirements). Obviously, a complete description of all the issues related to human–computer
interaction is beyond the scope of this book.12 However, from the user’s perspective, the user
interface is the system. We present the basic issues in user interface design in Chapter 10.
Physical Architecture Th e physical architecture layer addresses how the soft ware will exe-
cute on specifi c computers and networks. Th is layer includes classes that deal with commu-
nication between the soft ware and the computer’s operating system and the network. For
example, classes that address how to interact with the various ports on a specifi c computer
are included in this layer.
Evolving the Analysis Models into Design Models 261
11 Th ere are many good database design books that are relevant to this layer; see, for example, M. Gillenson,
Fundamentals of Database Management Systems (Hoboken, NJ: John Wiley & Sons, 2005); F. R. McFadden,
J. A. Hoff er, and Mary B. Prescott, Modern Database Management, 4th Ed. (Reading, MA: Addison-Wesley, 1998);
M. Blaha and W. Premerlani, Object-Oriented Modeling and Design for Database Applications (Englewood Cliff s,
NJ: Prentice Hall, 1998); R. J. Muller, Database Design for Smarties: Using UML for Data Modeling (San Francisco:
Morgan Kaufmann, 1999).
12 Books on user interface design that address these issues include B. Schheiderman, Designing the User Interface:
Strategies for Eff ective Human Computer Interaction, 3rd Ed. (Reading, MA: Addison-Wesley, 1998); J. Tidwell,
Designing Interfaces: Patterns for Eff ective Interaction Design, 2nd Ed. (Sebastopol, CA: O’Reilly Media, 2010);
S. Krug, Don’t Make Me Th ink: A Common Sense Approach to Web Usability (Berkeley, CA: New Riders Publishing,
2006); N. Singh and A. Pereira, Th e Culturally Customized Web Site: Customizing Web Sites for the Global Market-
place (Oxford, UK: Elsevier, 2005).
Unlike in the foundation layer, many design issues must be addressed before choosing the
appropriate set of classes for this layer. Th ese design issues include the choice of a computing
or network architecture (such as the various client-server architectures), the actual design of a
network, hardware and server soft ware specifi cation, and security issues. Other issues that must
be addressed with the design of this layer include computer hardware and soft ware confi gu-
ration (choice of operating systems, such as Linux, Mac OSX, and Windows; processor types
and speeds; amount of memory; data storage; and input/output technology), standardization,
virtualization, grid computing, distributed computing, and Web services. Th is then leads us to
one of the proverbial gorillas on the corner. What do you do with the cloud? Th e cloud is essen-
tially a form of distributed computing. In this case, the cloud allows you to treat the platform,
infrastructure, soft ware, and even business processes as remote services that can be managed by
another fi rm. In many ways, the cloud allows much of IT to be outsourced (see the discussion
of outsourcing later in this chapter). Also as brought up with the human–computer interaction
layer, the whole issue of mobile computing is very relevant to this layer. In particular, the diff erent
devices, such as phones and tablets, are relevant and the way they will communicate with each
other, such as through cellular networks or WiFi, is also important.
Finally, given the amount of power that IT requires today, the whole topic of Green IT must
be addressed. Topics that need to be addressed related to Green IT are the location of the data
center, data center cooling, alternative power sources, reduction of consumables, the idea of a
paperless offi ce, Energy Star compliance, and the potential impact of virtualization, the cloud,
and mobile computing. Like the data management and human–computer interaction layers, a
complete description of all the issues related to the physical architecture is beyond the scope of
this book.13 However, we do present the basic issues in Chapter 11.
PACKAGES AND PACKAGE DIAGRAMS
In UML, collaborations, partitions, and layers can be represented by a higher-level construct: a
package.14 In fact, a package serves the same purpose as a folder on your computer. When pack-
ages are used in programming languages such as Java, packages are actually implemented as fold-
ers. A package is a general construct that can be applied to any of the elements in UML models.
In Chapter 4, we introduced the idea of packages as a way to group use cases together to make
the use-case diagrams easier to read and to keep the models at a reasonable level of complexity.
In Chapters 5 and 6, we did the same thing for class and communication diagrams, respectively.
In this section, we describe a package diagram: a diagram composed only of packages. A package
diagram is eff ectively a class diagram that only shows packages.
Th e symbol for a package is similar to a tabbed folder (see Figure 7-18). Depending
on where a package is used, packages can participate in diff erent types of relationships. For
example, in a class diagram, packages represent groupings of classes. Th erefore, aggregation
and association relationships are possible.
In a package diagram, it is useful to depict a new relationship, the dependency rela-
tionship. A dependency relationship is portrayed by a dashed arrow (see Figure 7-18). A
dependency relationship represents the fact that a modifi cation dependency exists between
two packages. Th at is, it is possible that a change in one package could cause a change to
262 Chapter 7 Moving on to Design
13 Some books that cover these topics include S. D. Burd, Systems Architecture, 6th Ed. (Boston: Course Technology,
2011); I. Englander, Th e Architecture of Computer Hardware, Systems Soft ware, & Networking: An Information
Technology Approach (Hoboken, NJ: Wiley, 2009); K.K. Hausman and S. Cook, IT Architecture for Dummies
(Hoboken, NJ: Wiley Publishing, 2011).
14 Th is discussion is based on material in Chapter 7 of M. Fowler with K. Scott, UML Distilled: A Brief Guide to the
Standard Object Modeling Language, 3rd Ed. (Reading, MA: Addison-Wesley, 2004).
Packages and Package Diagrams 263
be required in another package. Figure 7-19 portrays the dependencies among the diff erent
layers (foundation, problem domain, data management, human–computer interaction, and
physical architecture). For example, if a change occurs in the problem domain layer, it most
likely will cause changes to occur in the human–computer interaction, physical architecture,
and data management layers. Notice that these layers point to the problem domain layer and
therefore are dependent on it. However, the reverse is not true.15 Also note that all layers
are dependent upon the foundation layer. Th is is due to the contents of the foundation layer
being the fundamental classes from which all other classes will be built. Consequently, any
changes made to this layer could have ramifi cabitons to all other layers.
At the class level, there could be many causes for dependencies among classes. For
example, if the protocol for a method is changed, then this causes the interface for all
A package:
■ Is a logical grouping of UML elements.
■ Is used to simplify UML diagrams by grouping related elements into a single
higher-level element.
A dependency relationship:
■ Represents a dependency between packages: If a package is changed, the
dependent package also could have to be modified.
■ Has an arrow drawn from the dependent package toward the package on which it
is dependent.
Package
FIGURE 7-18 Syntax for Package Diagram
Human–Computer Interaction
Problem Domain
Physical Architecture Data Management
Foundation
FIGURE 7-19
Package Diagram
of Dependency
Relationships
among Layers
15 A useful side eff ect of the dependencies among the layers is that the project manager can divide the project team up
into separate subteams: one for each design layer. Th is is possible because each of the design layers is dependent on
the problem domain layer, which has been the focus of analysis. In design, the team can gain some productivity-based
effi ciency by working on the diff erent layer designs in parallel.
264 Chapter 7 Moving on to Design
objects of this class to change. Th erefore, all classes that have objects that send messages
to the instances of the modifi ed class might have to be modifi ed. Capturing dependency
relationships among the classes and packages helps the organization in maintaining
object-oriented information systems.
Collaborations, partitions, and layers are modeled as packages in UML. Collaborations
are normally factored into a set of partitions, which are typically placed on a layer. Partitions
can be composed of other partitions. Also, it is possible to have classes in partitions, which are
contained in another partition, which is placed on a layer. All these groupings are represented
using packages in UML. Remember that a package is simply a generic grouping construct
used to simplify UML models through the use of composition.16
A simple package diagram, based on the appointment system example from the previ-
ous chapters, is shown in Figure 7-20. Th is diagram portrays only a very small portion of the
entire system. In this case, we see that the Patient UI, Patient-DAM, and Patient Table classes
depend on the Patient class. Furthermore, the Patient-DAM class depends on the Patient Table
class. Th e same can be seen with the classes dealing with the actual appointments. By isolating
the Problem Domain classes (such as the Patient and Appt classes) from the actual object-
persistence classes (such as the Patient Table and Appt Table classes) through the use of the
intermediate Data Management classes (Patient-DAM and Appt-DAM classes), we isolate the
Problem Domain classes from the actual storage medium.17 Th is greatly simplifi es the main-
tenance and increases the reusability of the Problem Domain classes. Of course, in a complete
description of a real system, there would be many more dependencies.
Guidelines for Creating Package Diagrams
As with the UML diagrams described in the earlier chapters, we provide a set of guidelines
that we have adapted from Ambler to create package diagrams.18 In this case, we off er six
guidelines.
■ Use package diagrams to logically organize designs. Specifi cally, use packages to
group classes together when there is an inheritance, aggregation, or composition
relationship between them or when the classes form a collaboration.
■ In some cases, inheritance, aggregation, or association relationships exist
between packages. In those cases, for readability purposes, try to support inher-
itance relationships vertically, with the package containing the superclass being
placed above the package containing the subclass. Use horizontal placement
to support aggregation and association relationships, with the packages being
placed side by side.
■ When a dependency relationship exists on a diagram, it implies that there is
at least one semantic relationship between elements of the two packages. Th e
direction of the dependency is typically from the subclass to the superclass,
from the whole to the part, and with contracts, from the client to the server. In
other words, a subclass is dependent on the existence of a superclass, a whole is
dependent upon its parts existing, and a client can’t send a message to a nonexist-
ent server.
16 For those familiar with traditional approaches, such as structured analysis and design, packages serve a similar
purpose as the leveling and balancing processes used in data fl ow diagramming.
17 Th ese issues are described in more detail in Chapter 9.
18 S. W. Ambler, Th e Elements of UML 2.0 Style (Cambridge, UK: Cambridge University Press, 2005).
■ When using packages to group use cases together, be sure to include the actors and
the associations that they have with the use cases grouped in the package. Th is will
allow the diagram’s user to better understand the context of the diagram.
■ Give each package a simple, but descriptive name to provide the package dia-
gram user with enough information to understand what the package encapsulates.
Otherwise, the user will have to drill-down or open up the package to understand
the package’s purpose.
■ Be sure that packages are cohesive. For a package to be cohesive, the classes con-
tained in the package, in some sense, belong together. A simple, but not perfect, rule
to follow when grouping classes together in a package is that the more the classes
depend on each other, the more likely they belong together in a package.
HCI Layer
Appt Sys UI
Patient UI Appt UI
PD Layer
Appt Sys
Patient Appt
DM Layer
Patient-DAM Appt-DAM
Patient Table Appt Table
FIGURE 7-20
Partial Package
Diagram of the
Appointment System
Packages and Package Diagrams 265
266 Chapter 7 Moving on to Design
Creating Package Diagrams
In this section, we describe a simple fi ve-step process to create package diagrams. Th e fi rst
step is to set the context for the package diagram. Remember, packages can be used to model
partitions and/or layers. Revisiting the appointment system again, let’s set the context as the
problem domain layer.
Th e second step is to cluster the classes together into partitions based on the relationships
that the classes share. Th e relationships include generalization, aggregation, the various
associations, and the message sending that takes place between the objects in the system.
To identify the packages in the appointment system, we should look at the diff erent anal-
ysis models [e.g., the class diagram (see Figure 7-15), the communication diagrams (see
Figure 7-10)], and the CRUDE matrix (see Figure 7-12). Classes in a generalization hierarchy
should be kept together in a single partition.
Th e third step is to place the clustered classes together in a partition and model the partitions
as packages. Figure 7-21 portrays fi ve packages in the PD Layer: Account Pkg, Participant
Pkg, Patient Pkg, Appointment Pkg, and Treatment Pkg.
Th e fourth step is to identify the dependency relationships among the packages. We accom-
plish this by reviewing the relationships that cross the boundaries of the packages to uncover
potential dependencies. In the appointment system, we see association relationships that
connect the Account Pkg with the Appointment Pkg (via the associations between the Entry
class and the Appointment class), the Participant Pkg with the Appointment Pkg (via the
association between the Doctor class and the Appointment class), the Patient Pkg, which is
contained within the Participant Pkg, with the Appointment Pkg (via the association between
the Patient and Appointment classes), and the Patient Pkg with the Treatment Pkg (via the
association between the Patient and Symptom classes).
Th e fi fth step is to lay out and draw the diagram. Using the guidelines, place the packages
and dependency relationships in the diagram. In the case of the Appointment system, there
are dependency relationships between the Account Pkg and the Appointment Pkg, the
Participant Pkg and the Appointment Pkg, the Patient Pkg and the Appointment Pkg, and
the Patient Pkg and the Treatment Pkg. To increase the understandability of the dependency
relationships among the diff erent packages, a pure package diagram that shows only the
dependency relationships among the packages can be created (see Figure 7-22).
Verifying and Validating Package Diagrams
Like all the previous models, package diagrams need to be verifi ed and validated. In this case,
the package diagrams were derived primarily from the class diagram, the communications
diagrams, and the CRUDE matrix. Only two areas need to be reviewed.
First, the identifi ed packages must make sense from a problem domain point of view. For
example, in the context of an appointment system, the packages in Figure 7-22 (Participant,
Patient, Appt, Account, and Treatment) seem to be reasonable.
Second, all dependency relationships must be based on message-sending relationships on
the communications diagram, cell entries in the CRUDE matrix, and associations on the class
diagram. In the case of the appointment system, the identifi ed dependency relationships are
reasonable (see Figures 7-10, 7-12, 7-15, and 7-22).
1. Set Context
2. Cluster Classes
3. Create Packages
4. Identify
Dependencies
5. Lay Out and Draw
Diagram
267
PD
L
ay
er
A
cc
ou
nt
Pk
g
A
cc
ou
nt
ha
s
1.
.1
1.
.1
..*
1.
.1
1.
.1
1.
.1
1.
.1
1.
.1
0.
.1
1.
.1
1.
.1
1.
.1
1.
.1
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
1.
.*
0.
.*
0.
.*
1.
.*
1.
.*
1.
.*
ha
s
It
em
co
nt
ai
ns
En
tr
y
D
eb
it
C
re
di
t
A
pp
oi
nt
m
en
t
Pk
g
Pa
rt
ic
ip
an
t
Pk
g
-t
im
e
-d
at
e
-r
ea
so
n
+
ca
nc
el
w
ith
ou
t n
ot
ic
e(
)
lo
ca
te
dA
t Pl
ac
e
Pr
es
cr
ip
ti
on
G
oo
d
Se
rv
ic
e
B
ra
ce
Ph
ys
ic
al
C
he
ck
up
Tr
an
sa
ct
io
n
Li
ne
I
te
m
A
ss
ig
ne
dT
o
Pa
ti
en
t
Pk
g
Pa
rt
ic
ip
an
t
-l
as
tn
am
e
-fi
rs
tn
am
e
-a
dd
re
ss
-p
ho
ne
-b
ir
th
da
te
-/
a
ge
schedules
Em
pl
oy
ee
D
oc
to
r
R
ec
ep
ti
on
is
t
N
ur
se
Pa
ti
en
t
-a
m
ou
nt
-i
ns
ur
an
ce
c
ar
ri
er
+
m
ak
e
ap
po
in
tm
en
t()
+
ca
lc
ul
at
e
la
st
v
is
it(
)
+
ch
an
ge
s
ta
tu
s(
)
+
pr
ov
id
es
m
ed
ic
al
h
is
to
ry
()
su
ffe
rs
pr
ov
id
es
+
p
ri
m
ar
y
in
su
ra
nc
e
ca
rr
ie
r
U
np
ai
d
B
ill
M
ed
ic
al
H
is
to
ry
-h
ea
rt
d
is
ea
se
-h
ig
h
bl
oo
d
pr
es
su
re
-d
ia
be
te
s
-a
lle
rg
ie
s
Sy
m
pt
om
-d
es
cr
ip
tio
n
Tr
ea
tm
en
t
m
ed
ic
at
io
n
in
st
ru
ct
io
ns
sy
m
pt
om
s
ev
er
ity
-n
am
e
Il
ln
es
s
ow
es
A
pp
oi
nt
m
en
t
Tr
ea
tm
en
t
Pk
g
FI
G
U
R
E
7
-2
1
Pa
ck
ag
e
D
ia
gr
am
o
f
th
e
PD
L
ay
er
o
f
th
e
A
p
p
o
in
tm
en
t
Pr
o
b
le
m
268 Chapter 7 Moving on to Design
PD Layer
Account Pkg
Patient Pkg
Appointment
Pkg
Treatment
Pkg
Participant
Pkg
FIGURE 7-22 Overview Package Diagram of the PD Layer for the Appointment System
DESIGN STRATEGIES
Until now, we have assumed that the system will be built and implemented by the project
team; however, there are actually three ways to approach the creation of a new system: devel-
oping a custom application in-house, buying and customizing a packaged system, and rely-
ing on an external vendor, developer, or service provider to build the system. Each of these
choices has strengths and weaknesses, and each is more appropriate in diff erent scenarios.
Th e following sections describe each design choice in turn, and then we present criteria that
you can use to select one of the three approaches for your project.
Custom Development
Many project teams assume that custom development, or building a new system from scratch,
is the best way to create a system. For one thing, teams have complete control over the way
the system looks and functions. Custom development also allows developers to be fl exible
and creative in the way they solve business problems. Additionally, a custom application is
easier to change to include components that take advantage of current technologies that can
support such strategic eff orts.
Design Strategies 269
Building a system in-house also builds technical skills and functional knowledge within
the company. As developers work with business users, their understanding of the business
grows and they become better able to align IS with strategies and needs. Th ese same develop-
ers climb the technology learning curve so that future projects applying similar technology
require much less eff ort.
Custom application development, however, requires dedicated eff ort that involves long
hours and hard work. Many companies have a development staff who already is overcom-
mitted to fi lling huge backlogs of systems requests and just does not have time for another
project. Also, a variety of skills—technical, interpersonal, functional, project management,
and modeling—must be in place for the project to move ahead smoothly. IS professionals,
especially highly skilled individuals, are quite diffi cult to hire and retain.
Th e risks associated with building a system from the ground up can be quite high, and
there is no guarantee that the project will succeed. Developers could be pulled away to work
on other projects, technical obstacles could cause unexpected delays, and the business users
could become impatient with a growing timeline.
Packaged Soft ware
Many business needs are not unique, and because it makes little sense to reinvent the wheel,
many organizations buy packaged soft ware that has already been written rather than develop-
ing their own custom solution. In fact, there are thousands of commercially available soft ware
programs that have already been written to serve a multitude of purposes. Th ink about your
own need for a word processor—did you ever consider writing your own word processing
soft ware? Th at would be very silly considering the number of good soft ware packages availa-
ble that are relatively inexpensive.
Similarly, most companies have needs that can be met quite well by packaged soft –
ware, such as payroll or accounts receivable. It can be much more effi cient to buy pro-
grams that have already been created, tested, and proven. Moreover, a packaged system
can be bought and installed in a relatively short time when compared with a custom
system. Plus, packaged systems incorporate the expertise and experience of the vendor
who created the soft ware.
Packaged soft ware can range from reusable components to small, single-function
tools to huge, all-encompassing systems such as enterprise resource planning (ERP) appli-
cations that are installed to automate an entire business. Implementing ERP systems is a
process in which large organizations spend millions of dollars installing packages by com-
panies such as SAP or Oracle and then change their businesses accordingly. Installing ERP
soft ware is much more diffi cult than installing small application packages because benefi ts
can be harder to realize and problems are much more serious.
However, there are problems related to packaged soft ware. For example, companies
buying packaged systems must accept the functionality that is provided by the system, and
rarely is there a perfect fi t. If the packaged system is large in scope, its implementation could
mean a substantial change in the way the company does business. Letting technology drive
the business can be dangerous.
Most packaged applications allow customization, or the manipulation of system param-
eters to change the way certain features work. For example, the package might have a way to
accept information about your company or the company logo that would then appear on input
screens. Or an accounting soft ware package could off er a choice of various ways to handle cash
fl ow or inventory control so that it can support the accounting practices in diff erent organiza-
tions. If the amount of customization is not enough and the soft ware package has a few fea-
tures that don’t quite work the way the company needs it to work, the project team can create
workarounds.
270 Chapter 7 Moving on to Design
A workaround is a custom-built add-on program that interfaces with the packaged appli-
cation to handle special needs. It can be a nice way to create needed functionality that does not
exist in the soft ware package. But workarounds should be a last resort for several reasons. First,
workarounds are not supported by the vendor who supplied the packaged soft ware, so upgrades
to the main system might make the workaround ineff ective. Also, if problems arise, vendors
have a tendency to blame the workaround as the culprit and refuse to provide support.
Although choosing a packaged soft ware system is simpler than custom development, it
too can benefi t from following a formal methodology, just as if a custom application were
being built.
Systems integration refers to the process of building new systems by combining pack-
aged software, existing legacy systems, and new software written to integrate these. Many
consulting firms specialize in systems integration, so it is not uncommon for companies
to select the packaged software option and then outsource the integration of a variety of
packages to a consulting firm. (Outsourcing is discussed in the next section.)
Th e key challenge in systems integration is fi nding ways to integrate the data produced by
the diff erent packages and legacy systems. Integration oft en hinges on taking data produced by
one package or system and reformatting it for use in another package or system. Th e project team
starts by examining the data produced by and needed by the diff erent packages or systems and
identifying the transformations that must occur to move the data from one to the other. In many
cases, this involves fooling the diff erent packages or systems into thinking that the data were
produced by an existing program module that the package or system expects to produce the data
rather than the new package or system that is being integrated.
A third approach is through the use of an object wrapper.19 An object wrapper is essen-
tially an object that “wraps around” a legacy system, enabling an object-oriented system to
send messages to the legacy system. Eff ectively, object wrappers create an application pro-
gram interface (API) to the legacy system. Th e creation of an object wrapper protects the
corporation’s investment in the legacy system.
Outsourcing
Th e design choice that requires the least amount of in-house resources is outsourcing—
hiring an external vendor, developer, or service provider to create the system. Outsourcing
has become quite popular in recent years. Some estimate that as many as 50 percent of
companies with IT budgets of more than $5 million are currently outsourcing or evaluat-
ing the approach.
With outsourcing, the decision making and/or management control of a business
function is transferred to an outside supplier. Th is transfer requires two-way coordination,
exchange of information, and trust between the supplier and the business. From an IT per-
spective, IT outsourcing can include hiring consultants to solve a specifi c problem, hiring
contract programmers to implement a solution, hiring a fi rm to manage the IT function and
assets of a company, or actually outsourcing the entire IT function to a separate fi rm. Today,
through the use of application service providers (ASPs), Web services technology, and cloud
services, it is possible to use a pay-as-you-go approach for a soft ware package.20 Essentially,
IT outsourcing involves hiring a third party to perform some IT function that traditionally
would be performed in-house.
Th ere can be great benefi t to having someone else develop a company’s system. Th e out-
side company may be more experienced in the technology or have more resources, such as
19 Ian Graham, Object-Oriented Methods: Principles & Practice, 3rd Ed. (Reading, MA: Addison-Wesley, 2001).
20 For an economic explanation of how this could work, see H. Baetjer, Soft ware as Capital: An Economic Perspective
on Soft ware Engineering (Los Alamitos, CA: IEEE Computer Society Press, 1997).
experienced programmers. Many companies embark upon outsourcing deals to reduce costs,
whereas others see it as an opportunity to add value to the business.
For whatever reason, outsourcing can be a good alternative for a new system. However,
it does not come without costs. If you decide to leave the creation of a new system in the
hands of someone else, you could compromise confi dential information or lose control
over future development. In-house professionals are not benefi ting from the skills that
could be learned from the project; instead, the expertise is transferred to the outside organ-
ization. Ultimately, important skills can walk right out the door at the end of the contract.
Furthermore, when off shore outsourcing is being considered, we must also be cognizant
of language issues, time-zone diff erences, and cultural diff erences (e.g., acceptable business
practices as understood in one country that may be unacceptable in another). All these
concerns, if not dealt with properly, can prevail over any advantage that outsourcing or off –
shore outsourcing could realize.
Most risks can be addressed if a company decides to outsource, but two are par-
ticularly important. First, the company must thoroughly assess the requirements for the
project—a company should never outsource what is not understood. If rigorous planning
and analysis has occurred, then the company should be well aware of its needs. Second, the
company should carefully choose a vendor, developer, or service with a proven track record
with the type of system and technology that its system needs.
Th ree primary types of contracts can be drawn to control the outsourcing deal. A
time-and-arrangements contract is very fl exible because a company agrees to pay for whatever
time and expenses are needed to get the job done. Of course, this agreement could result in a
large bill that exceeds initial estimates. Th is works best when the company and the outsourcer
are unclear about what it is going to take to fi nish the job.
A company will pay no more than expected with a fi xed-price contract because if the out-
sourcer exceeds the agreed-upon price, it will have to absorb the costs. Outsourcers are much
more careful about defi ning requirements clearly up front, and there is little fl exibility for change.
Th e type of contract gaining in popularity is the value-added contract, whereby the out-
sourcer reaps some percentage of the completed system’s benefi ts. Th e company has very little
risk in this case, but it must expect to share the wealth once the system is in place.
Creating fair contracts is an art because fl exibility must be carefully balanced with clearly
defi ned terms. Oft en, needs change over time. Th erefore, the contract should not be so specifi c and
rigid that alterations cannot be made. Th ink about how quickly mobile technology has changed. It
is diffi cult to foresee how a project might evolve over a long period of time. Short-term contracts
help leave room for reassessment if needs change or if relationships are not working out the way
both parties expected. In all cases, the relationship with the outsourcer should be viewed as a part-
nership where both parties benefi t and communicate openly.
Managing the outsourcing relationship is a full-time job. Th us, someone needs to be assigned
full time to manage the outsourcer, and the level of that person should be appropriate for the size
of the job (a multimillion dollar outsourcing engagement should be handled by a high-level
executive). Th roughout the relationship, progress should be tracked and measured against prede-
termined goals. If a company does embark upon an outsourcing design strategy, it should be sure
to get adequate information. Many books have been written that provide much more detailed
information on the topic.21 Figure 7-23 summarizes some guidelines for outsourcing.
21 For more information on outsourcing, we recommend M. Lacity and R. Hirschheim, Information Systems Out-
sourcing: Myths, Metaphors, and Realities (New York, NY: Wiley, 1993); L. Willcocks and G. Fitzgerald, A Business
Guide to Outsourcing Information Technology (London: Business Intelligence, 1994); E. Carmel, Off shoring Infor-
mation Technology: Sourcing and Outsourcing to a Global Workforce (Cambridge, England: Cambridge University
Press, 2005); J. K. Halvey and B. M. Melby, Information Technology Outsourcing Transactions: Process, Strategies, and
Contracts, 2nd Ed. (Hoboken, NJ: Wiley, 2005); T. L. Friedman, Th e World Is Flat: A Brief History of the Twenty-First
Century, Updated and Expanded Edition (New York: Farrar, Straus, and Giroux, 2006).
Design Strategies 271
272 Chapter 7 Moving on to Design
Selecting a Design Strategy
Each of the design strategies just discussed has its strengths and weaknesses, and no one strat-
egy is inherently better than the others. Th us, it is important to understand the strengths and
weaknesses of each strategy and when to use each. Figure 7-24 summarizes the characteristics
of each strategy.
Business Need If the business need for the system is common and technical solutions
already exist that can meet the business need of the system, it makes little sense to build a
custom application. Packaged systems are good alternatives for common business needs.
A custom alternative should be explored when the business need is unique or has special
requirements. Usually, if the business need is not critical to the company, then outsourcing is
the best choice—someone outside of the organization can be responsible for the application
development.
In-house Experience If in-house experience exists for all the functional and technical needs
of the system, it will be easier to build a custom application than if these skills do not exist.
A packaged system may be a better alternative for companies that do not have the technical
skills to build the desired system. For example, a project team that does not have mobile tech-
nology skills might want to consider outsourcing those aspects of the system.
Project Skills Th e skills that are applied during projects are either technical (e.g., Java, SQL)
or functional (e.g., security), and diff erent design alternatives are more viable, depending on
• Keep the lines of communication open between you and your outsourcer.
• Defi ne and stabilize requirements before signing a contact.
• View the outsourcing relationship as a partnership.
• Select the vendor, developer or service provider carefully.
• Assign a person to managing the relationship.
• Don’t outsource what you don’t understand.
• Emphasize fl exible requirements, long-term relationships and short-term contracts.
FIGURE 7-23
Outsourcing
Guidelines
Outsourcing
Business Need The business need is unique. The business need is common. The business need is not
core to the business.
In-house Experience In-house functional and In-house functional In-house functional or technical
technical experience exists. experience exists. experience does not exist.
Project Skills There is a desire to The skills are not strategic. The decision to outsource is a
build in-house skills. strategic decision.
Project Management The project has a highly The project has a project The project has a highly skilled
skilled project manager manager who can coordinate project manager at the level of
and a proven methodology. the vendor’s efforts. the organization that matches the
scope of the outsourcing deal.
Time frame The time frame is fl exible. The time frame is short. The time frame is short or fl exible.
Use Custom Use a Packaged Use Outsourcing
Development When… System When… When…
FIGURE 7-24 Selecting a Design Strategy
Selecting an Acquisition Strategy 273
how important the skills are to the company’s strategy. For example, if certain functional and
technical expertise that relates to mobile application development is important to an organi-
zation because it expects mobile to play an important role in its sales over time, then it makes
sense for the company to develop mobile applications in-house, using company employees
so that the skills can be developed and improved. On the other hand, some skills, such as
network security, may be beyond the technical expertise of employees or not of interest to the
company’s strategists—it is just an operational issue that needs to be addressed. In this case,
packaged systems or outsourcing should be considered so that internal employees can focus
on other business-critical applications and skills.
Project Management Custom applications require excellent project management and
a proven methodology. So many things, such as funding obstacles, staffi ng holdups, and
overly demanding business users, can push a project off -track. Th erefore, the project team
should choose to develop a custom application only if it is certain that the underlying
coordination and control mechanisms will be in place. Packaged and outsourcing alternatives
also need to be managed; however, they are more shielded from internal obstacles because the
external parties have their own objectives and priorities (e.g., it may be easier for an outside
contractor to say no to a user than it is for a person within the company). Typically, packaged
and outsourcing alternatives have their own methodologies, which can benefi t companies that
do not have an appropriate methodology to use.
Time Frame When time is a factor, the project team should probably start looking for a
system that is already built and tested. In this way, the company will have a good idea of
how long the package will take to put in place and what the fi nal result will contain. Th e
time frame for custom applications is hard to pin down, especially when you consider how
many projects end up missing important deadlines. If a company must choose the custom
development alternative and the time frame is very short, it should consider using techniques
such as timeboxing to manage this problem. Th e time to produce a system using outsourcing
really depends on the system and the outsourcer’s resources. If a service provider has ser-
vices in place that can be used to support the company’s needs, then a business need could
be implemented quickly. Otherwise, an outsourcing solution could take as long as a custom
development initiative.
SELECTING AN ACQUISITION STRATEGY
Once the project team has a good understanding of how well each design strategy fi ts with the
project’s needs, it must begin to understand exactly how to implement these strategies. For exam-
ple, what tools and technology would be used if a custom alternative were selected? What vendors
make packaged systems that address the project’s needs? What service providers would be able to
build this system if the application were outsourced? Th is information can be obtained from peo-
ple working in the IS department and from recommendations by business users. Alternatively,
the project team can contact other companies with similar needs and investigate the types of
systems that they have put in place. Vendors and consultants usually are willing to provide
information about various tools and solutions in the form of brochures, product demonstrations,
and information seminars. However, a company should be sure to validate the information it
receives from vendors and consultants. Aft er all, they are trying to make a sale. Th erefore, they
may stretch the capabilities of their tool by focusing on only the positive aspects of the tool while
omitting the tool’s drawbacks.
It is likely that the project team will identify several ways that a system could be con-
structed aft er weighing the specifi c design options. For example, the project team might have
274 Chapter 7 Moving on to Design
found three vendors that make packaged systems that potentially could meet the project’s
needs. Or the team may be debating over whether to develop a system using Java as a devel-
opment tool and the database management system from Oracle or to outsource the develop-
ment eff ort to a consulting fi rm such as Accenture or CGI. Each alternative has pros and cons
associated with it that need to be considered, and only one solution can be selected in the end.
To aid in this decision, additional information should be collected. Project teams
employ several approaches to gather additional information that is needed. One helpful
tool is the request for proposal (RFP), a document that solicits a formal proposal from
a potential vendor, developer, or service provider. RFPs describe in detail the system or
service that is needed, and vendors respond by describing in detail how they could supply
those needs.
Although there is no standard way of writing an RFP, it should include certain key facts
that the vendor requires, such as a detailed description of needs, any special technical needs
or circumstances, evaluation criteria, procedures to follow, and a timetable. In a large project,
the RFP can be hundreds of pages long, since it is essential that all required project details
are included.
Th e RFP is not just a way to gather information. Rather, it results in a vendor proposal
that is a binding off er to accomplish the tasks described in the RFP. Th e vendor proposal
includes a schedule and a price for which the work is to be performed. Once the winning
vendor proposal is chosen, a contract for the work is developed and signed by both parties.
For smaller projects with smaller budgets, the request for information (RFI) may be
suffi cient. An RFI is a shorter, less detailed request that is sent to potential vendors to obtain
general information about their products and services. Sometimes, the RFI is used to deter-
mine which vendors have the capability to perform a service. It is oft en then followed up with
an RFP to the qualifi ed vendors.
When a list of equipment is so complete that the vendor need only provide a price, with-
out any analysis or description of what is needed, the request for quote (RFQ) may be used.
For example, if twenty long-range RFID tag readers are needed from the manufacturer on a
certain date at a certain location, the RFQ can be used. If an item is described, but a specifi c
manufacturer’s product is not named, then extensive testing will be required to verify fulfi ll-
ment of the specifi cations.
Alternative Matrix
An alternative matrix can be used to organize the pros and cons of the design alternatives
so that the best solution will be chosen in the end (see Figure 7-25). Th is matrix is created
using the same steps as the feasibility analysis, which was presented in Chapter 2. Th e only
diff erence is that the alternative matrix combines several feasibility analyses into one matrix
so that the alternatives can easily be compared. An alternative matrix is a grid that contains
the technical, budget, and organizational feasibilities for each system candidate, pros and cons
associated with adopting each solution, and other information that is helpful when making
comparisons. Sometimes weights are provided for diff erent parts of the matrix to show when
some criteria are more important to the fi nal decision.
To create the alternative matrix, draw a grid with the alternatives across the top and dif-
ferent criteria (e.g., feasibilities, pros, cons, and other miscellaneous criteria) along the side.
Next, fi ll in the grid with detailed descriptions about each alternative. Th is becomes a useful
document for discussion because it clearly presents the alternatives being reviewed and com-
parable characteristics for each one.
Sometimes, weights and scores are added to the alternative matrix to create a weighted
alternative matrix that communicates the project’s most important criteria and the alternatives
that best address them. A scorecard is built by adding a column labeled “weight” that includes
Selecting an Acquisition Strategy 275
a number depicting how much each criterion matters to the fi nal decision. Typically, analysts
take 100 points and spread them out across the criteria appropriately. If fi ve criteria were used
and all mattered equally, then each criterion would receive a weight of 20. However, if costs
were the most important criterion for choosing an alternative, it might receive 60 points, and
the other four criteria might get only 10 points each.
Th en, the analysts add to the matrix a column called “Score” that communicates how well
each alternative meets the criteria. Usually, number ranges like 1 to 5 or 1 to 10 are used to
rate the appropriateness of the alternatives by the criteria. So, for the cost criterion, the least
expensive alternative may receive a 5 on a 1-to-5 scale, whereas a costly alternative would
receive a 1. Weighted scores are computed with each criterion’s weight multiplied by the score
it was given for each alternative. Th en, the weighted scores are totaled for each alternative.
Th e highest weighted score achieves the best match for our criteria. When numbers are used
in the alternative matrix, project teams can make decisions quantitatively and on the basis of
hard numbers.
It should be pointed out, however, that the score assigned to the criteria for each alternative
is nothing more than a subjective assignment. Consequently, it is entirely possible for an analyst
to skew the analysis according to his or her own biases. In other words, the weighted alternative
matrix can be made to support whichever alternative you prefer and yet retains the appearance
of an objective, rational analysis. To avoid the problem of a biased analysis, each analyst on the
team could develop ratings independently; then, the ratings could be compared and discrepan-
cies resolved in an open team discussion.
Th e fi nal step, of course, is to decide which solution to design and implement. Th e deci-
sion should be made by a combination of business users and technical professionals aft er
the issues involved with the diff erent alternatives are well understood. Once the decision is
fi nalized, design can continue as needed, based on the selected alternative.
Technical
Issues:
Criterion 1 20 5 100 3 60 3 60
Criterion 2 10 3 30 3 30 5 50
Criterion 3 10 2 20 1 10 3 30
Economic
Issues:
Criterion 4 25 Supporting 3 75 Supporting 3 75 Supporting 5 125
Criterion 5 10 Information 3 30 Information 1 10 Information 5 50
Organizational
Issues:
Criterion 6 10 5 50 5 50 3 30
Criterion 7 10 3 30 3 30 1 10
Criterion 8 5 3 15 1 5 1 5
TOTAL 100 350 270 360
* This denotes how well the alternative meets the criteria. 1 = poor fi t; 5 = perfect fi t.
Relative Alternative 1: Alternative 2: Alternative 3:
Evaluation Importance Custom Score Weighted Custom Score Weighted Packaged Score Weighted
Criteria (Weight) Application (1–5)* Score Application (1–5)* Score Software (1–5)* Score
Using VB.NET Using Java Product ABC
FIGURE 7-25 Sample Alternative Matrix Using Weights
▲
▲
▲
▲
▲
▲
276 Chapter 7 Moving on to Design
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Th e team had one major task to complete before moving into design. Aft er developing and
verifying the functional, structural, and behavioral models, they now had to validate that
the functional, structural, and behavioral models developed in analysis agreed with each
other. In other words, they needed to balance the functional, structural, and behavioral
models. As you will see, this activity revealed inconsistencies and uncovered new informa-
tion about the system that they hope to implement. Aft er creating corrected iterations of
each of the three types of models, the team explored design alternatives and determined a
design strategy.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the purpose of balancing the analysis models.
Balance the functional models with the structural models.
Balance the functional models with the behavioral models.
Balance the structural models with the behavioral models.
Describe the purpose of the factoring, refi nement, and abstraction processes.
Describe the purpose of partitions and collaborations.
Name and describe the layers.
Explain the purpose of a package diagram.
Describe the diff erent elements of the package diagram.
Create a package diagram to model partitions and layers.
Verify and validate package diagrams using walkthroughs.
Describe the pros and cons of the three basic design strategies.
Describe the basis of selecting a design strategy.
Explain how and when to use RFPs, RFIs, and RFQs to gather information from vendors.
Describe how to use a weighted alternative matrix to select an acquisition strategy.
KEY TERMS
A-kind-of
Abstract classes
Abstraction
Aggregation
Alternative matrix
Balancing the models
Class
Client
Collaboration
Concrete classes
Contract
Controller
Custom development
Customization
Data management layer
Dependency relationship
Enterprise resource
systems (ERP)
Factoring
Fixed-price contract
Foundation layer
Generalization
Has-parts
Human–computer
interaction layer
Layer
Message
Method
Model
Model-View-Controller
(MVC)
Module
Object wrapper
Outsourcing
Package
Package diagram
Packaged soft ware
Partition
Physical architecture
layer
Problem domain layer
Refi nement
Request for information
(RFI)
Request for proposals
(RFP)
Server
Smalltalk
Systems integration
Time-and-arrangements
contract
Validation
Value-added contract
Verifi cation
View
Workaround
Exercises 277
QUESTIONS
1. Explain the primary diff erence between an analysis
model and a design model.
2. What is meant by balancing the models?
3. What are the interrelationships among the functional,
structural, and behavioral models that need to be tested?
4. What does factoring mean? How is it related to
abstraction and refi nement?
5. What is a partition? How does a partition relate to a
collaboration?
6. What is a layer? Name the diff erent layers.
7. What is the purpose of the diff erent layers?
8. Describe the diff erent types of classes that can appear
on each of the layers.
9. What issues or questions arise on each of the diff erent
layers?
10. What is a package? How are packages related to parti-
tions and layers?
11. What is a dependency relationship? How do you iden-
tify them?
12. What are the fi ve steps for identifying packages and
creating package diagrams?
13. What needs to be verifi ed and validated in package
diagrams?
14. When drawing package diagrams, what guidelines
should you follow?
15. What situations are most appropriate for a custom
development design strategy?
16. What are some problems with using a packaged soft –
ware approach to building a new system? How can
these problems be addressed?
17. Why do companies invest in ERP systems?
18. What are the pros and cons of using a workaround?
19. When is outsourcing considered a good design strat-
egy? When is it not appropriate?
20. What is an object wrapper?
21. What is systems integration? Explain the challenges.
22. What are the diff erences between the time-and-
arrangements, fi xed-price, and value-added contracts
for outsourcing?
23. How are the alternative matrix and feasibility analysis
related?
24. What is an RFP? How is this diff erent from an RFI?
EXERCISES
A. For the A Real Estate Inc. problem in Chapters 4 (exer-
cises I, J, and K), 5 (exercises P and Q), and 6 (exercise D):
1. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
2. Using the communication diagrams and the CRUDE
matrix, create a package diagram of the problem
domain layer.
3. Perform a verifi cation and validation walkthrough
of the package diagram.
4. Based on the analysis models that have been created
and your current understanding of the fi rm’s posi-
tion, what design strategy would you recommend?
Why?
B. For the A Video Store problem in Chapters 4 (exercises
L, M, and N), 5 (exercises R and S), and 6 (exercise E):
1. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
2. Using the communication diagrams and the CRUDE
matrix, create a package diagram of the problem
domain layer.
3. Perform a verifi cation and validation walkthrough
of the package diagram.
4. Based on the analysis models that have been
created and your current understanding of the
fi rm’s position, what design strategy would you
recommend? Why?
C. For the health club membership problem in Chap-
ters 4 (exercises O, P, and Q), 5 (exercises T and U), and 6
(exercise F):
1. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
2. Using the communication diagrams and the CRUDE
matrix, create a package diagram of the problem
domain layer.
3. Perform a verifi cation and validation walkthrough
of the package diagram.
278 Chapter 7 Moving on to Design
4. Based on the analysis models that have been created
and your current understanding of the fi rm’s posi-
tion, what design strategy would you recommend?
Why?
D. For the Picnics R Us problem in Chapters 4 (exercises
R, S, and T), 5 (exercises V and W), and 6 (exercise G):
1. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
2. Using the communication diagrams and the CRUDE
matrix, create a package diagram of the problem
domain layer.
3. Perform a verifi cation and validation walkthrough
of the package diagram.
4. Based on the analysis models that have been created
and your current understanding of the fi rm’s posi-
tion, what design strategy would you recommend?
Why?
E. For the Of-the-Month-Club problem in Chapters 4
(exercises U, V, and W), 5 (exercises X and Y), and 6
(exercise H):
1. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
2. Using the communication diagrams and the CRUDE
matrix, create a package diagram of the problem
domain layer.
3. Perform a verifi cation and validation walkthrough
of the package diagram.
4. Based on the analysis models that have been created
and your current understanding of the fi rm’s posi-
tion, what design strategy would you recommend?
Why?
F. Suppose you are leading a project that will implement
a new course-enrollment system for your univer-
sity. You are thinking about either using a packaged
course-enrollment application or outsourcing the job
to an external consultant. Create an outline for an RFP
to which interested vendors and consultants could
respond.
G. Suppose you and your friends are starting a small busi-
ness painting houses in the summertime. You need to
buy a soft ware package that handles the fi nancial trans-
actions of the business. Create an alternative matrix
that compares three packaged systems (e.g., Quicken,
MS Money, Quickbooks). Which alternative appears to
be the best choice?
MINICASES
1. Susan, president of MOTO, Inc., a human resources
management fi rm, is refl ecting on the client man-
agement soft ware system her organization purchased
four years ago. At that time, the fi rm had just gone
through a major growth spurt, and the mixture of
automated and manual procedures that had been
used to manage client accounts became unwieldy.
Susan and Nancy, her IS department head, researched
and selected the package that is currently used. Susan
had heard about the soft ware at a professional con-
ference she attended, and, at least initially, it worked
fairly well for the fi rm. Some of their procedures had
to change to fi t the package, but they expected that
and were prepared for it.
Since that time, MOTO, Inc., has continued to grow,
not only through an expansion of the client base
but also through the acquisition of several smaller
employment-related businesses. MOTO, Inc., is a
much diff erent business than it was four years ago.
Along with expanding to off er more diversifi ed human
resources management services, the fi rm’s support
staff has also expanded. Susan and Nancy are particu-
larly proud of the IS department they have built up
over the years. Using strong ties with a local university,
an attractive compensation package, and a good working
environment, the IS department is well staff ed with
competent, innovative people, plus a steady stream
of college interns that keeps the department fresh
and lively. One of the IS teams pioneered the use of
the Internet to off er MOTO’s services to a whole new
market segment, an experiment that has proved very
successful.
It seems clear that a major change is needed in the
client-management soft ware, and Susan has already
begun to plan fi nancially to undertake such a project.
Th is soft ware is a central part of MOTO’s operations,
and Susan wants to be sure that a high-quality system
is obtained this time. She knows that the vendor of
their current system has made some revisions and
additions to its product line. A number of other
Minicases 279
soft ware vendors also off er products that may be
suitable. Some of these vendors did not exist when the
purchase was made four years ago. Susan is also con-
sidering Nancy’s suggestion that the IS department
develop a custom soft ware application.
a. Outline the issues that Susan should consider that
would support the development of a custom soft –
ware application in-house.
b. Outline the issues that Susan should consider
that would support the purchase of a software
package.
c. Within the context of a systems-development pro-
ject, when should the decision of make-versus-buy
be made? How should Susan proceed? Explain
your answer.
2. Refer to minicase 1 (West Star Marinas) in Chapter 5.
Aft er all the analysis models (both the as-is and to-be
models) for West Star Marinas were completed, the
director of operations fi nally understood why it was
important to understand the as-is system before delv-
ing into the development of the to-be system. How-
ever, you now tell him that the to-be models are only
the problem-domain portion of the design. He is now
very confused. Aft er explaining to him the advantages
of using a layered approach to developing the system,
he says, “I don’t care about reusability or mainte-
nance. I only want the system to be implemented as
soon as possible. You IS types are always trying to pull
a fast one on the users. Just get the system completed.”
What is your response to the Director of Operations?
Do you jump into implementation as he seems to
want? What do you do next?
3. Refer to the analysis models that you created for pro-
fessional and scientifi c staff management (PSSM) for
minicase 2 in Chapter 4 and for minicase 1 in Chapter 6.
a. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
b. Using the communication diagrams and the
CRUDE matrix, create a package diagram of the
problem domain layer.
c. Perform a verifi cation and validation walkthrough
of the package diagram.
d. Based on the analysis models that have been cre-
ated and your current understanding of the fi rm’s
position, what design strategy would you recom-
mend? Why?
4. Refer to the analysis models that you created for
Holiday Travel Vehicles for minicase 2 in Chapter 5
and for minicase 2 in Chapter 6.
a. Perform a verifi cation and validation walkthrough
of the functional, structural, and behavioral models
to ensure that all between-model issues have been
resolved.
b. Using the communication diagrams and the
CRUDE matrix, create a package diagram of the
problem domain layer.
c. Perform a verifi cation and validation walkthrough
of the package diagram.
d. Based on the analysis models that have been
created and your current understanding of the
fi rm’s position, what design strategy would you
recommend? Why?
The most important step of the design phase is designing the individual classes and meth-
ods. Object-oriented systems can be quite complex, so analysts need to create instructions
and guidelines for programmers that clearly describe what the system must do. Th is chapter
presents a set of criteria, activities, and techniques used to design classes and methods. Together
they are used to ensure that the object-oriented design communicates how the system needs
to be coded.
OBJECTIVES
■ Become familiar with coupling, cohesion, and connascence.
■ Be able to specify, restructure, and optimize object designs.
■ Be able to identify the reuse of predefi ned classes, libraries, frameworks, and components.
■ Be able to specify constraints and contracts.
■ Be able to create a method specifi cation.
INTRODUCTION
WARNING: Th is material may be hazardous to your mental stability. Not really, but now that
we have your attention, you must realize that this material is fairly technical in nature and
that it is extremely important in today’s “fl at” world. Today, much of the actual implemen-
tation will be done in a diff erent geographic location than where the analysis and design are
performed. We must ensure that the design is specifi ed in a “correct” manner and that there
is no, or at least minimal, ambiguity in the design specifi cation.
In today’s fl at world, the common language spoken among developers is very likely to be
UML and some object-oriented language, such as Java, and not English. English has always
been and always will be ambiguous. Furthermore, to what variety of English do we refer? As
both Oscar Wilde and George Bernard Shaw independently pointed out, the United States
and England are divided by a common language.
Practically speaking, Class and Method design is where all the work actually gets done
during design. No matter which layer you are focusing on, the classes, which will be used
to create the system objects, must be designed. Some people believe that with reusable class
libraries and off -the-shelf components, this type of low-level, or detailed, design is a waste
of time and that we should jump immediately into the “real” work: coding the system.
However, past experience shows that low-level, or detailed, design is critical despite the
use of libraries and components. Detailed design is still very important for three reasons.
First, with today’s modern CASE tools, quite a bit of the actual code can be generated by the
tool from the detailed design. Second, even preexisting classes and components need to be
understood, organized, and pieced together. Th ird, it is still common for the project team
C H A P T E R 8
Class and Method Design
280
Introduction 281
to have to write some code and produce original classes that support the application logic
of the system.
Jumping right into coding will guarantee disastrous results. For example, even though
the use of layers can simplify the individual classes, they can increase the complexity of the
interactions between them. If the classes are not designed carefully, the resulting system can
be very ineffi cient. Or worse, the instances of the classes (i.e., the objects) will not be capable
of communicating with each other, which will result in the system’s not working properly.
In an object-oriented system, changes can take place at diff erent levels of abstraction.
Th ese levels include variable, method, class/object, package,1 library, and/or application/
system levels (see Figure 8-1). Th e changes that take place at one level can aff ect other levels
(e.g., changes to a class can aff ect the package level, which can aff ect both the system level
and the library level, which in turn can cause changes back down at the class level). Finally,
changes can occur at diff erent levels at the same time.
Th e good news is that the detailed design of the individual classes and methods is
fairly straightforward. Th e interactions among the objects on the problem-domain layer
have been designed, in some detail, during analysis (see Chapters 4 through 6). Th e other
layers (data management, human–computer interaction, and physical architecture) are
highly dependent on the problem-domain layer. Th erefore, if the problem-domain classes
are designed correctly, the design of the classes on the other layers will fall into place,
relatively speaking.
Th at being said, it has been our experience that many project teams are much too quick
at jumping into writing code for the classes without fi rst designing them. Some of this has
been caused by the fact that object-oriented systems analysis and design has evolved from
object-oriented programming. Until recently there has been a general lack of accepted guide-
lines on how to design and develop eff ective object-oriented systems. However, with the
Class/Object
Package
System
Library
Variable Method
FIGURE 8-1
Levels of Abstraction
in Object-Oriented
Systems
Source: Based on material from David P. Tegarden, Steven D. Sheetz, and David E. Monarchi, “A Software
Complexity Model of Object-Oriented Systems,” Decision Support Systems 13 (March 1995): 241–262.
1 A package is a group of collaborating objects. Other names for a package include cluster, partition, pattern, subject,
and subsystem.
282 Chapter 8 Class and Method Design
acceptance of UML as a standard object notation, standardized approaches based on work of
many object methodologists have begun to emerge.2
REVIEW OF THE BASIC CHARACTERISTICS
OF OBJECT ORIENTATION
Object-oriented systems can be traced back to the Simula and the Smalltalk programming lan-
guages. However, until the increase in processor power and the decrease in processor cost that
occurred in the 1980s, object-oriented approaches were not practical. Many of the specifi c details
concerning the basic characteristics of object-orientation are language dependent; that is, each
object-oriented programming language tends to implement some of the object-oriented basics
in a diff erent way. Consequently, we need to know which programming language is going to be
used to implement the diff erent aspects of the solution. Otherwise, the system could behave in a
manner diff erent than the analyst, designer, and client expect. Today, the C11, Java, Objective-C,
and Visual Basic programming languages tend to be the more predominant languages used. In
this section, we review the basic characteristics of object orientation and point out where the
language-specifi c issues emerge.
Classes, Objects, Methods, and Messages
Th e basic building block of the system is the object. Objects are instances of classes. Classes are
templates that we use to defi ne both the data and processes that each object contains. Each
object has attributes that describe data about the object. Objects have state, which is defi ned
by the value of its attributes and its relationships with other objects at a particular point in
time. And each object has methods, which specify what processes the object can perform.
From our perspective, methods are used to implement the operations that specifi ed the behav-
ior of the objects (see Chapter 5). To get an object to perform a method (e.g., to delete itself),
a message is sent to the object. A message is essentially a function or procedure call from one
object to another object.
Encapsulation and Information Hiding
Encapsulation is the mechanism that combines the processes and data into a single object.
Information hiding suggests that only the information required to use an object be available
outside the object; that is, information hiding is related to the visibility of the methods and
attributes (see Chapter 5). Exactly how the object stores data or performs methods is not
relevant, as long as the object functions correctly. All that is required to use an object are the
set of methods and the messages needed to be sent to trigger them. Th e only communication
between objects should be through an object’s methods. Th e fact that we can use an object by
sending a message that calls methods is the key to reusability because it shields the internal
workings of the object from changes in the outside system, and it keeps the system from being
aff ected when changes are made to an object.
Polymorphism and Dynamic Binding
Polymorphism means having the ability to take several forms. By supporting polymor-
phism, object-oriented systems can send the same message to a set of objects, which can be
2 For example, OPEN [I. Graham, B. Henderson-Seller, and H. Yanoussi, Th e Open Process Specifi cation (Reading,
MA: Addison-Wesley, 1997)], RUP [P. Kruchten, Th e Rational Unifi ed Process: An Introduction, 2nd ed. (Reading,
MA: Addison-Wesley, 2000)], and the Enhanced Unifi ed Process (see Chapter 1).
Review of the Basic Characteristics of Object Orientation 283
interpreted diff erently by diff erent classes of objects. Based on encapsulation and information
hiding, an object does not have to be concerned with how something is done when using other
objects. It simply sends a message to an object and that object determines how to interpret the
message. Th is is accomplished through the use of dynamic binding.
Dynamic binding refers to the ability of object-oriented systems to defer the data typing
of objects to run time. For example, imagine that you have an array of type employee that
contains instances of hourly employees and salaried employees (see Figure 8-2). Both these
types of employees implement a compute pay method. An object can send the message to
each instance contained in the array to compute the pay for that individual instance. Depend-
ing on whether the instance is an hourly employee or a salaried employee, a diff erent method
will be executed. Th e specifi c method is chosen at run time. With this ability, individual
classes are easier to understand. However, the specifi c level of support for polymorphism and
dynamic binding is language specifi c. Most object-oriented programming languages support
dynamic binding of methods, and some support dynamic binding of attributes.
But polymorphism can be a double-edged sword. Th rough the use of dynamic binding,
there is no way to know before run time which specifi c object will be asked to execute its
method. In eff ect, there is a decision made by the system that is not coded anywhere.3 Because
all these decisions are made at run time, it is possible to send a message to an object that it
does not understand (i.e., the object does not have a corresponding method). Th is can cause
a run-time error that, if the system is not programmed to handle it correctly, can cause the
system to abort.4
Finally, if the methods are not semantically consistent, the developer cannot assume
that all methods with the same name will perform the same generic operation. For example,
imagine that you have an array of type person that contains instances of employees and cus-
tomers (see Figure 8-3). Th ese both implement a compute pay method. An object can send
the message to each instance contained in the array to execute the compute pay method for
that individual instance. In the case of an instance of employee, the compute pay method
computes the amount that the employee is owed by the fi rm, whereas the compute pay
method associated with an instance of a customer computes the amount owed the fi rm by
the customer. Depending on whether the instance is an employee or a customer, a diff erent
meaning is associated with the method. Th erefore, the semantics of each method must be
determined individually. Th is substantially increases the diffi culty of understanding individ-
ual objects. Th e key to controlling the diffi culty of understanding object- oriented systems
Array Employeecontains
+computePay()
HourlyEmployee
+computePay()
SalariedEmployee
+computePay()
* *
FIGURE 8-2
Example of
Polymorphism
3 From a practical perspective, there is an implied case statement. Th e system chooses the method based on the type of
object being asked to execute it and the parameters passed as arguments to the method. Th is is typically done through
message dispatch tables that are hidden from the programmer.
4 In most object-oriented programming languages, these errors are referred to as exceptions that the system “throws”
and must “catch.” In other words, the programmer must correctly program the throw and catch or the systems will
abort. Again, each programming language can handle these situations in a unique manner.
284 Chapter 8 Class and Method Design
when using polymorphism is to ensure that all methods with the same name implement that
same generic operation (i.e., they are semantically consistent).
Inheritance
Inheritance allows developers to defi ne classes incrementally by reusing classes defi ned pre-
viously as the basis for new classes. Although we could defi ne each class separately, it might
be simpler to defi ne one general superclass that contains the data and methods needed by
the subclasses and then have these classes inherit the properties of the superclass. Subclasses
inherit the attributes and methods from the superclasses above them. Inheritance makes it
simpler to defi ne classes.
Th ere have been many diff erent types of inheritance mechanisms associated with
object-oriented systems.5 Th e most common inheritance mechanisms include diff erent forms
of single and multiple inheritance. Single inheritance allows a subclass to have only a single
parent class. Currently, all object-oriented methodologies, databases, and programming lan-
guages permit extending the defi nition of the superclass through single inheritance.
Some object-oriented methodologies, databases, and programming languages allow a sub-
class to redefi ne some or all the attributes and/or methods of its superclass. With redefi nition
capabilities, it is possible to introduce an inheritance confl ict [i.e., an attribute (or method) of
a subclass with the same name as an attribute (or method) of a super-class]. For example in
Figure 8-4, Doctor is a subclass of Employee. Both have methods named ComputePay(). Th is
causes an inheritance confl ict. Furthermore, when the defi nition of a superclass is modifi ed, all
its subclasses are aff ected. Th is can introduce additional inheritance confl icts in one (or more)
of the superclass’s subclasses. For example in Figure 8-4, Employee could be modifi ed to include
an additional method, UpdateSchedule(). Th is would add another inheritance confl ict between
Employee and Doctor. Th erefore, developers must be aware of the eff ects of the modifi cation
not only in the superclass but also in each subclass that inherits the modifi cation.
Finally, through redefi nition capabilities, it is possible for a programmer to arbitrarily
cancel the inheritance of methods by placing stubs6 in the subclass that will override the
PersonArray
Customer
HourlyEmployee SalariedEmployee
Employee
Contains
+computePay()
+computePay()+computePay()
+computePay() +computePay()
* *
FIGURE 8-3
Example of
Polymorphism Misuse
5 See, for example, M. Lenzerini, D. Nardi, and M. Simi, Inheritance Hierarchies in Knowledge Representation and
Programming Languages (New York: Wiley, 1991).
6 In this case, a stub is simply the minimal defi nition of a method to prevent syntax errors from occurring.
Review of the Basic Characteristics of Object Orientation 285
defi nition of the inherited method. If the cancellation of methods is
necessary for the correct defi nition of the subclass, then it is likely that
the subclass has been misclassifi ed (i.e., it is inheriting from the wrong
superclass).
As you can see, from a design perspective, inheritance confl icts and
redefi nition can cause all kinds of problems with interpreting the fi nal
design and implementation.7 However, most inheritance confl icts are
due to poor classifi cation of the subclass in the inheritance hierarchy (the
generalization a-kind-of semantics are violated), or the actual inheritance
mechanism violates the encapsulation and information hiding princi-
ple (i.e., subclasses are capable of directly addressing the attributes or
methods of a superclass). To address these issues, Jim Rumbaugh and his
colleagues suggested the following guidelines:8
■ Do not redefi ne query operations.
■ Methods that redefi ne inherited ones should restrict only the
semantics of the inherited ones.
■ Th e underlying semantics of the inherited method should never
be changed.
■ Th e signature (argument list) of the inherited method should
never be changed.
However, many existing object-oriented programming languages violate
these guidelines. When it comes to implementing the design, diff erent object-oriented pro-
gramming languages address inheritance confl icts diff erently. Th erefore, it is important at
this point in the development of the system to know what the chosen programming language
supports. We must be sure that the design can be implemented as intended. Otherwise, the
design needs to be modifi ed before it is turned over to remotely located programmers.
When considering the interaction of inheritance with polymorphism and dynamic bind-
ing, object-oriented systems provide the developer with a very powerful, but dangerous, set
of tools. Depending on the object-oriented programming language used, this interaction can
allow the same object to be associated with diff erent classes at diff erent times. For example, an
instance of Doctor can be treated as an instance of Employee or any of its direct and indirect
superclasses, such as SalariedEmployee and Person, respectively (see Figure 8-4). Th erefore,
depending on whether static or dynamic binding is supported, the same object may exe-
cute diff erent implementations of the same method at diff erent times. Or, if the method is
defi ned only with the SalariedEmployee class and it is currently treated as an instance of the
Employee class, the instance could cause a run-time error to occur.9 It is important to know
what object-oriented programming language is going to be used so that these kinds of issues
can be solved with the design, instead of the implementation, of the class.
With multiple inheritance, a subclass may inherit from more than one superclass. In this
situation, the types of inheritance confl icts are multiplied. In addition to the possibility of
having an inheritance confl ict between the subclass and one (or more) of its superclasses, it
is now possible to have confl icts between two (or more) superclasses. In this latter case, three
diff erent types of additional inheritance confl icts can occur:
Person
Employee
SalariedEmployee
Doctor
+computePay()
+computePay()
+computePay()
+computePay()
+updateSchedule()
FIGURE 8-4
Example of
Redefi nition and
Inheritance Confl ict
7 For more information, see Ronald J. Brachman, “I Lied about the Trees Or, Defaults and Defi nitions in Knowledge
Representation,” AI Magazine 5, no. 3 (Fall 1985): 80–93.
8 J. Rumbaugh, M. Blaha, W. Premerlani, F. Eddy, and W. Lorensen, Object-Oriented Modeling and Design (Engle-
wood Cliff s, NJ: Prentice Hall, 1991).
9 Th is happens with novices quite regularly when using C11.
286 Chapter 8 Class and Method Design
■ Two inherited attributes (or methods) have the same name (spelling) and semantics.
■ Two inherited attributes (or methods) have diff erent names but identical semantics
(i.e., they are synonyms).
■ Two inherited attributes (or methods) have the same name but diff erent semantics
(i.e., they are heteronyms, homographs, or homonyms). Th is also violates the proper
use of polymorphism.
For example, in Figure 8-5, Robot-Employee is a subclass of both Employee and Robot.
In this case, Employee and Robot confl ict with the attribute name. Which one should
Robot-Employee inherit? Because they are the same, semantically speaking, does it really
matter? It is also possible that Employee and Robot could have a semantic confl ict on the
classifi cation and type attributes if they have the same semantics. Practically speaking,
the only way to prevent this situation is for the developer to catch it during the design of the
subclass. Finally, what if the runningTime attributes have diff erent semantics? In the case
of Employee objects, the runningTime attribute stores the employee’s time running a mile,
whereas the runningTime attribute for Robot objects stores the average time between check-
ups. Should Robot-Employee inherit both of them? It really depends on whether the robot
employees can run the mile or not. With the potential for these additional types of confl icts,
there is a risk of decreasing the understandability in an object-oriented system instead of
increasing it through the use of multiple inheritance. Our advice is to use great care when
using multiple inheritance.
DESIGN CRITERIA
When considering the design of an object-oriented system, a set of criteria exists that can be
used to determine whether the design is a good one or a bad one. According to Coad and
Yourdon,10 “A good design is one that balances trade-off s to minimize the total cost of the
system over its entire lifetime.” Th ese criteria include coupling, cohesion, and connascence.
Coupling
Coupling refers to how interdependent or interrelated the modules (classes, objects, and meth-
ods) are in a system. Th e higher the interdependency, the more likely changes in part of a design
Employee Robot
Robot-Employee
-name
-classification
-runningTime
-name
-type
-runningTime
FIGURE 8-5
Additional
Inheritance
Confl icts with
multiple Inheritance
10 Peter Coad and Edward Yourdon, Object-Oriented Design (Englewood Cliff s, NJ: Yourdon Press, 1991), p. 128.
Design Criteria 287
Purchase Order
-PO Number[1..1] : Unsigned long
-Sub Total[0..1] : Currency
-Tax[0..1] : Currency
-Shipping[0..1] : Currency
-Total[0..1] : Currency
-Customer[1..1] : Customer
-State[1..1] : State
PO1 : Purchase Order
Date : Date
PO Number : Unsigned long
Sub Total : Currency
Tax : Currency
Shipping : Currency
Total : Currency
Customer : Customer
State : State
Message1
Invoice
Object1 AcctsPayForms
-Date : Date
(a) (b)
(c)
sd Make Old Patient Appt Use Case
RequestAppt(name, address)
NewCancelChangeAppt?()
ApptTimes?()
aPatient
LookUpPatient()
aReceptionist
[aPatientExists] LookupBills()
MatchAppts()
CreateAppt()
aPatient:Patient :UnpaidBill :Appointment
FIGURE 8-6 Examples of Interaction Coupling
can cause changes to be required in other parts of the design. For object- oriented systems, Coad
and Yourdon11 identifi ed two types of coupling to consider: interaction and inheritance.
Interaction coupling deals with the coupling among methods and objects through message
passing. Lieberherr and Holland put forth the law of Demeter as a guideline to minimize this type
11 Ibid.
288 Chapter 8 Class and Method Design
of coupling.12 Essentially, the law minimizes the number of objects that can receive messages from
a given object. Th e law states that an object should send messages only to one of the following:
■ Itself (For example in Figure 8-6a, Object1 can send Message1 to itself. In other words,
a method associated with Object1 can use other methods associated with Object1.13)
■ An object that is contained in an attribute of the object or one of its superclasses
(For example in Figure 8-6b, the P01 instance of the Purchase Order class should
be able to send messages using its Customer, State, and Date attributes.)
■ An object that is passed as a parameter to the method (For example in Figure 8-6c,
the aPatient instance sends the message RequestAppt(name, address) to the aRecep-
tionist instance, which is allowed to send messages to the instances contained in
the name and address parameters.)
■ An object that is created by the method (For example in Figure 8-6c, the method
RequestAppt associated with the aReceptionist instance creates an instance of the
Appointment class. Th e RequestAppt method is allowed to send messages to that
instance.)
■ An object that is stored in a global variable14
Even though the law of Demeter attempts to minimize interaction coupling among methods
and objects, each of the above allowed forms of message sending in fact increases coupling. For
example, the coupling increases between the objects if the calling method passes attributes to the
called method or if the calling method depends on the value being returned by the called method.
Th ere are six types of interaction coupling, each falling on diff erent parts of a good-to-bad
continuum. Th ey range from no direct coupling to content coupling. Figure 8-7 presents the
12 Karl J. Lieberherr and Ian M. Holland, “Assuring Good Style for Object-Oriented Programs,” IEEE Soft ware 6,
no. 5 (September, 1989): 38–48; Karl J. Lieberherr, Adaptive Object-Oriented Soft ware: Th e Demeter Method with
Propagation Patterns (Boston, MA: PWS Publishing, 1996).
13 Obviously, this is stating what is expected.
14 From a design perspective, global variables should be avoided. Most pure object-oriented programming languages
do not explicitly support global variables, and we do not address them any further.
FIGURE 8-7
Types of Interaction
Coupling
Level Type Description
Good No Direct Coupling The methods do not relate to one another; that is, they do
not call one another.
Data The calling method passes a variable to the called method. If
the variable is composite (i.e., an object), the entire object is
used by the called method to perform its function.
Stamp The calling method passes a composite variable (i.e., an
object) to the called method, but the called method only
uses a portion of the object to perform its function.
Control The calling method passes a control variable whose value
will control the execution of the called method.
Common or Global The methods refer to a “global data area” that is outside the
individual objects.
Bad Content or Pathological A method of one object refers to the inside (hidden parts) of
another object. This violates the principles of encapsulation
and information hiding. However, C++ allows this to take
place through the use of “friends.”
Source: These types are based on material from Meilir Page-Jones, The Practical Guide to Structured
Systems Design, 2nd Ed. (Englewood Cliffs, NJ: Yardon Press, 1988); Glenford Myers, Composite/
Structured Design (New York: Van Nostrand Reinhold, 1978).
Design Criteria 289
diff erent types of interaction coupling. In general, interaction coupling should be minimized.
Th e one possible exception is that non–problem-domain classes must be coupled to their cor-
responding problem-domain classes. For example, a report object (on the human–computer
interaction layer) that displays the contents of an employee object (on the problem-domain
layer) will be dependent on the employee object. In this case, for optimization purposes, the
report class may be even content or pathologically coupled to the employee class. However,
problem-domain classes should never be coupled to non–problem-do-
main classes.
Inheritance coupling, as its name implies, deals with how tightly
coupled the classes are in an inheritance hierarchy. Most authors tend
to say simply that this type of coupling is desirable. However, depending
on the issues raised previously with inheritance—inheritance confl icts,
redefi nition capabilities, and dynamic binding—a high level of inher-
itance coupling might not be a good thing. For example, in Figure 8-8,
should Method2() defi ned in Subclass be allowed to call Method1()
defi ned in Superclass? Or, should Method2() defi ned in Subclass refer
to Attribute1 defi ned in Superclass? Or, even more confusing, assuming
that Superclass is an abstract class, can a Method1() call Method2() or
use Attribute2 defi ned in Subclass? Obviously, the fi rst two examples
have some intuitive sense. Using the properties of a superclass is the
primary purpose of inheriting from it in the fi rst place. On the other
hand, the third example is somewhat counterintuitive. However, owing to the way that dif-
ferent object-oriented programming languages support dynamic binding, polymorphism,
and inheritance, all these examples could be possible.
As Snyder has pointed out, most problems with inheritance involve the ability
within the object-oriented programming languages to violate the encapsulation and
information-hiding principles.15 From a design perspective, the developer needs to opti-
mize the trade-off s of violating the encapsulation and information-hiding principles and
increasing the desirable coupling between subclasses and its superclasses. Th e best way to
solve this conundrum is to ensure that inheritance is used only to support generalization/
specialization (a-kind-of) semantics and the principle of substitutability (see Chapter 5). All
other uses should be avoided.
Cohesion
Cohesion refers to how single-minded a module (class, object, or method) is within a sys-
tem. A class or object should represent only one thing, and a method should solve only a
single task. Th ree general types of cohesion have been identifi ed by Coad and Yourdon for
object-oriented systems: method, class, and generalization/specialization.16
Method cohesion addresses the cohesion within an individual method (i.e., how
single-minded a method is). Methods should do one and only one thing. A method that
actually performs multiple functions is more diffi cult to understand—and, therefore, to
implement and maintain—than one that performs only a single function. Seven types
of method cohesion have been identifi ed (see Figure 8-9). Th ey range from functional
Subclass
Superclass
+Method2()
-Attribute2
+Method1()
-Attribute1
FIGURE 8-8
Example of Inher-
itance Coupling
15 Alan Snyder, “Encapsulation and Inheritance in Object-Oriented Programming Languages,” in N. Meyrowitz
(ed.), OOPSLA ’86 Conference Proceedings, ACM SigPlan Notices 21, no. 11 (November 1986); Alan Snyder, “Inher-
itance and the Development of Encapsulated Soft ware Components,” in B. Shriver and P. Wegner (eds.), Research
Directions in Object-Oriented Programming (Cambridge, MA: MIT Press, 1987).
16 Coad and Yourdon, Object-Oriented Design.
290 Chapter 8 Class and Method Design
FIGURE 8-9
Types of Method
Cohesion
Good Functional A method performs a single problem-related task (e.g.,
calculate current GPA).
Sequential The method combines two functions in which the output
from the fi rst one is used as the input to the second one
(e.g., format and validate current GPA).
Communicational The method combines two functions that use the same
attributes to execute (e.g., calculate current and cumula-
tive GPA).
Procedural The method supports multiple weakly related functions.
For example, the method could calculate student GPA,
print student record, calculate cumulative GPA, and print
cumulative GPA.
Temporal or Classical The method supports multiple related functions in time
(e.g., initialize all attributes).
Logical The method supports multiple related functions, but the
choice of the specifi c function is chosen based on a con-
trol variable that is passed into the method. For example,
the called method could open a checking account, open
a savings account, or calculate a loan, depending on the
message that is sent by its calling method.
Bad Coincidental The purpose of the method cannot be defi ned or it
performs multiple functions that are unrelated to one
another. For example, the method could update customer
records, calculate loan payments, print exception reports,
and analyze competitor pricing structure.
Source: These types are based on material from Page-Jones, The Practical Guide to Structured Sys-
tems; Myers, Composite/Structured Design; Edward Yourdon and Larry L. Constantine, Structured
Design: Fundamentals of a Discipline of Computer Program and Systems Design (Englewood Cliffs, NJ:
Prentice-Hall, 1979).
Level Type Description
cohesion (good) down to coincidental cohesion (bad). In general, method cohesion
should be maximized.
Class cohesion is the level of cohesion among the attributes and methods of a class
(i.e., how single-minded a class is). A class should represent only one thing, such as an
employee, a department, or an order. All attributes and methods contained in a class should
be required for the class to represent the thing. For example, an employee class should
have attributes that deal with a social security number, last name, fi rst name, middle initial,
addresses, and benefi ts, but it should not have attributes such as door, engine, or hood.
Furthermore, there should be no attributes or methods that are never used. In other words, a
class should have only the attributes and methods necessary to fully defi ne instances for the
problem at hand. In this case, we have ideal class cohesion. Glenford Meyers suggested that a
cohesive class17 should have these attributes:
■ It should contain multiple methods that are visible outside the class (i.e., a
single-method class rarely makes sense).
■ Each visible method performs only a single function (i.e., it has functional cohesion;
see Figure 8-9).
17 We have adapted his informational-strength module criteria from structured design to object-oriented design. [See
Glenford J. Myers, Composite/Structured Design (New York, NY: Van Nostrand Reinhold, 1978).]
Design Criteria 291
FIGURE 8-10
Types of Class
Cohesion
Good Ideal The class has none of the mixed cohesions.
Mixed-Role The class has one or more attributes that relate objects
of the class to other objects on the same layer (e.g., the
problem domain layer), but the attribute(s) has nothing to
do with the underlying semantics of the class.
Mixed-Domain The class has one or more attributes that relate objects of
the class to other objects on a different layer. As such, they
have nothing to do with the underlying semantics of the
thing that the class represents. In these cases, the offending
attribute(s) belongs in another class located on one of the
other layers. For example, a port attribute located in a prob-
lem domain class should be in a system architecture class
that is related to the problem domain class.
Worse Mixed-Instance The class represents two different types of objects. The class
should be decomposed into two separate classes. Typically,
different instances only use a portion of the full defi nition
of the class.
Based upon material from Page-Jones, Fundamentals of Object-Oriented Design in UML.
Level Type Description
■ All methods reference only attributes or other methods defi ned within the
class or one of its superclasses (i.e., if a method is going to send a message to
another object, the remote object must be the value of one of the local object’s
attributes).18
■ It should not have any control couplings between its visible methods (see Figure 8-7).
Page-Jones19 has identifi ed three less-than-desirable types of class cohesion: mixed-instance,
mixed-domain, and mixed-role (see Figure 8-10). An individual class can have a mixture of
any of the three types.
Generalization/specialization cohesion addresses the sensibility of the inheritance hierar-
chy. How are the classes in the inheritance hierarchy related? Are the classes related through a
generalization/specialization (a-kind-of) semantics? Or, are they related via some association,
aggregation, or membership type of relationship that was created for simple reuse purposes?
Recall all the issues raised previously on the use of inheritance. For example, in Figure 8-11,
the subclasses ClassRooms and Staff inherit from the superclass Department. Obviously,
instances of the ClassRooms and Staff classes are not a-kind-of Department. However, in the
early days of object-oriented programming, this use of inheritance was quite common. When
a programmer saw that there were some common properties that a set of classes shared, the
programmer would create an artifi cial abstraction that defi ned the commonalities. Th is was
potentially useful in a reuse sense, but it turned out to cause many maintenance nightmares.
In this case, instances of the ClassRooms and Staff classes are associated with or a-part-of an
instance of Department. Today we know that highly cohesive inheritance hierarchies should
support only the semantics of generalization and specialization (a-kind-of) and the principle
of substitutability.
18 Th is restricts messages passing to only the fi rst, second, and fourth conditions supported by the law of Demeter.
For example, in Figure 8-6c, aReceptionist must have attributes associated with it that contains objects for Patients,
Unpaid Bills, and Appointments. Furthermore, once an instance of Appointment is created, aReceptionist must have
an attribute with the instance as its value to send any additional messages.
19 See Meilir Page-Jones, Fundamentals of Object-Oriented Design in UML (Reading, MA: Addison-Wesley, 2000).
292 Chapter 8 Class and Method Design
Connascence
Connascence20 generalizes the ideas of cohesion and coupling, and it combines them with
the arguments for encapsulation. To accomplish this, three levels of encapsulation have been
identifi ed. Level-0 encapsulation refers to the amount of encapsulation realized in an individ-
ual line of code, level-1 encapsulation is the level of encapsulation attained by combining lines
of code into a method, and level-2 encapsulation is achieved by creating classes that contain
both methods and attributes. Method cohesion and interaction coupling address primarily
level-1 encapsulation. Class cohesion, generalization/specialization cohesion, and inheritance
coupling address only level-2 encapsulation. Connascence, as a generalization of cohesion
and coupling, addresses both level-1 and level-2 encapsulation.
But what exactly is connascence? Connascence literally means to be born together. From
an object-oriented design perspective, it really means that two modules (classes or methods)
are so intertwined that if you make a change in one, it is likely that a change in the other will
be required. On the surface, this is very similar to coupling and, as such, should be minimized.
However, when you combine it with the encapsulation levels, it is not quite that simple. In
this case, we want to minimize overall connascence by eliminating any unnecessary connas-
cence throughout the system; minimize connascence across any encapsulation boundaries,
such as method boundaries and class boundaries; and maximize connascence within any
encapsulation boundary.
Based on these guidelines, a subclass should never directly access any hidden attribute
or method of a superclass [i.e., a subclass should not have special rights to the properties
of its superclass(es)]. If direct access to the nonvisible attributes and methods of a super-
class by its subclass is allowed—and is permitted in most object-oriented programming
languages—and a modifi cation to the superclass is made, then owing to the connascence
between the subclass and its superclass, it is likely that a modifi cation to the subclass also
is required.21 In other words, the subclass has access to something across an encapsulation
boundary (the class boundary between the subclass and the superclass). Practically speak-
ing, you should maximize the cohesion (connascence) within an encapsulation boundary
and minimize the coupling (connascence) between the encapsulation boundaries. Th ere are
many possible types of connascence. Figure 8-12 describes fi ve of the types.
Department
ClassRooms StaffFIGURE 8-11
Generalization/
Specialization vs.
Inheritance Abuse
20 See Meilir Page-Jones, “Comparing Techniques by Means of Encapsulation and Connascence,” Communications
of the ACM 35, no. 9 (September 1992): 147–151.
21 Based on these guidelines, the use of the protected visibility, as supported in Java and C11, should be minimized,
if not avoided. “Friends” as defi ned in C11 also should be minimized or avoided. Owing to the level of dependencies
these language features create, any convenience aff orded to a programmer is more than off set in potential design,
understandability, and maintenance problems. Th ese features must be used with great caution and must be fully
documented.
Object Design Activities 293
Name If a method refers to an attribute, it is tied to the name of the attribute. If the
attribute’s name changes, the content of the method will have to change.
Type or Class If a class has an attribute of type A, it is tied to the type of the attribute.
If the type of the attribute changes, the attribute declaration will have to
change.
Convention A class has an attribute in which a range of values has a semantic meaning
(e.g., account numbers whose values range from 1000 to 1999 are assets).
If the range would change, then every method that used the attribute would
have to be modifi ed.
Algorithm Two different methods of a class are dependent on the same algorithm to
execute correctly (e.g., insert an element into an array and fi nd an element
in the same array). If the underlying algorithm would change, then the insert
and fi nd methods would also have to change.
Position The order of the code in a method or the order of the arguments to a
method is critical for the method to execute correctly. If either is wrong,
then the method will, at least, not function correctly.
Based upon material from Meilir Page-Jones, “Comparing Techniques by Means of Encapsulation
and Connascence” and Meilir Page-Jones, Fundamentals of Object-Oriented Design in UML.
FIGURE 8-12
Types of Connascence
Type Description
OBJECT DESIGN ACTIVITIES
Th e design activities for classes and methods are really an extension of the analysis and evo-
lution activities presented previously (see Chapters 4 through 7). In this case, we expand the
descriptions of the partitions, layers, and classes. Practically speaking, the expanded descrip-
tions are created through the activities that take place during the detailed design of the classes
and methods. Th e activities used to design classes and methods include additional specifi ca-
tion of the current model, identifying opportunities for reuse, restructuring the design, opti-
mizing the design, and, fi nally, mapping the problem-domain classes to an implementation
language. Of course, any changes made to a class on one layer can cause the classes on the
other layers that are coupled to it to be modifi ed as well.
Adding Specifi cations
At this point in the development of the system, it is crucial to review the current set of
functional, structural, and behavioral models. First, we should ensure that the classes on the
problem-domain layer are both necessary and suffi cient to solve the underlying problem.
To do this, we need to be sure that there are no missing attributes or methods and no extra
or unused attributes or methods in each class. Furthermore, are there any missing or extra
classes? If we have done our job well during analysis, there will be few, if any, attributes,
methods, or classes to add to the models. And it is unlikely that we have any extra attributes,
methods, or classes to delete from the models. However, we still need to ensure that we have
factored, abstracted, and refi ned the evolving models and created the relevant partitions and
collaborations (see Chapter 7).
Second, we need to fi nalize the visibility (hidden or visible) of the attributes and methods
in each class. Depending on the object-oriented programming language used, this could be
predetermined. [For example, in Smalltalk, attributes are hidden and methods are visible.
Other languages allow the programmer to set the visibility of each attribute or method. For
example, in C11 and Java, you can set the visibility to private (hidden), public (visible), or
294 Chapter 8 Class and Method Design
protected (visible to subclasses, but not to other classes).]22 By default, most object-oriented
analysis and design approaches assume Smalltalk’s approach.
Th ird, we need to decide on the signature of every method in every class. Th e signature
of a method comprises three parts: the name of the method, the parameters or arguments
that must be passed to the method, including their object type, and the type of value that the
method will return to the calling method. Th e signature of a method is related to the method’s
contract.23
Fourth, we need to defi ne any constraints that must be preserved by the objects (e.g., an
attribute of an object that can have values only in a certain range). Th ere are three diff erent
types of constraints: preconditions, postconditions, and invariants.24 Th ese are captured in
the form of contracts and assertions added to the CRC cards and class diagrams. We also
must decide how to handle a violation of a constraint. Should the system simply abort? Should
the system automatically undo the change that caused the violation? Should the system let the
end user determine the approach to correct the violation? In other words, the designer must
design the errors that the system is expected to handle. It is best not to leave these types of
design decisions for the programmer to solve. Violations of a constraint are known as excep-
tions in languages such as C11 and Java.
Even though we have described these activities in the context of the problem-domain
layer, they are also applicable to the other layers: data management (Chapter 9), human–
computer interaction (Chapter 10), and physical architecture (Chapter 11).
Identifying Opportunities for Reuse
Previously, we looked at possibly employing reuse in our models in analysis through the use
of patterns (see Chapter 5). In design, in addition to using analysis patterns, there are oppor-
tunities for using design patterns, frameworks, libraries, and components. Th e opportunities
vary depending on which layer is being reviewed. For example, it is doubtful that a class
library will be of much help on the problem-domain layer, but a class library could be of great
help on the foundation layer.
Like analysis patterns, design patterns are simply useful grouping of collaborating classes
that provide a solution to a commonly occurring problem. Th e primary diff erence between
analysis and design patterns is that design patterns are useful in solving “a general design
problem in a particular context,”25 whereas analysis patterns tended to aid in fi lling out a
problem-domain representation. For example, a useful design pattern is the Whole-Part pat-
tern (see Figure 8-13a). Th e Whole-Part pattern explicitly supports the Aggregation and Com-
position relationships within the UML. Another useful design pattern is the Iterator pattern
(see Figure 8-13b). Th e primary purpose of the Iterator pattern is to provide the designer with
a standard approach to support traversing diff erent types of collections. By using this pattern,
regardless of the collection type (ConcreteAggregate), the designer knows that the collection
will need to create an iterator (ConcreteIterator) that customizes the standard operations used
to traverse the collection: fi rst(), next(), isDone(), and currentItem(). Given the number of col-
lections typically found in business applications, this pattern is one of the more useful ones. For
example in Figure 8-14a, we replicate a portion of both the Appointment and Library problems
discussed in previous chapters, and in Figure 8-14b we show how the Iterator pattern can be
22 It is also possible to control visibility through packages and friends (see Footnote 21).
23 Contracts were introduced in Chapter 5, and they are described in detail later in this chapter.
24 Constraints are described in more detail later in this chapter.
25 Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides, Design Patterns: Elements of Reusable Object-
Oriented Soft ware (Reading, MA: Addison-Wesley, 1995).
Object Design Activities 295
(a)
(b)
(c)
calls service
combines
Client Whole
Part1
+serviceA1()
+serviceA2()
+serviceN1()
+serviceN2()
PartN+doTask() +service1()
+service2()
sendMsg
receiveMsg
receiveMsg
Receiver
Peer1
+service()
+receive()
+unmarshal()
+receiveMsg()
Receiver Forwarder
InterProcessCommunication
InterProcessCommunication
sendMsg
+marshal()
+deliver()
+sendMsg()
+receive()
+unmarshal()
+receiveMsg()
Forwarder
+marshal()
+deliver()
+sendMsg()
Peer2
+service()1
1
1
1
<>
Aggregate
+createlterator(): +first() : Object
+next() : Object
+isDone() :
+currentltem() : Object
<>
lterator
Concretelterator ConcreteAggregate
Iterator createlterator()
{ return new Concretelterator(this); }
Client
FIGURE 8-13 Sample Design Patterns
Source: Based upon material from F. Buschmann, R. Meunier, H. Rohnert, P. Sommerlad, and M. Stal, Pattern-Oriented Software Architecture: A System of
Patterns (Chichester, UK: Wiley, 1996); E. Gamma, R. Helm, R. Johnson, and J. Vlissides, Design Patterns: Elements of Reusable Object-Oriented Software
(Reading, MA: Addison-Wesley, 1995).
296 Chapter 8 Class and Method Design
Check Out Trans Transaction Line Item
Transaction Line Item
1..1 1..*
1..*
Patient
-amount
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
0..*
0..*1..1
1..1
1..1
Appointment
-time
-date
-reason
+cancel without notice()
has
+ primary
insurance
carrier
schedules
(a)
(b)
Transaction Line Item 1..* 1..1
1..* 1..1
Appointment
-time
-date
-reason
+cancel without notice()
<>
Aggregate
+createlterator() : Iterator
<>
Iterator
+first() : Object
+next() : Object
+isDone() : boolean
+currentItem() : Object
Transaction Line Item
<>
Iterator
+first() : Object
+next() : Object
+isDone() : boolean
+currentItem() : Object
Client
Client
<>
Aggregate
+createlterator() : Iterator
Check Out Trans
FIGURE 8-14 Iterator Design Pattern Applied to Library and Appointment Problems
Object Design Activities 297
applied to those sections of their evolving designs. Finally, some of the design patterns support
diff erent physical architectures (see Chapter 11). For example, the Forwarder-Receiver pattern
(see Figure 8-13c) supports a peer-to-peer architecture. Many design patterns are available in
C11 or Java source code.
A framework is composed of a set of implemented classes that can be used as a basis for
implementing an application. Most frameworks allow us to create subclasses to inherit from
classes in the framework. Th ere are object-persistence frameworks that can be purchased and
used to add persistence to the problem-domain classes, which would be helpful on the data
management layer. Of course, when inheriting from classes in a framework, we are creating
a dependency (i.e., increasing the inheritance coupling from the subclass to the superclass).
Th erefore, if we use a framework and the vendor makes changes to the framework, we will
have to at least recompile the system when we upgrade to the new version of the framework.
A class library is similar to a framework in that it typically has a set of implemented
classes that were designed for reuse. However, frameworks tend to be more domain spe-
cifi c. In fact, frameworks may be built using a class library. A typical class library could be
purchased to support numerical or statistical processing, fi le management (data manage-
ment layer), or user interface development (human–computer interaction layer). In some
cases, instances of classes contained in the class library can be created, and in other cases,
classes in the class library can be extended by creating subclasses based on them. As with
frameworks, if we use inheritance to reuse the classes in the class library, we will run into
all the issues dealing with inheritance coupling and connascence. If we directly instantiate
classes in the class library, we will create a dependency between our object and the library
object based on the signatures of the methods in the library object. Th is increases the
interaction coupling between the class library object and our object.
A component is a self-contained, encapsulated piece of soft ware that can be plugged
into a system to provide a specifi c set of required functionalities. Today, there are many
components available for purchase. A component has a well-defi ned API (application pro-
gram interface). An API is essentially a set of method interfaces to the objects contained in
the component. Th e internal workings of the component are hidden behind the API. Com-
ponents can be implemented using class libraries and frameworks. However, components
also can be used to implement frameworks. Unless the API changes between versions of
the component, upgrading to a new version normally requires only linking the component
back into the application. As such, recompilation typically is not required.
Which of these approaches should we use? It depends on what we are trying to build.
In general, frameworks are used mostly to aid in developing objects on the physical archi-
tecture, human–computer interaction, or data management layers; components are used
primarily to simplify the development of objects on the problem-domain and human–
computer interaction layers; and class libraries are used to develop frameworks and com-
ponents and to support the foundation layer. Whichever of these reuse approaches you
use, you must remember that reuse brings many potential benefi ts and possible problems.
For example, the soft ware has previously been verifi ed and validated, which should reduce
the amount of testing required for our system. However as stated before, if the soft ware on
which we are basing our system changes, then most likely, we will also have to change our
system. Furthermore, if the soft ware is from a third-party fi rm, we are creating a depend-
ency from our fi rm (or our client’s fi rm) to the third-party vendor. Consequently, we need
to have some confi dence that the vendor will be in business for a while.
Restructuring the Design
Once the individual classes and methods have been specifi ed and the class libraries, frame-
works, and components have been incorporated into the evolving design, we should use
298 Chapter 8 Class and Method Design
factoring to restructure the design. Factoring (Chapter 7) is the process of separating out
aspects of a method or class into a new method or class to simplify the overall design. For
example, when reviewing a set of classes on a particular layer, we might discover that a subset
of them shares a similar defi nition. In that case, it may be useful to factor out the similarities
and create a new class. Based on the issues related to cohesion, coupling, and connascence,
the new class may be related to the old classes via inheritance (generalization) or through an
aggregation or association relationship.
Another process that is useful for restructuring the evolving design is normalization.
Normalization is described in Chapter 9 in relation to relational databases. However, normali-
zation can be useful at times to identify potential classes that are missing from the design. Also
related to normalization is the requirement to implement the actual association and aggregation
relationships as attributes. Virtually no object-oriented programming language diff erentiates
between attributes and association and aggregation relationships. Th erefore, all association and
aggregation relationships must be converted to attributes in the classes. For example in Figure
8-15a, the Customer and State classes are associated with the Order class. Furthermore, the
Product-Order association class is associated with both the Order and Product classes. One of
the fi rst things that must be done is to convert the Product Order Association class to a normal
class. Notice the multiplicity values for the new associations between the Order and the Product
Order classes and the Product Order and Product classes (see Figure 8-15b). Next, we need to
convert all associations to attributes that represent the relationships between the aff ected classes.
In this case, the Customer class must have an Orders attribute added to represent the set of
orders that an instance of the Customer class may possess; the Order class must add attributes to
reference instances of the Customer, State, and Product Order classes; the State class must have
an attribute added to it to reference all of the instances of the Order class that is associated with
that particular state; the new Product Order class must have attributes that allow an instance of
the Product Order class to reference which instance of the Order class and which instance of the
Product class is relevant to it; and, fi nally, the Product class must add an attribute that references
the relevant instances of the Product Order class (see Figure 8-15c). As you can see, even in this
very small example, many changes need to be made to ready the design for implementation.
Finally, all inheritance relationships should be challenged to ensure that they sup-
port only a generalization/specialization (a-kind-of) semantics. Otherwise, all the prob-
lems mentioned previously with inheritance coupling, class cohesion, and generalization/
specialization cohesion will come to pass.
Optimizing the Design26
Up until now, we have focused our energy on developing an understandable design. With
all the classes, patterns, collaborations, partitions, and layers designed and with all the class
libraries, frameworks, and components included in the design, understandability has been
our primary focus. However, increasing the understandability of a design typically creates
an ineffi cient design. Conversely, focusing on effi ciency issues will deliver a design that is
more diffi cult to understand. A good practical design manages the inevitable trade-off s that
must occur.27
26 Th e material contained in this section is based on James Rumbaugh, Michael Blaha, William Premerlani, Frederick
Eddy, and William Lorensen, Object-Oriented Modeling and Design (Englewood Cliff s, NJ: Prentice Hall, 1991);
Bernd Brugge and Allen H. Dutoit, Object-Oriented Soft ware Engineering: Conquering Complex and Changing
Systems (Englewood Cliff s, NJ: Prentice Hall, 2000).
27 Th e optimizations described here are only suggestions. In all cases, the decision to implement one or more
of these optimizations really depends on the problem domain of the system and the environment on which the
system will reside, i.e., the data management layer (see Chapter 9), the human–computer interaction layer (see
Chapter 10), and the physical architecture layer (see Chapter 11).
FIGURE 8-15 Converting Associations to Attributes
299
(a)
-Order Number[1..1] : unsigned long
-Date[1..1] : Date
-Sub Total[0..1] : double
-Tax[0..1] : double
-Shipping[0..1] : double
-Total[0..1] : double
1..1
1..1
0..*
1..*0..*
0..*
Order
Customer
State
-Cust ID[1..1]
-Last Name[1..1]
-First Name[1..1]
Product
-Product Number[1..1] :
unsigned long(idl)
-Product Desc[1..1] : String
-Price[..] : double
-State[1..1] : String
-TaxRate[1..1] : float
(b)
-Order Number[1..1] : unsigned long
-Date[1..1] : Date
-Sub Total[0..1] : double
-Tax[0..1] : double
-Shipping[0..1] : double
-Total[0..1] : double
Product Order
Customer
-Cust ID[1..1]
-Last Name[1..1]
-First Name[1..1]
Product
-Product Number[1..1] :
unsigned long(idl)
-Product Desc[1..1] : String
-Price[..] : double
1..1
0..*1..1
0..*
1..1
0..*
State
-State[1..1] : String
-TaxRate[1..1] : float
1..1
1..*
Order
-Qty[1..1] : Integer
-Extension[1..1] : Decimal
Product Order
-Qty[1..1] : unsigned long
-Extension[1..1] : double
(c)
State
-State[1..1] : String
-TaxRate[1..1] : float
-Orders[0..*] : Order
-Order Number[1..1] : unsigned long
-Date[1..1] : Date
-Sub Total[0..1] : double
-Tax[0..1] : double
-Shipping[0..1] : double
-Total[0..1] : double
-Customer{1..1] : Customer
-State{1..] : State
-Product Orders[1..*] : Product Order
Order Product Order
-Qty[1..1] : Integer
-Extension[1..1] : Decimal
-Order[1..1] : Order
-Product[1..1] : Product
-Cust ID[1..1] :
unsigned long
-Last Name[1..1] : String
-First Name[1..1] : String
-Orders[0..*] : Order
Customer
-Product Number[1..1] :
unsigned long(idl)
-Product Desc[1..1] : String
-Price[1..1] : double
-Product Orders[0..*] :
Product order
Product
300 Chapter 8 Class and Method Design
Th e fi rst optimization to consider is to review the access paths between objects. In some
cases, a message from one object to another has a long path to traverse (i.e., it goes through
many objects). If the path is long and the message is sent frequently, a redundant path should
be considered. Adding an attribute to the calling object that will store a direct connection to
the object at the end of the path can accomplish this.
A second optimization is to review each attribute of each class. It should be determined
which methods use the attributes and which objects use the methods. If the only methods
that use an attribute are read and update methods and only instances of a single class send
messages to read and update the attribute, then the attribute may belong with the calling class
instead of the called class. Moving the attribute to the calling class will substantially speed up
the system.
A third optimization is to review the direct and indirect fan-out of each method. Fan-out
refers to the number of messages sent by a method. Th e direct fan-out is the number of messages
sent by the method itself, whereas the indirect fan-out also includes the number of messages
sent by the methods called by the other methods in a message tree. If the fan-out of a method is
high relative to the other methods in the system, the method should be optimized. One way to
do this is to consider adding an index to the attributes used to send the messages to the objects
in the message tree.
A fourth optimization is to look at the execution order of the statements in oft en-used
methods. In some cases, it is possible to rearrange some of the statements to be more effi cient.
For example, if based on the objects in the system, it is known that a search routine can be
narrowed by searching on one attribute before another one, then the search algorithm should
be optimized by forcing it to always search in a predefi ned order.
A fi ft h optimization is to avoid recomputation by creating a derived attribute (or active
value) (e.g., a total that stores the value of the computation). Th is is also known as caching
computational results, and it can be accomplished by adding a trigger to the attributes con-
tained in the computation (i.e., attributes on which the derived attribute is dependent). Th is
would require a recomputation to take place only when one of the attributes that go into the
computation is changed. Another approach is to simply mark the derived attribute for rec-
omputation and delay the recomputation until the next time the derived attribute is accessed.
Th is last approach delays the recomputation as long as possible. In this manner, a computa-
tion does not occur unless it must occur. Otherwise, every time a derived attribute needs to
be accessed, a computation will be required.
A sixth optimization that should be considered deals with objects that participate in a one-
to-one association; that is, they both must exist for either to exist. In this case, it might make
sense, for effi ciency purposes, to collapse the two defi ning classes into a single class. However,
this optimization might need to be reconsidered when storing the “fatter” object in a database.
Depending on the type of object persistence used (see Chapter 9), it can actually be more effi –
cient to keep the two classes separate. Alternatively, it could make more sense for the two classes
to be combined on the problem-domain layer but kept separate on the data management layer.
Mapping Problem-Domain Classes to Implementation Languages28
Up until this point, it has been assumed that the classes and methods in the models would
be implemented directly in an object-oriented programming language. However, now it is
important to map the current design to the capabilities of the programming language used.
For example, if we have used multiple inheritance in our design but we are implementing
in a language that supports only single inheritance, then the multiple inheritance must be
factored out of the design. If the implementation is to be done in an object-based language,
28 Th e mapping rules presented in this section are based on material in Coad and Yourdon, Object-Oriented Design.
Object Design Activities 301
one that does not support inheritance,29 or a non–object-based language, such as C, we must
map the problem-domain objects to programming constructs that can be implemented
using the chosen implementation environment.
Implementing Problem Domain Classes in a Single-Inheritance Language Th e only
issue associated with implementing problem-domain objects is the factoring out of any multiple
inheritance—i.e., the use of more than one superclass—used in the evolving design. For example,
if you were to implement the solution in Java, Smalltalk, or Visual Basic.net, you must factor out
any multiple inheritance. Th e easiest way to do this is to use the following rule:
RULE 1a: Convert the additional inheritance relationships to association relationships.
Th e multiplicity of the new association from the subclass to the superclass
should be 1..1. If the additional superclasses are concrete, that is, they can be
instantiated themselves, then the multiplicity from the superclass to the sub-
class is 0..1. Otherwise, it is 1..1. Furthermore, an exclusive-or (XOR) constraint
must be added between the associations. Finally, you must add appropriate
methods to ensure that all information is still available to the original class.
or
RULE 1b: Flatten the inheritance hierarchy by copying the attributes and methods of
the additional superclass(es) down to all of the subclasses and remove the
additional superclass from the design.30
Figure 8-16 demonstrates the application of these rules. Figure 8-16a portrays a simple
example of multiple inheritance where Flying Car inherits from both Airplane and Car, and
Amphibious Car inherits from both Car and Boat. Assuming that Car is concrete, we apply
Rule 1a to part a, and we end up with the diagram in part b, where we have added the associa-
tion between Flying Car and Car and the association between Amphibious Car and Boat. Th e
multiplicities have been added correctly, and the XOR constraint has been applied. If we apply
Rule 1b to part a, we end up with the diagram in part c, where all the attributes of Car have
been copied down into Flying Car and Amphibious Car. In this latter case, you might have to
deal with the eff ects of inheritance confl icts (see earlier in the chapter).
Th e advantage of Rule 1a is that all problem-domain classes identifi ed during analysis
are preserved. Th is allows maximum fl exibility of maintenance of the design of the problem
domain layer. However, Rule 1a increases the amount of message passing required in the sys-
tem, and it has added processing requirements involving the XOR constraint, thus reducing
the overall effi ciency of the design. Accordingly, our recommendation is to limit Rule 1a to
be applied only when dealing with “extra” superclasses that are concrete because they have an
independent existence in the problem domain. Use Rule 1b when they are abstract because
they do not have an independent existence from the subclass.
Implementing Problem Domain Objects in an Object-Based Language If we are going to
implement our solution in an object-based language (i.e., a language that supports the creation
of objects but does not support implementation inheritance), we must factor out all uses of
inheritance from the problem-domain class design. Applying the preceding rule to all super-
classes enables us to restructure our design without any inheritance.
Figure 8-17 demonstrates the application of the preceding rules. Figure 8-17a shows
the same simple example of multiple inheritance portrayed in Figure 8-16, where Flying Car
29 In this case, we are talking about implementation inheritance, not the interface inheritance. Interface inheritance
supported by Visual Basic and Java supports only inheriting the requirements to implement certain methods, not any
implementation. Java and Visual Basic.net also support single inheritance as described in this text.
30 It is also a good idea to document this modifi cation in the design so that in the future, modifi cations to the design
can be maintained easily.
302 Chapter 8 Class and Method Design
Flying Car
-mfg
-yr
Airplane
-EngineType
-Fuel Type
Car
-NumberOfDoors
-RegNo
-attribute1
-attribute2
Amphibious Car
-mfg
-yr
Boat
-Weight
-Length
(a)
(b)
(c)
{XOR}
0..* 0..*
1..11..1
Flying Car
-mfg
-yr
Flying Car
-mfg
-yr
Airplane
Airplane
-EngineType
-Fuel Type
Car
-NumberOfDoors
-RegNo
-NumberOfDoors
-RegNo
-NumberOfDoors
-RegNo
Amphibious Car
-mfg
-yr
Boat
-Weight
-Length
Boat
-Weight
-Length
Amphibious Car
-mfg
-yr
-EngineType
-Fuel Type
FIGURE 8-16 Factoring Out Multiple-Inheritance Effect for a Single-Inheritance Language
Object Design Activities 303
-attribute1
-attribute2
Amphibious Car
-Weight
-Length
(a)
(b)
(c)
{XOR}
0..*
0..1
0..*
1..1 0..1
0..1
0..1
1..1
Flying Car
-mfg
-yr
Airplane
-EngineType
-Fuel Type
Car
-NumberOfDoors
-RegNo
Amphibious Car
-mfg
-yr
Boat
-Weight
-Length
Airplane
-EngineType
-Fuel Type
Car
-NumberOfDoors
-RegNo
Boat
Flying Car
-mfg
-yr
Flying Car
-EngineType
-FuelType
-mfg
-yr
-NumberOfDoors
-RegNo
Amphibious Car
-Weight
-Length
-mfg
-yr
-NumberOfDoors
-RegNo
-mfg
-yr
FIGURE 8-17 Factoring Out Multiple Inheritance Effect for an Object-Based Language
inherits from both Airplane and Car, and Amphibious Car inherits from both Car and Boat.
Assuming that Airplane, Car, and Boat are concrete, we apply Rule 1a to part a and we end up
with the diagram in part b, where we have added the associations, the multiplicities, and the
XOR constraint. If we apply Rule 1b to part a, we end up with the diagram in part c, where all
the attributes of the superclasses have been copied down into Flying Car and Amphibious Car.
In this latter case, you might have to deal with the eff ects of inheritance confl icts.
304 Chapter 8 Class and Method Design
Implementing Problem-Domain Objects in a Traditional Language From a practical
perspective, we are much better off implementing an object-oriented design in an object-
oriented programming language, such as C11, Java, Objective-C, or Visual Basic.net. Prac-
tically speaking, the gulf between an object-oriented design and a traditional programming
language is simply too great for mere mortals to be able to cross. Th e best advice that we can
give about implementing an object-oriented design in a traditional programming language is
to run away as fast and as far as possible from the project. However, if we are brave (foolish?)
enough to attempt this, we must realize that in addition to factoring out inheritance from
the design, we have to factor out all uses of polymorphism, dynamic binding, encapsulation,
and information hiding. Th is is quite a bit of additional work to be accomplished. Th e way
we factor these object-oriented features out of the detailed design of the system tends to be
language dependent. Th is is beyond the scope of this text.
CONSTRAINTS AND CONTRACTS
Contracts were introduced in Chapter 5 in association with collaborations. A contract formal-
izes the interactions between the client and server objects, where a client (consumer) object is
an instance of a class that sends a message to a server (supplier) object that executes one of its
methods in response to the request. Contracts are modeled on the legal notion of a contract,
where both parties, client and server objects, have obligations and rights. Practically speaking, a
contract is a set of constraints and guarantees. If the constraints are met, then the server object
guarantees certain behavior.31 Constraints can be written in a natural language (e.g., English),
a semiformal language (e.g., Structured English32), or a formal language (e.g., UML’s Object
Constraint Language). Given the need for precise, unambiguous specifi cation of constraints,
we recommend using UML’s Object Constraint Language.
Th e Object Constraint Language (OCL)33 is a complete language designed to specify con-
straints. In this section, we provide a short overview of some of the more useful constructs
contained in the language (see Figure 8-18). Essentially, all OCL expressions are simply a
declarative statement that evaluates to either being true or false. If the expression evaluates
to true, then the constraint has been satisfi ed. For example, if a customer had to have a less
than a one hundred dollar balance owed to be allowed to place another credit order, the OCL
expression would be:
balance owed ,5 100.00
OCL also has the ability to traverse relationships between objects, e.g., if the amount on a
purchase order is required to be the sum of the values of the individual purchase order lines,
this can be modeled as:
amount 5 OrderLine.sum(getPrice())
OCL also provides the ability to model more-complex constraints with a set of logical opera-
tors: and, or, xor, and not. For example, if customers were to be given a discount only if they
were a senior citizen or a “prime” customer, OCL could be used to model the constraint as:
age . 65 or customerType 5 “prime”
31 Th e idea of using contracts in design evolved from the “Design by Contract” technique developed by Bertrand
Meyer. See Bertrand Meyer, Object-Oriented Soft ware Construction (Englewood Cliff s, NJ: Prentice Hall, 1988).
32 We describe Structured English with Method Specifi cation later in this chapter.
33 For a complete description of the object constraint language, see Jos Warmer and Anneke Kleppe, Th e Object
Constraint Language: Precise Modeling with UML (Reading, MA: Addison-Wesley, 1999).
Constraints and Contracts 305
OCL provides many other constructs that can be used to build unique constraints. Th ese
include math-oriented operators, string operators, and relationship traversal operators. For
example, if the printed name on a customer order should be the concatenation of the custom-
er’s fi rst name and last name, then OCL could represent this constraint as:
printedName 5 fi rstName.concat(lastName)
We already have seen an example of the ‘.’ operator being used to traverse a relationship from
Order to OrderLine above. Th e ‘::’ operator allows the modeling of traversing inheritance
relationships.
OCL also provides a set of operations that are used to support constraints over a collection
of objects. For example, we demonstrated the use of the sum() operator above where we wanted
to guarantee that the amount was equal to the summation of all of the prices of the items in the
collection. Th e size operation returns the number of items in the collection. Th e count operation
returns the number of occurrences in the collection of the specifi c object passed as its argument.
Th e includes operation tests whether the object passed to it is already included in the collection.
Th e isEmpty operation determines whether the collection is empty or not. Th e select operation
provides support to model the identifi cation of a subset of the collection based on the expression
that is passed as its argument. Obviously, OCL provides a rich set of operators and operations in
which to model constraints.
Comparison 5 a 5 5
, a , 100
,5 a ,5 100
. a . 100
.5 a . 5 100
,. a ,. 100
Logical and a and b
or a or b
xor a xor b
not not a
Math 1 a 1 b
2 a 2 b
* a * b
/ a / b
String concat a 5 b.concat(c)
Relationship Traversal .
::
relationshipAttributeName.b
superclassName::propertyName
Collection size a.size
count(object) a.count(b)
includes(object) a.includes(b)
isEmpty a.isEmpty
sum() a.sum(b,c,d)
select(expression) a.select(b . d)
FIGURE 8-18
Sample OCL
Constructs
Operator Type Operator Example
306 Chapter 8 Class and Method Design
Types of Constraints
Th ree diff erent types of constraints are typically captured in object-oriented design: precon-
ditions, postconditions, and invariants.
Contracts are used primarily to establish the preconditions and postconditions for a
method to be able to execute properly. A precondition is a constraint that must be met for
a method to execute. For example, the parameters passed to a method must be valid for the
method to execute. Otherwise, an exception should be raised. A postcondition is a constraint
that must be met aft er the method executes, or the eff ect of the method execution must be
undone. For example, the method cannot make any of the attributes of the object take on
an invalid value. In this case, an exception should be raised, and the eff ect of the method’s
execution should be undone.
Whereas preconditions and postconditions model the constraints on an individual
method, invariants model constraints that must always be true for all instances of a class.
Examples of invariants include domains or types of attributes, multiplicity of attributes, and
the valid values of attributes. Th is includes the attributes that model association and aggrega-
tion relationships. For example, if an association relationship is required, an invariant should
be created that will enforce it to have a valid value for the instance to exist. Invariants are
normally attached to the class. We can attach invariants to the CRC cards or class diagram by
adding a set of assertions to them.
In Figure 8-19, the back of the CRC card constrains the attributes of an Order to specifi c
types. For example, Order Number must be an unsigned long, and Customer must be an
instance of the Customer class. Furthermore, additional invariants were added to four of the
attributes. For example, Cust ID must not only be an unsigned long, but it also must have one
and only one value [i.e., a multiplicity of (1..1)], and it must have the same value as the result
of the GetCustID() message sent to the instance of Customer stored in the Customer attrib-
ute. Also shown is the constraint for an instance to exist, an instance of the Customer class,
an instance of the State class, and at least one instance of the Product class must be associated
with the Order object (see the Relationships section of the CRC card where the multiplicities
are 1..1, 1..1, and 1..*, respectively). Figure 8-20 portrays the same set of invariants on a class
diagram. However, if all invariants are placed on a class diagram, the diagram becomes very
diffi cult to understand. Consequently, we recommend either extending the CRC card to doc-
ument the invariants instead of attaching them all to the class diagram or creating a separate
text document that contains them (see Figure 8-21).
Elements of a Contract
Contracts document the message passing that takes place between objects. Technically speak-
ing, a contract should be created for each message sent and received by each object, one for
each interaction. However, there would be quite a bit of duplication if this were done. In
practice, a contract is created for each method that can receive messages from other objects
(i.e., one for each visible method).
A contract should contain the information necessary for a programmer to understand
what a method is to do (i.e., they are declarative in nature). Th is information includes the
method name, class name, ID number, client objects, associated use cases, description, argu-
ments received, type of data returned, and the pre- and postconditions.34 Contracts do not
34 Currently, there is no standard format for a contract. Th e contract in Figure 8-22 is based on material contained in
Ian Graham, Migrating to Object Technology (Reading, MA: Addison-Wesley, 1995); Craig Larman, Applying UML
and Patterns: An Introduction to Object-Oriented Analysis and Design (Englewood Cliff s, NJ: Prentice Hall, 1998);
Meyer, Object-Oriented Soft ware Construction; R. Wirfs-Brock, B. Wilkerson, and L. Wiener, Designing Object-
Oriented Soft ware (Englewood Cliff s, NJ: Prentice Hall, 1990).
Constraints and Contracts 307
Front:
Class Name: Order ID: 2
Calculate tax
Calculate subtotal
Calculate shipping
Calculate total
Responsibilities
Associated Use Cases: 3Description: An Individual who needs to receive or has
received medical attention
Type: Concrete, Domain
Collaborators
(a)
Back:
Attributes:
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: Customer {1..1} State {1..1}Product {1..*}
(b)
Order Number (1..1) (unsigned long)
Date (1..1) (Date)
Sub Total (0..1) (double) {Sub Total = ProductOrder. sum(GetExtension())}
Tax (0..1) (double) (Tax = State.GetTaxRate() * Sub Total)
Shipping (0..1) (double)
Total (0..1) (double)
Customer (1..1) (Customer)
Cust ID (1..1) (unsigned long) {Cust ID = Customer. GetCustID()}
State (1..1) (State)
StateName (1..1) (String) {State Name = State. GetState()}
FIGURE 8-19
Invariants on a
CRC Card
308 Chapter 8 Class and Method Design
have a detailed algorithmic description of how the method is to work. Detailed algorithmic
descriptions typically are documented in a method specifi cation (as described later in this
chapter). In other words, a contract is composed of the information required for the devel-
oper of a client object to know what messages can be sent to the server objects and what the
client can expect in return. Figure 8-22 shows a sample format for a contract.
Because each contract is associated with a specifi c method and a specifi c class, the con-
tract must document them. Th e ID number of the contract is used to provide a unique iden-
tifi er for every contract. Th e Clients (Consumers) element of a contract is a list of classes and
Order
-Order Number[1..1] : unsigned long
-Date[1..1] : Date
-SubTotal[0..1] : double
-Tax[0..1] : double
-Shipping[0..1] : double
-Total[0..1] : double
-Customer[1..1] : Customer
-Cust ID[1..1] : unsigned long
-State[1..1] : State
-StateName[1..1] : String
Product
-Product Number
-Product Desc
-Price
Customer
-Cust ID
-Last Name
-First Name 1..1
1..1
0..*
1..*0..*
0..*
State
-State
-TaxRate
<>
{Cust ID = Customer.GetCustID()}
<>
{State Name = State.GetState()}
<>
{Tax =State.GetTaxRate()*SubTotal}
<>
{Sub Total = Product Order.sum(GetExtension())}
Product Order
Order
Product
Qty
Extension
FIGURE 8-20 Invariants on a Class Diagram
Order class invariants:
Cust ID 5 Customer.GetCustID()
State Name 5 Sate.GetState()
Sub Total 5 ProductOrder.sum(GetExtension())
Tax 5 State.GetTaxRate() * Sub Total1
FIGURE 8-21
Invariants in a Text File
Constraints and Contracts 309
methods that send a message to this specifi c method. Th is list is determined by reviewing the
sequence diagrams associated with the server class. Th e Associated Use Cases element is a list
of use cases in which this method is used to realize the implementation of the use case. Th e
use cases listed here can be found by reviewing the server class’s CRC card and the associated
sequence diagrams.
Th e Description of Responsibilities provides an informal description of what the
method is to perform, not how it is to do it. Th e arguments received are the data types of the
parameters passed to the method, and the value returned is the data type of the value that
the method returns to its clients. Together with the method name, they form the signature
of the method.
Th e precondition and postcondition elements are where the pre- and postconditions for
the method are recorded. Recall that pre- and postconditions can be written in a natural lan-
guage, a semiformal language, or a formal language. As with invariants, we recommend that
you use UML’s Object Constraint Language.35
Example In this example, we return to the order example shown in Figures 8-15, 8-19,
8-20, and 8-21. In this case, we limit the discussion to the design of the addOrder method
for the Customer class. Th e fi rst decision we must make is how to specify the design of the
relationship from Customer to Order. By reviewing Figures 8-15, 8-19, and 8-20, we see that
the relationship has a multiplicity of 0..* which means that an instance of customer may exist
without having any orders or an instance of customer could have many orders. As shown
35 See Warmer and Kleppe, Th e Object Constraint Language: Precise Modeling with UML.
Method Name: Class Name:
Clients (Consumers):
ID:
Associated Use Cases:
Type of Value Returned:
Description of Responsibilities:
Arguments Received:
Pre-Conditions:
Post-Conditions:
FIGURE 8-22
Sample Contract Form
310 Chapter 8 Class and Method Design
in Figure 8-15c, the relationship has been converted to an attribute that can contain many
instances of the Order class.
However, an important question that would not typically come up during analysis is
whether the order objects should be kept in sorted order or not. Another question that is
necessary to have answered for design purposes is how many orders could be expected by
a customer. Th e answers to these two questions will determine how we should organize
the orders from the customer object’s perspective. If the number of orders is going to be
relatively small and the orders don’t have to be kept in sorted order, then using a built-in
programming language construct such as a vector is suffi cient. However, if the number of
orders is going to be large or the orders must be kept in sorted order, then some form of a
sorted data structure, such as a linked list, is necessary. For example purposes, we assume
that a customer’s orders will need to be kept in sorted order and that there will be a large
number of them. Th erefore, instead of using a vector to contain the orders, we use a sorted
singly linked list.
To keep the design of the Customer class as close to the problem domain representation
as possible, the design of the Customer class is based on the Iterator pattern in Figure 8-13.
For simplicity purposes, we assume that an order is created before it is associated with the
specifi c customer. Otherwise, given the additional constraints of the instance of State class and
the instance of the Product Order class existing before an instance of Order can be created
would also have to be taken into consideration. Th is assumption allows us to ignore the fact
that an instance of State can have many orders, an instance of Order can have many instances
of Product Order associated with it, and an instance of Product can have many instances of
Product Order associated with it, which would require us to design many additional containers
(vectors or other data structures).
Based on all of the above, a new class diagram fragment was created that represents
a linked list-based relationship between instances of the Customer class and instances of
the Order class (see Figure 8-23). By carefully comparing Figures 8-15 and 8-23, we see
that the Iterator pattern idea has been included between the Customer and Order classes.
Th e domain of the Orders relationship-based attribute of the Customer class has been
replaced with OrderList to show that the list of orders will be contained in a list data struc-
ture. Figure 8-24 portrays an object diagram-based representation of how the relationship
between a customer instance and a set of order instances is stored in a sorted singly linked
list data structure. In this case, we see that a Customer object has an OrderList object asso-
ciated with it, each OrderList object could have N OrderNode objects, and each OrderNode
object will have an Order object. We see that each Order object is associated with a single
Customer object. By comparing Figures 8-15 and 8-24, we see that the intention of the mul-
tiplicity constraints of the Orders attribute of Customer, where a customer can have many
orders, and the multiplicity constraints of the Customer attribute of Orders is being mod-
eled correctly. Finally, notice that one of the operations contained in the OrderList class
is a private method. We will return to this specifi c point in the next section that addresses
method specifi cation.
Using Figures 8-22, 8-23, and 8-24, contracts for the addOrder method of the Customer
class and the insertOrder method for the OrderList class can be specifi ed (see Figure 8-25).
In the case of the addOrder method of the Customer class, we see that only instances of the
Order class use the method (see Clients section), that the method only implements part of
the logic that supports the addCustomerOrder use case (see Associated Use Cases section),
and that the contract includes a short description of the methods responsibilities. We also
see that the method receives a single argument of type Order and that it does not return any-
thing (void). Finally, we see that both a precondition and a postcondition were specifi ed. Th e
precondition simply states that the new Order object cannot be included in the current list
-O
rd
er
N
um
be
r[
1.
.1
]
: u
ns
ig
ne
d
lo
ng
-D
at
e[
1.
.1
]
: D
at
e
-S
ub
To
ta
l[
0.
.1
]
: d
ou
bl
e
-T
ax
[0
..1
]
: d
ou
bl
e
-S
hi
pp
in
g[
0.
.1
]
: d
ou
bl
e
-T
ot
al
[0
..1
]
: d
ou
bl
e
-C
us
to
m
er
[1
..1
]
: C
us
to
m
er
-S
ta
te
[1
..1
]
: S
ta
te
-P
ro
du
ct
O
rd
er
s[
1.
.*
]
: P
ro
du
ct
O
rd
er
O
rd
er
-c
re
at
eO
rd
er
Li
st
()
: O
rd
er
Li
st
+
ad
dO
rd
er
(in
a
nO
rd
er
:
O
rd
er
) :
v
oi
d
-C
us
t I
D
[1
..1
]
: u
ns
ig
ne
d
lo
ng
-L
as
t N
am
e[
1.
.1
]
: S
tr
in
g
-F
ir
st
N
am
e[
1.
.1
]
: S
tr
in
g
-O
rd
er
s[
1.
.1
]
: O
rd
er
Li
st
C
us
to
m
er
+
ad
va
nc
e(
) :
v
oi
d
+
be
gO
fL
is
t?
()
: b
oo
le
an
+
en
dO
fL
is
t?
()
: b
oo
le
an
+
em
pt
yL
is
t?
()
: b
oo
le
an
+
re
se
tL
is
t()
:
vo
id
+
ge
tC
ur
re
nt
O
rd
er
N
od
e(
) :
O
rd
er
N
od
e
-m
id
dl
eL
is
tIn
se
rt
(in
n
ew
O
rd
er
N
od
e
: O
rd
er
N
od
e)
:
vo
id
+
in
se
rt
O
rd
er
(in
a
nO
rd
er
:
O
rd
er
) :
v
oi
d
-F
ir
st
N
od
e[
0.
.1
]
: O
rd
er
N
od
e
-C
ur
re
nt
N
od
e[
0.
.1
]
: O
rd
er
N
od
e
-L
as
tN
od
e[
0.
.1
]
: O
rd
er
N
od
e
O
rd
er
Li
st
+
O
rd
er
N
od
e(
in
a
nO
rd
er
:
O
rd
er
)
+
ge
tO
rd
er
()
: O
rd
er
+
ge
tN
ex
tO
rd
er
N
od
e(
) :
O
rd
er
N
od
e
+
se
tN
ex
tO
rd
er
N
od
e(
in
a
nO
rd
er
N
od
e
: O
rd
er
N
od
e)
:
vo
id
-N
ex
tN
od
e[
0.
.1
]
: O
rd
er
N
od
e
-O
rd
er
[1
..1
]
: O
rd
er
O
rd
er
N
od
e
FI
G
U
R
E
8
-2
3
C
la
ss
D
ia
gr
am
F
ra
gm
en
t
o
f
th
e
C
u
st
o
m
er
t
o
O
rd
er
R
el
at
io
n
sh
ip
M
o
d
el
ed
a
s
a
So
rt
ed
S
in
gl
y
Li
n
ke
d
L
is
t
311
312 Chapter 8 Class and Method Design
FIGURE 8-24 Object Diagram of the Customer to Order Relationship Modeled as a Sorted Singly
Linked List
OrderList OrderNode1
OrderNode2
OrderNode3
Order1
Order2
Order3
OrderNodeN OrderN
Customer
of Orders; that is, the order cannot have previously been associated with this customer. Th e
postcondition, on the other hand, specifi es that the new list of orders must be equal to the old
list of orders (@pre) plus the new order object (including).
Th e contract for the insertOrder method for the OrderList class is somewhat simpler
than the addOrder method’s contract. From a practical perspective, the insertOrder method
implements part of the addOrder method’s logic. Specifi cally speaking, it implements that
actual insertion of the new order object into the specifi c data structure chosen to manage
the list of Order objects associated with the specifi c Customer object. Consequently, because
we already have specifi ed the precondition and postcondition for the addOrder method, we
do not have to further specify the same constraints for the insertOrder method. However,
this does implicitly increase the dependence of the Customer objects on the implementation
chosen for the list of customer orders. Th is is a good example of moving from the problem
domain to the solution domain. While we were focusing on the problem domain during
analysis, the actual implementation of the list of orders was never considered. However,
because we now are designing the implementation of the relationship between the Customer
objects and the Order objects, we have had to move away from the language of the end user
and toward the language of the programmer. During design, the focus moves toward opti-
mizing the code to run faster on the computer and not worrying about the end user’s ability
to understand the inner workings of the system; from an end user’s perspective, the system
should become more of a black box with which they interact. As we move farther into the
detailed design of the implementation of the problem domain classes, some solution domain
classes, such as the approach to implement relationships, will creep into the specifi cation of
Constraints and Contracts 313
Method Name: Class Name: ID:
Associated Use Cases:
Clients (consumers):
Type of Value Returned:
Description of Responsibilities:
Arguments Received:
Pre-Conditions:
Post-Conditions:
addOrder Customer 36
Order
addCustomerOrder
anOrder:Order
void
not orders.includes(anOrder)
Orders = Orders@pre.including(anOrder)
Implement the necessary behavior to add a new order to an existing customer keeping
the orders in sorted order by the order’s order number.
Method Name: Class Name: ID:
Associated Use Cases:
Type of Value Returned:
Description of Responsibilities:
Arguments Received:
Pre-Conditions:
Post-Conditions:
insertOrder OrderList 123
Customer
addCustomerOrder
anOrder:Order
void
None.
None.
Implement inserting an Order object into an OrderNode object and manage the
insertion of the OrderNode object into the current location in the sorted singly
linked list of orders.
Clients (consumers):
FIGURE 8-25
Sample Contract for the
addOrder Method of
the Customer Class and
the insertOrder Method
of the OrderList Class
314 Chapter 8 Class and Method Design
the problem domain layer. In this particular example, the OrderList and OrderNode classes
also could be used to implement the relationships from State objects to Order objects, from
Order Objects to Product Order objects, and from Product objects to Product Order objects
(see Figure 8-15). Given our simple example, one can clearly see that specifying the design
of the problem domain layer could include many additional solution domain classes to be
specifi ed on the problem domain layer.
METHOD SPECIFICATION
Once the analyst has communicated the big picture of how the system needs to be put
together, he or she needs to describe the individual classes and methods in enough detail so
that programmers can take over and begin writing code. Methods on the CRC cards, class
diagram, and contracts are described using method specifi cations. Method specifi cations are
written documents that include explicit instructions on how to write the code to implement
the method. Typically, project team members write a specifi cation for each method and then
pass them all along to programmers who write the code during implementation of the project.
Specifi cations need to be very clear and easy to understand, or programmers will be slowed
down trying to decipher vague or incomplete instructions.
Th ere is no formal syntax for a method specifi cation, so every organization uses its
own format, oft en using a form like the one in Figure 8-26. Typical method specifi cation
forms contain four components that convey the information that programmers will need
for writing the appropriate code: general information, events, message passing, and algo-
rithm specifi cation.
General Information
Th e top of the form in Figure 8-26 contains general information, such as the name of the
method, name of the class in which this implementation of the method will reside, ID num-
ber, Contract ID (which identifi es the contract associated with this method implementation),
programmer assigned, the date due, and the target programming language. Th is information
is used to help manage the programming eff ort.
Events
Th e second section of the form is used to list the events that trigger the method. An event
is a thing that happens or takes place. Clicking the mouse generates a mouse event, press-
ing a key generates a keystroke event—in fact, almost everything the user does generates
an event.
In the past, programmers used procedural programming languages that contained
instructions that were implemented in a predefi ned order, as determined by the computer
system, and users were not allowed to deviate from the order. Many programs today are
event driven (e.g., programs written in languages such as Visual Basic, Objective C, C11, or
Java), and event-driven programs include methods that are executed in response to an event
initiated by the user, system, or another method. Aft er initialization, the system waits for an
event to occur. When it does, a method is fi red that carries out the appropriate task, and then
the system waits once again.
We have found that many programmers still use method specifi cations when program-
ming in event-driven languages, and they include the event section on the form to capture
when the method will be invoked. Other programmers have switched to other design tools
that capture event-driven programming instructions, such as the behavioral state machine
described in Chapter 6.
Method Specifi cation 315
Method Name:
Contract ID:
Class Name:
❏ Visual Basic ❏ Smalltalk ❏ C++ ❏ Java
ID:
Programmer: Date Due:
Programming Language:
Triggers/Events:
Algorithm Specification:
Misc. Notes:
Data Type: Notes:
Arguments Received:
Data Type: Notes:
Arguments Returned:
ClassName.MethodName: Data Type: Notes:
Messages Sent & Arguments Passed:
FIGURE 8-26
Method Specifi cation
Form
Message Passing
Th e next sections of the method specifi cation describe the message passing to and from the
method, which are identifi ed on the sequence and collaboration diagrams. Programmers
need to understand what arguments are being passed into, passed from, and returned by the
method because the arguments ultimately translate into attributes and data structures within
the actual method.
316 Chapter 8 Class and Method Design
Algorithm Specifi cations
Algorithm specifi cations can be written in Structured English or some type of formal lan-
guage.36 Structured English is simply a formal way of writing instructions that describe the
steps of a process. Because it is the fi rst step toward the implementation of the method, it
looks much like a simple programming language. Structured English uses short sentences
that clearly describe exactly what work is performed on what data. Th ere are many versions
of Structured English because there are no formal standards; each organization has its own
type of Structured English. Figure 8-27 shows some examples of commonly used Structured
English statements.
Action statements are simple statements that perform some action. An If statement
controls actions that are performed under different conditions, and a For statement
(or a While statement) performs some actions until some condition is reached. A Case
statement is an advanced form of an If statement that has several mutually exclusive
branches.
If the algorithm of a method is complex, a tool that can be useful for algorithm spec-
ification is UML’s activity diagram (see Figure 8-28 and Chapter 4). Recall that activity
diagrams can be used to specify any type of process. Obviously, an algorithm specification
represents a process. However, owing to the nature of object orientation, processes tend
to be highly distributed over many little methods over many objects. Needing to use an
activity diagram to specify the algorithm of a method can, in fact, hint at a problem in
the design. For example, the method should be further decomposed or there could be
missing classes.
The last section of the method specification provides space for other information
that needs to be communicated to the programmer, such as calculations, special business
rules, calls to subroutines or libraries, and other relevant issues. This also can point out
36 For our purposes, Structured English will suffi ce. However, there has been some work with the Catalysis, Fusion,
and Syntropy methodologies to include formal languages, such as VDM and Z, into specifying object- oriented systems.
Profi ts 5 Revenues – Expenses
Action Statement Generate Inventory-Report
IF Customer Not in the Customer Object Store
THEN Add Customer record to Customer Object Store
If Statement ELSE Add Current-Sale to Customer’s Total-Sales
Update Customer record in Customer Object Store
FOR all Customers in Customer Object Store DO
For Statement Generate a new line in the Customer-Report
Add Customer’s Total-Sales to Report-Total
CASE
IF Income < 10,000: Marginal-tax-rate = 10 percent
IF Income < 20,000: Marginal-tax-rate = 20 percent
Case Statement IF Income < 30,000: Marginal-tax-rate = 31 percent
IF Income < 40,000: Marginal-tax-rate = 35 percent
ELSE Marginal-Tax-Rate = 38 percent
ENDCASE
Common Statements Example
FIGURE 8-27
Structured English
Activity
Action
A decision node:
■ Is used to represent a test condition to ensure that the control flow or object flow
only goes down one path.
■ Is labeled with the decision criteria to continue down the specific path.
An object flow:
■ Shows the flow of an object from one activity (or action) to another activity
(or action).
A control flow:
■ Shows the sequence of execution.
A final-activity node:
■ Is used to stop all control flows and object flows in an activity (or action).
A merge node:
■ Is used to bring back together different decision paths that were created using a
decision node.
An initial node:
■ Portrays the beginning of a set of actions or activities.
A final-flow node:
■ Is used to stop a specific control flow or object flow.
Class Name
[Decision
Criteria]
[Decision
Criteria]
An action:
■ Is a simple, nondecomposable piece of behavior.
■ Is labeled by its name.
An activity:
■ Is used to represent a set of actions.
■ Is labeled by its name.
An object node:
■ Is used to represent an object that is connected to a set of object flows.
■ Is labeled by its class name.
A Swimlane:
Is used to break up an activity diagram into rows and columns to assign the
individual activities (or actions) to the individuals or objects that are responsible
for executing the activity (or action).
Is labeled with the name of the individual or object responsible.
A Fork node:
Is used to split behavior into a set of parallel or concurrent flows of activities
(or actions).
A Join node:
Is used to bring back together a set of parallel or concurrent flows of activities
(or actions).
Swimlane
FIGURE 8-28 Syntax for an Activity Diagram (Figure 4-7)
317
318 Chapter 8 Class and Method Design
Method Name:
Contract ID:
Class Name:
❏ Visual Basic ❏ Smalltalk ❏ C++ ❏ Java
ID:
Programmer: Date Due:
Programming Language:
Triggers/Events:
Algorithm Specification:
Misc. Notes:
Data Type: Notes:
Arguments Received:
Data Type:
Notes:
Arguments Returned:
ClassName.MethodName: Data Type: Notes:
Messages Sent & Arguments Passed:
insertOrder
Order
void
None.
OrderNode.new() Order
OrderNode.getOrder()
Order.getOrderNumber()
OrderNode.setNextNode() OrderNode
OrderNodeself.middleListInsert()
See Figures 8-30 and 8-31.
The new customer’s new order.
Customer places an order
123 J. Doe 1/1/12
OrderList 100
FIGURE 8-29 Method Specifi cation for the insertOrder Method
changes or improvements that will be made to any of the other design documentation
based on problems that the analyst detected during the specification process.37
Example
Th is example continues the addition of a new order for a customer described in the
previous section (see Figure 8-29). Even though in most cases, because there are libraries
37 Remember that the development process is very incremental and iterative. Th erefore, changes could be cascaded back
to any point in the development process (e.g., to use-case descriptions, use-case diagrams, CRC cards, class diagrams,
object diagrams, sequence diagrams, communication diagrams, behavioral state machines, and package diagrams).
Verifying and Validating Class and Method Design 319
of data structure classes available that you could simply reuse and therefore would not
need to specify the algorithm to insert into a sorted singly linked list, we use it as an example
of how method specifi cation can be accomplished. Th e general information section of the
specifi cation documents the method’s name, its class, its unique ID number, the ID num-
ber of its associated contract, the programmer assigned, the date that its implementation
is due, and the programming language to be used. Second, the trigger/event that caused
this method to be executed is identifi ed. Th ird, the data type of the argument passed to this
method is documented (Order). Fourth, owing to the overall complexity of inserting a new
node into the list, we have factored out one specifi c aspect of the algorithm into a separate
private method (middleListInsert()) and we have specifi ed that this method will be sending
messages to instances of the OrderNode class and the Order class. Fift h, we specify the type
of return value that insertOrder will produce. In this case, the insertOrder method will not
return anything (void). Finally, we specify the actual algorithm. In this example, for the sake
of completeness, we provide both a Structured English–based (see Figure 8-30) and an activ-
ity diagram–based algorithm specifi cation (see Figure 8-31). Previously, we stated that we had
factored out the logic of inserting into the middle of the list into a separate private method:
middleList Insert(). Figure 8-32 shows the logic of this method. Imagine collapsing this logic
back into the logic of the insertOrder method, i.e., replace the middleListInsert(newOrder-
Node) activity in Figure 8-31 with the contents of Figure 8-32. Obviously, the insertOrder
method would be more complex.
VERIFYING AND VALIDATING CLASS AND METHOD DESIGN
Like all of the previous problem domain models, the constraints, contracts, and method spec-
ifi cations need to be verifi ed and validated. Given that we are primarily dealing with the prob-
lem domain in this chapter, the constraints and contracts were derived from the functional
requirements and the problem domain representations. However, they are applicable to the
other layers. In that case, they would be derived from the solution domain representations
associated with the data management (Chapter 9), human–computer interaction (Chapter
10), and system architecture (Chapter 11) layers. Given all of the issues described earlier with
the design criteria (coupling, cohesion, and connascence), additional specifi cations, reuse
opportunities, design restructuring and optimization, and mapping to implementation lan-
guages, it is likely that many modifi cations have taken place to the analysis representations of
the problem domain. Consequently, virtually everything must be re-verifi ed and re-validated.
First, we recommend that a walkthrough of all of the evolved problem domain representa-
tions be performed. Th at is, all functional models (Chapter 4) must be consistent; all structural
Create new OrderNode with the new Order
IF emptyList?()
FirstNode 5 LastNode 5 CurrentNode 5 newOrderNode
ELSE IF newOrderNode.getOrder().getOrderNumber() , FirstNode.getOrder().getOrderNumber()
newOrderNode.setNextNode(FirstNode)
FirstNode 5 newOrderNode
ELSE IF newOrderNode.getOrder().getOrderNumber() . LastNode.getOrder().getOrderNumber()
LastNode.setNextNode(newOrderNode)
LastNode 5 newOrderNode
ELSE
middleListInsert(newOrderNode)
FIGURE 8-30
Structured English-
based Algorithm
Specifi cation for the
insertOrder Method
320
C
re
at
e
ne
w
O
rd
er
N
od
e
[e
m
pt
yL
is
t?
()]
Fi
rs
tN
od
e
=
L
as
tN
od
e
=
C
ur
re
nt
N
od
e
=
n
ew
O
rd
er
N
od
e
[n
ew
O
rd
er
N
od
e.
ge
tO
rd
er
().
ge
tO
rd
er
N
um
be
r(
)
L
as
tN
od
e.
ge
tO
rd
er
().
ge
tO
rd
er
N
um
be
r(
)]
ne
w
O
rd
er
N
od
e.
se
tN
ex
tN
od
e(
Fi
rs
tN
od
e)
Fi
rs
tN
od
e
=
n
ew
O
rd
er
N
od
e
La
st
N
od
e
=
n
ew
O
rd
er
N
od
e
m
id
dl
eL
is
tI
ns
er
t(
ne
w
O
rd
er
N
od
e)
La
st
N
od
e.
se
tN
ex
tN
od
e(
ne
w
O
rd
er
N
od
e)
FI
G
U
R
E
8
-3
1
A
ct
iv
it
y
D
ia
gr
am
-b
as
ed
A
lg
o
ri
th
m
S
p
ec
ifi
ca
ti
o
n
f
o
r
th
e
in
se
rt
O
rd
er
M
et
h
o
d
Verifying and Validating Class and Method Design 321
models (Chapter 5) must be consistent; all behavioral models (Chapter 6) must be consistent;
and the functional, structural, and behavioral models must be balanced (Chapter 7).
Second, all constraints, contracts, and method specifi cations must be tested. Th e best way
to do this is to role-play the system using the diff erent scenarios of the use cases. In this case,
we must enforce the invariants on the evolved CRC cards (see Figure 8-19), the pre- and post-
conditions on the contract forms (see Figures 8-22 and 8-25), and the design of each method
specifi ed with the method specifi cation forms (see Figures 8-26 and 8-29) and algorithm
specifi cations (see Figures 8-30, 8-31, and 8-32).
Given the amount of verifying and validating the fi delity of all of the models that we
have performed on the evolving system, it might seem like overkill to perform the above
again. However, given the pure volume of changes that can take place during design, it is
crucial to thoroughly test the models again before the system is implemented. In fact, testing
is so important to the agile development approaches, testing forms the virtual backbone of
those methodologies. Without thorough testing, there is no guarantee that the system being
implemented will address the problem being solved. Once the system has been implemented,
testing becomes even more important (see Chapter 12).
[CurrentNode = NULL]
[CurrentNode.getNextNode().getOrder().getOrderNumber()
>newOrderNode.getOrder().getOrderNumber()]
[CurrentNode != NULL]
newOrderNode.setNextNode(CurrentNode.getNextNode())
CurrentNode.setNextNode(newOrderNode)
CurrentNode = NULL
advance()
resetList()
FIGURE 8-32 Activity Diagram–based Algorithm Specifi cation for the middleListInsert Method
322 Chapter 8 Class and Method Design
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
In this installment of the Patterson Superstore case, Ruby and her team have shift ed their
focus from capturing the requirements, behavior, and structure of the evolving system to
the design of the individual classes and method for the system. First, they had to return
once more to the functional, structural, and behavior models to ensure that the classes
defi ned in analysis (the problem domain layer) are both suffi cient and necessary. In eval-
uating these models, they checked for coupling, cohesion, and connascence. Th ey then
moved to designing the contracts and method specifi cations.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the basic characteristics of object orientation.
Describe the problems that can arise when using polymorphism and inheritance.
Describe the diff erent types of inheritance confl icts.
Describe the diff erent types of coupling and why coupling should be minimized.
Describe the law of Demeter.
Describe the diff erent types of cohesion and why cohesion should be maximized.
Describe connascence.
Identify opportunities for reuse through the use of patterns, frameworks, class libraries, and components.
Optimize a design.
Map the problem domain classes to a single-inheritance language.
Map the problem domain classes to an object-based language.
Understand the diffi culties in implementing an object-oriented design in a traditional programming language.
Use the OCL to defi ne precondition, postcondition, and invariant constraints.
Create contracts to specify the interaction between client and server objects.
Specify methods using the method specifi cation form.
Specify the logic of a method using Structured English and activity diagrams.
Understand how to verify and validate both the design of the classes and the design of their methods.
KEY TERMS
Active value
Activity diagram
API (application program
interface)
Attribute
Behavior
Class
Class cohesion
Class library
Client
Cohesion
Component
Connascence
Constraint
Consumer
Contract
Coupling
Derived attribute
Design pattern
Dynamic binding
Encapsulation
Event
Event driven
Exceptions
Factoring
Fan-out
Framework
Generalization/specialization
cohesion
Heteronyms
Homographs
Homonyms
Ideal class cohesion
Information hiding
Inheritance
Inheritance confl ict
Inheritance coupling
Instance
Interaction coupling
Invariant
Law of Demeter
Message
Method
Method cohesion
Method specifi cation
Multiple inheritance
Normalization
Object
Object-based language
Object constraint language
(OCL)
Operations
Exercises 323
EXERCISES
A. For the A Real Estate Inc. problem in Chapters 4 (exer-
cises I, J, and K), 5 (exercises P and Q), 6 (exercise D),
and 7 (exercise A):
1. Choose one of the classes and create a set of
invariants for attributes and relationships and add
them to the CRC card for the class.
2. Choose one of the methods in the class that you chose
and create a contract and a method specifi cation for
it. Use OCL to specify any pre- or postcondition and
use both Structured English and an activity diagram to
specify the algorithm.
Patterns
Polymorphism
Postcondition
Precondition
Redefi nition
Server
Signature
Single inheritance
State
Structured English
Supplier
Synonyms
Trigger
Visibility
QUESTIONS
1. What are the basic characteristics of object-oriented
systems?
2. What is dynamic binding?
3. Defi ne polymorphism. Give one example of a good
use of polymorphism and one example of a bad use of
polymorphism.
4. What is an inheritance confl ict? How does an inher-
itance confl ict aff ect the design?
5. Why is cancellation of methods a bad thing?
6. Give the guidelines to avoid problems with inher-
itance confl icts.
7. Why is it important to know which object-oriented
programming language is going to be used to imple-
ment the system?
8. What additional types of inheritance confl icts are
there when using multiple inheritance?
9. What is the law of Demeter?
10. What are the six types of interaction coupling? Give
one example of good interaction coupling and one
example of bad interaction coupling.
11. What are the seven types of method cohesion? Give
one example of good method cohesion and one
example of bad method cohesion.
12. What are the four types of class cohesion? Give one
example of each type.
13. What are the fi ve types of connascence described in
your text? Give one example of each type.
14. When designing a specifi c class, what types of addi-
tional specifi cation for a class could be necessary?
15. What are exceptions?
16. What are constraints? What are the three diff erent
types of constraints?
17. What are patterns, frameworks, class libraries, and
components? How are they used to enhance the
evolving design of the system?
18. How are factoring and normalization used in design-
ing an object system?
19. What are the diff erent ways to optimize an object
system?
20. What is the typical downside of system optimization?
21. What is the purpose of a contract? How are contracts
used?
22. What is the Object Constraint Language? What is its
purpose?
23. What is the Structured English? What is its purpose?
24. What is an invariant? How are invariants modeled in
a design of a class? Give an example of an invariant for
an hourly employee class using the Object Constraint
Language.
25. Create a contract for a compute pay method asso-
ciated with an hourly employee class. Specify the
preconditions and postconditions using the Object
Constraint Language.
26. How do you specify a method’s algorithm? Give an
example of an algorithm specifi cation for a compute
pay method associated with an hourly employee class
using Structured English.
27. How do you specify a method’s algorithm? Give an
example of an algorithm specifi cation for a compute
pay method associated with an hourly employee class
using an activity diagram.
28. How are methods specifi ed? Give an example of a
method specifi cation for a compute pay method asso-
ciated with an hourly employee class.
324 Chapter 8 Class and Method Design
MINICASES
1. Your boss has been in the soft ware development
fi eld for thirty years. He has always prided himself
on his ability to adapt his skills from one approach
to developing soft ware to the next approach. For
example, he had no problem learning structured
analysis and design in the early 1980s and information
B. For the A Video Store problem in Chapters 4 (exercises
L, M, N K), 5 (exercises R and S), 6 (exercise E), and 7
(exercise B):
1. Choose one of the classes and create a set of invari-
ants for attributes and relationships and add them to
the CRC card for the class.
2. Choose one of the methods in the class that you
chose and create a contract and a method spec-
ifi cation for it. Use OCL to specify any pre- or
postcondition and use both Structured English and
an activity diagram to specify the algorithm.
C. For the gym membership problem in Chapters 4
(exercises O, P, and Q), 5 (exercises T and U), 6
(exercise F), and 7 (exercise C):
1. Choose one of the classes and create a set of invari-
ants for attributes and relationships and add them to
the CRC card for the class.
2. Choose one of the methods in the class that you
chose and create a contract and a method spec-
ification for it. Use OCL to specify any pre- or
postcondition and use both Structured English
and an activity diagram to specify the algorithm.
D. For the Picnics R Us problem in Chapters 4 (exercises R,
S, and T), 5 (exercises V and W), 6 (exercise G), and 7
(exercise D):
1. Choose one of the classes and create a set of invari-
ants for attributes and relationships and add them to
the CRC card for the class.
2. Choose one of the methods in the class that you
chose and create a contract and a method specifi –
cation for it. Use OCL to specify any pre- or post-
condition and use both Structured English and an
activity diagram to specify the algorithm.
E. For the Of-the-Month-Club problem in Chapters 4
(exercises U, V, and W), 5 (exercises X and Y), 6
(exercise H), and 7 (exercise E):
1. Choose one of the classes and create a set of invari-
ants for attributes and relationships and add them to
the CRC card for the class.
2. Choose one of the methods in the class that you chose
and create a contract and a method specifi cation for
it. Use OCL to specify any pre- or postcondition and
use both Structured English and an activity diagram
to specify the algorithm.
F. Describe the diff erence in meaning between the follow-
ing two class diagrams. Which is a better model? Why?
Name-Address
Employee
PersonPerson
Employee
Name-Address
1..10..*
G. From a cohesion, coupling, and connascence perspec-
tive, is the following class diagram a good model? Why
or why not?
Customer
Credit Customer Cash Customer Check Customer
Person
H. From a cohesion, coupling, and connascence perspec-
tive, are the following class diagrams good models?
Why or why not?
Car
Car-Person
Person Robot
Robot-Employee
Employee
I. Create a set of inheritance confl icts for the two inher-
itance structures in the class diagrams of exercise H.
Minicases 325
engineering in the early 1990s. He even understands
the advantage of rapid application development. But
the other day, when you and he were talking about
the advantages of object-oriented approaches, he
became totally confused. He thought that character-
istics such as polymorphism and inheritance were
an advantage for object-oriented systems. However,
when you explained the problems with inheritance
confl icts, redefi nition capabilities, and the need for
semantic consistency across diff erent implementa-
tions of methods, he was ready to simply give up. To
make matters worse, you then went on to explain the
importance of contracts in controlling the develop-
ment of the system. At this point in the conservation,
he basically threw in the towel. As he walked off , you
heard him say something like “I guess it’s true, it’s too
hard to teach an old dog new tricks.”
Being a loyal employee and friend, you decided
to write a short tutorial to give your boss on object-
oriented systems development. As a fi rst step, create a
detailed outline for the tutorial. As a subtle example,
use good design criteria, such as coupling and cohe-
sion, in the design of your tutorial outline.
2. You have been working with the professional and sci-
entifi c management (PSSM) problem for quite a while.
You should go back and refresh your memory about
the problem before attempting to solve this situation.
Refer back to your solutions to Minicase 3 in Chapter 7.
a. For each class in the structural model, using OCL,
create a set of invariants for attributes and relation-
ships and add them to the CRC cards for the classes.
b. Choose one of the classes in the structural model.
Create a contract for each method in that class. Be
sure to use OCL to specify the preconditions and
the postconditions. Be as complete as possible.
c. Create a method specifi cation for each method
in the class you chose for question b. Use both
Structured English and activity diagrams for the
algorithm specifi cation.
3. You have been working with the Holiday Travel
Vehicle problem for quite a while. You should go back
and refresh your memory about the problem before
attempting to solve this situation. Refer back to your
solutions Minicase 4 in Chapter 7.
In the new system for Holiday Travel Vehicles,
the system users follow a two-stage process to record
complete information on all of the vehicles sold.
When an RV or trailer fi rst arrives at the company
from the manufacturer, a clerk from the inventory
department creates a new vehicle record for it in
the computer system. Th e data entered at this time
include basic descriptive information on the vehicle
such as manufacturer, name, model, year, base cost,
and freight charges. When the vehicle is sold, the new
vehicle record is updated to refl ect the fi nal sales terms
and the dealer-installed options added to the vehicle.
Th is information is entered into the system at the
time of sale when the salesperson completes the sales
invoice.
When it is time for the clerk to fi nalize the new
vehicle record, the clerk selects a menu option from
the system, which is called Finalize New Vehi-
cle Record. Th e tasks involved in this process are
described below.
When the user selects Finalize New Vehicle Record
from the system menu, the user is immediately
prompted for the serial number of the new vehicle.
Th is serial number is used to retrieve the new vehicle
record for the vehicle from system storage. If a record
cannot be found, the serial number is probably inva-
lid. Th e vehicle serial number is then used to retrieve
the option records that describe the dealer-installed
options that were added to the vehicle at the custom-
er’s request. Th ere may be zero or more options. Th e
cost of the option specifi ed on the option record(s) is
totaled. Th en, the dealer cost is calculated using the
vehicle’s base cost, freight charge, and total option
cost. Th e completed new vehicle record is passed back
to the calling module.
a. Update the structural model (CRC cards and class
diagram) with this additional information.
b. For each class in the structural model, using OCL,
create a set of invariants for attributes and relation-
ships and add them to the CRC cards for the classes.
c. Choose one of the classes in the structural model.
Create a contract for each method in that class. Be
sure to use OCL to specify the preconditions and
the postconditions. Be as complete as possible.
d. Create a method specifi cation for each method
in the class you chose for question b. Use both
Structured English and activity diagrams for the
algorithm specifi cation.
A project team designs the data management layer of a system using a four-step process:
selecting the format of the storage, mapping the problem domain classes to the selected
format, optimizing the storage to perform effi ciently, and then designing the necessary data
access and manipulation classes. Th is chapter describes the diff erent ways objects can be
stored and several important characteristics that should be considered when choosing among
object persistence formats. It describes a problem domain class to object persistence format
mapping process for the most important object persistence formats. Because the most pop-
ular storage format today is the relational database, the chapter focuses on the optimization
of relational databases from both storage and access perspectives. We describe the eff ect that
nonfunctional requirements have on the data-management layer. Th e chapter fi nally describes
how to design data access and manipulation classes and describes how to ensure the fi delity
of the data management layer.
OBJECTIVES
■ Become familiar with several object persistence formats.
■ Be able to map problem domain objects to diff erent object persistence formats.
■ Be able to apply the steps of normalization to a relational database.
■ Be able to optimize a relational database for object storage and access.
■ Become familiar with indexes for relational databases.
■ Be able to estimate the size of a relational database.
■ Understand the aff ect of nonfunctional requirements on the data management layer.
■ Be able to design the data access and manipulation classes.
INTRODUCTION
Applications are of little use without the data that they support. How useful is a multimedia
application that can’t support images or sound? Why would someone log into a system to fi nd
information if it took him or her less time to locate the information manually? One of the
leading complaints by end users is that the fi nal system is too slow, so to avoid such complaints
project team members must allow time during design to carefully make sure that the fi le or
database performs as fast as possible. At the same time, the team must keep hardware costs
down by minimizing the storage space that the application will require. Th e goals of maxi-
mizing access to the objects and minimizing the amount of space taken to store objects can
confl ict, and designing object persistence effi ciency usually requires trade-off s.
Th e design of the data management layer addresses these concerns. It includes both the
design of data access and manipulation classes and the actual data storage. Th e design of the data
access and manipulation classes should ensure the independence of the problem domain classes
from the data storage format. As such, the data access and manipulation classes handle all com-
munication with the database. In this manner, the problem domain is decoupled from the object
storage, allowing the object storage to be changed without aff ecting the problem domain classes.
C H A P T E R 9
Data Management Layer Design
326
Object Persistence Formats 327
Th e data storage component manages how data are stored and handled by the programs
that run the system. Th e data storage component is composed of a set of object persistence
classes. Eff ective object persistence design decreases the chances of ending up with ineffi cient
systems, long system response times, and users who cannot get to the information that they
need in the way that they need it—all of which can aff ect the success of the project. From a
practical perspective, there are fi ve basic types of formats that can be used to store objects for
application systems: fi les (sequential and random), object-oriented databases, object-relational
databases, relational databases, or NoSQL datastores.1 Each type has certain characteristics
that make it more appropriate for some types of systems over others. Once the object persis-
tence format is selected to support the system, the problem domain objects need to drive the
design of the actual object storage. Th en the object storage needs to be designed to optimize
its processing effi ciency.
OBJECT PERSISTENCE FORMATS
Each of the object persistence types is described in this section. Files are electronic lists of data
that have been optimized to perform a particular transaction. For example, Figure 9-1 shows a
customer order fi le with information about customers’ orders, in the form in which it is used,
so that the information can be accessed and processed quickly by the system.
A database is a collection of groupings of information, each of which is related to each
other in some way (e.g., through common fi elds). Logical groupings of information could
include such categories as customer data, information about an order, product information,
and so on. A database management system (DBMS) is soft ware that creates and manipulates
these databases (see Figure 9-2 for a relational database example). Such end-user DBMSs as
Microsoft Access support small-scale databases that are used to enhance personal productiv-
ity, whereas enterprise DBMSs, such as DB2, Versant, and Oracle, can manage huge volumes
of data and support applications that run an entire company. An end-user DBMS is signif-
icantly less expensive and easier for novice users to use than its enterprise counterpart, but
it does not have the features or capabilities that are necessary to support mission-critical or
large-scale systems.
Sequential and Random Access Files
From a practical perspective, most object-oriented programming languages support sequen-
tial and random access fi les as part of the language.2 In this section, we describe what sequen-
tial access and random access fi les are.3 We also describe how sequential access and random
access fi les are used to support an application. For example, they can be used to support
master fi les, look-up fi les, transaction fi les, audit fi les, and history fi les.
Sequential access fi les allow only sequential fi le operations to be performed (e.g., read,
write, and search). Sequential access fi les are very effi cient for sequential operations that
process all of the objects consecutively, such as report writing. However, for random opera-
tions, such as fi nding or updating a specifi c object, they are very ineffi cient. On the average,
50 percent of the contents of a sequential access fi le will have to be searched before fi nding
the specifi c object of interest in the fi le. Th ey come in two fl avors: ordered and unordered.
1 Th ere are other types of fi les, such as relative, indexed sequential, and multi-indexed sequential, and databases, such
as hierarchical, network, and multidimensional. However, these formats typically are not used for object persistence.
2 For example, see the FileInputStream, FileOutputStream, and RandomAccessFile classes in the java.io package.
3 For a more complete coverage of issues related to the design of fi les, see Owen Hanson, Design of Computer Data
Files (Rockville, MD: Computer Science Press, 1982).
328 Chapter 9 Data Management Layer Design
234 11/23/00 2242 DeBerry Ann $ 90.00 $5.85 $ 95.85 Y MC
235 11/23/00 9500 Chin April $ 12.00 $0.60 $ 12.60 Y VISA
236 11/23/00 1556 Fracken Chris $ 50.00 $2.50 $ 52.50 N VISA
237 11/23/00 2242 DeBerry Ann $ 75.00 $4.88 $ 79.88 Y AMEX
238 11/23/00 2242 DeBerry Ann $ 60.00 $3.90 $ 63.90 Y MC
239 11/23/00 1035 Black John $ 90.00 $4.50 $ 94.50 Y AMEX
240 11/23/00 9501 Kaplan Bruce $ 50.00 $2.50 $ 52.50 N VISA
241 11/23/00 1123 Williams Mary $120.00 $9.60 $129.60 N MC
242 11/24/00 9500 Chin April $ 60.00 $3.00 $ 63.00 Y VISA
243 11/24/00 4254 Bailey Ryan $ 90.00 $4.50 $ 94.50 Y VISA
244 11/24/00 9500 Chin April $ 24.00 $1.20 $ 25.20 Y VISA
245 11/24/00 2242 DeBerry Ann $ 12.00 $0.78 $ 12.78 Y AMEX
246 11/24/00 4254 Bailey Ryan $ 20.00 $1.00 $ 21.00 Y MC
247 11/24/00 2241 Jones Chris $ 50.00 $2.50 $ 52.50 N VISA
248 11/24/00 4254 Bailey Ryan $ 12.00 $0.60 $ 12.60 Y AMEX
249 11/24/00 5927 Lee Diane $ 50.00 $2.50 $ 52.50 N AMEX
250 11/24/00 2242 DeBerry Ann $ 12.00 $0.78 $ 12.78 Y MC
251 11/24/00 9500 Chin April $ 15.00 $0.75 $ 15.75 Y MC
252 11/24/00 2242 DeBerry Ann $132.00 $8.58 $140.58 Y MC
253 11/24/00 2242 DeBerry Ann $ 72.00 $4.68 $ 76.68 Y AMEX
Order Cust Last First Prior Payment
Number Date ID Name Name Amount Tax Total Customer Type
FIGURE 9-1 Customer Order File
An unordered sequential access fi le is basically an electronic list of information stored on
disk. Unordered fi les are organized serially (i.e., the order of the fi le is the order in which the
objects are written to the fi le). Typically, new objects simply are added to the fi le’s end.
Ordered sequential access fi les are placed into a specifi c sorted order (e.g., in ascending
order by customer number). Th ere is overhead associated with keeping fi les in a particular
sorted order. Th e fi le designer can keep the fi le in sorted order by always creating a new fi le
each time a delete or addition occurs, or he or she can keep track of the sorted order via the use
of a pointer, which is information about the location of the related record. A pointer is placed at
the end of each record, and it “points” to the next record in a series or set. Th e underlying data/
fi le structure in this case is the linked list4 data structure demonstrated in the previous chapter.
Random access fi les allow only random or direct fi le operations to be performed. Th is
type of fi le is optimized for random operations, such as fi nding and updating a specifi c object.
Random access fi les typically have a faster response time to fi nd and update operations than
any other type of fi le. However, because they do not support sequential processing, applica-
tions such as report writing are very ineffi cient. Th e various methods to implement random
access fi les are beyond the scope of this book.5
4 For more information on various data structures, see Ellis Horowitz and Sartaj Sahni, Fundamentals of Data Struc-
tures (Rockville, MD: Computer Science Press, 1982); Michael T. Goodrich and Roberto Tamassia, Data Structures
and Algorithms in Java (New York: Wiley, 1998).
5 For a more-detailed look at the underlying data and fi le structures of the diff erent types of fi les, see Mary E. S.
Loomis, Data Management and File Structures, 2nd Ed. (Englewood Cliff s, NJ: Prentice Hall, 1989); Michael J. Folk
and Bill Zoeellick, File Structures: A Conceptual Toolkit (Reading, MA: Addison-Wesley, 1987).
FI
G
U
R
E
9
-2
C
u
st
o
m
er
O
rd
er
D
at
ab
as
e
C
us
t
La
st
Fi
rs
t
Pr
io
r
ID
N
am
e
N
am
e
C
us
to
m
er
2
24
2
D
eB
er
ry
A
nn
Y
9
50
0
C
hi
n
A
pr
il
Y
1
55
6
Fr
ac
ke
n
C
hr
is
N
1
03
5
B
la
ck
Jo
hn
Y
9
50
1
K
ap
la
n
B
ru
ce
N
1
12
3
W
ill
ia
m
s
M
ar
y
N
4
25
4
B
ai
le
y
R
ya
n
Y
2
24
1
Jo
ne
s
C
hr
is
N
5
92
7
Le
e
D
ia
ne
N
C
us
to
m
er
Pa
ym
en
t
Pa
ym
en
t
Ty
pe
D
es
c
M
C
M
as
te
rc
ar
d
V
IS
A
V
is
a
A
M
EX
A
m
er
ic
an
E
xp
re
ss
Pa
ym
en
t T
yp
e
O
rd
er
C
us
t
Pa
ym
en
t
N
um
be
r
D
at
e
ID
A
m
ou
nt
Ta
x
To
ta
l
Ty
pe
23
4
11
/2
3/
00
22
42
$
90
.0
0
$5
.8
5
$
95
.8
5
M
C
23
5
11
/2
3/
00
95
00
$
12
.0
0
$0
.6
0
$
12
.6
0
V
IS
A
23
6
11
/2
3/
00
15
56
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
23
7
11
/2
3/
00
22
42
$
75
.0
0
$4
.8
8
$
79
.8
8
A
M
EX
23
8
11
/2
3/
00
22
42
$
60
.0
0
$3
.9
0
$
63
.9
0
M
C
23
9
11
/2
3/
00
10
35
$
90
.0
0
$4
.5
0
$
94
.5
0
A
M
EX
24
0
11
/2
3/
00
95
01
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
24
1
11
/2
3/
00
11
23
$ 1
20
.0
0
$9
.6
0
$ 1
29
.6
0
M
C
24
2
11
/2
4/
00
95
00
$
60
.0
0
$3
.0
0
$
63
.0
0
V
IS
A
24
3
11
/2
4/
00
42
54
$
90
.0
0
$4
.5
0
$
94
.5
0
V
IS
A
24
4
11
/2
4/
00
95
00
$
24
.0
0
$1
.2
0
$
25
.2
0
V
IS
A
24
5
11
/2
4/
00
22
42
$
12
.0
0
$0
.7
8
$
12
.7
8
A
M
EX
24
6
11
/2
4/
00
42
54
$
20
.0
0
$1
.0
0
$
21
.0
0
M
C
24
7
11
/2
4/
00
22
41
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
24
8
11
/2
4/
00
42
54
$
12
.0
0
$0
.6
0
$
12
.6
0
A
M
EX
24
9
11
/2
4/
00
59
27
$
50
.0
0
$2
.5
0
$
52
.5
0
A
M
EX
25
0
11
/2
4/
00
22
42
$
12
.0
0
$0
.7
8
$
12
.7
8
M
C
25
1
11
/2
4/
00
95
00
$
15
.0
0
$0
.7
5
$
15
.7
5
M
C
25
2
11
/2
4/
00
22
42
$ 1
32
.0
0
$8
.5
8
$ 1
40
.5
8
M
C
25
3
11
/2
4/
00
22
42
$
72
.0
0
$4
.6
8
$
76
.6
8
A
M
EX
O
rd
er
Ta
bl
es
r
el
at
ed
th
ro
ug
h
C
us
t I
D
Ta
bl
es
r
el
at
ed
th
ro
ug
h
Pa
ym
en
t T
yp
e
329
330 Chapter 9 Data Management Layer Design
Th ere are times when it is necessary to be able to process fi les in both a sequential and ran-
dom manner. One simple way to do this is to use a sequential fi le that contains a list of the keys
(the fi eld in which the fi le is to be kept in sorted order) and a random access fi le for the actual
objects. Th is minimizes the cost of additions and deletions to a sequential fi le while allowing the
random fi le to be processed sequentially by simply passing the key to the random fi le to retrieve
each object in sequential order. It also allows fast random processing to occur by using only
the random access fi le, thus optimizing the overall cost of fi le processing. However, if a fi le of
objects needs to be processed in both a random and sequential manner, the developer should
consider using a database (relational, object-relational, or object-oriented) instead.
Th ere are many diff erent application types of fi les—e.g., master fi les, lookup fi les, transac-
tion fi les, audit fi les, and history fi les. Master fi les store core information that is important to the
business and, more specifi cally, to the application, such as order information or customer mail-
ing information. Th ey usually are kept for long periods of time, and new records are appended
to the end of the fi le as new orders or new customers are captured by the system. If changes need
to be made to existing records, programs must be written to update the old information.
Lookup fi les contain static values, such as a list of valid ZIP codes or the names of the U.S.
states. Typically, the list is used for validation. For example, if a customer’s mailing address is
entered into a master fi le, the state name is validated against a lookup fi le that contains U.S.
states to make sure that the operator entered the value correctly.
A transaction fi le holds information that can be used to update a master fi le. Th e transac-
tion fi le can be destroyed aft er changes are added, or the fi le may be saved in case the trans-
actions need to be accessed again in the future. Customer address changes, for one, would be
stored in a transaction fi le until a program is run that updates the customer address master
fi le with the new information.
For control purposes, a company might need to store information about how data
change over time. For example, as human resources clerks change employee salaries in a
human resources system, the system should record the person who made the changes to the
salary amount, the date, and the actual change that was made. An audit fi le records before
and aft er images of data as they are altered so that an audit can be performed if the integrity
of the data is questioned.
Sometimes fi les become so large that they are unwieldy, and much of the information
in the fi le is no longer used. Th e history fi le (or archive fi le) stores past transactions (e.g., old
customers, past orders) that are no longer needed by system users. Typically the fi le is stored
off -line, yet it can be accessed on an as-needed basis. Other fi les, such as master fi les, can then
be streamlined to include only active or very recent information.
Relational Databases
A relational database is the most popular kind of database for application development today.
A relational database is based on collections of tables with each table having a primary key—a
fi eld or fi elds whose values are unique for every row of the table. Th e tables are related to one
another by placing the primary key from one table into the related table as a foreign key (see
Figure 9-3). Most relational database management systems (RDBMS) support referential integ-
rity, or the idea of ensuring that values linking the tables together through the primary and
foreign keys are valid and correctly synchronized. For example, if an order-entry clerk using
the tables in Figure 9-3 attempted to add order 254 for customer number 1111, he or she would
have made a mistake because no customer exists in the Customer table with that number. If the
RDBMS supported referential integrity, it would check the customer numbers in the Customer
table, discover that the number 1111 is invalid, and return an error to the entry clerk. Th e clerk
would then go back to the original order form and recheck the customer information.
FI
G
U
R
E
9
-3
R
el
at
io
n
al
D
at
ab
as
e
C
us
t
La
st
Fi
rs
t
Pr
io
r
ID
N
am
e
N
am
e
C
us
to
m
er
2
24
2
D
eB
er
ry
A
nn
Y
9
50
0
C
hi
n
A
pr
il
Y
1
55
6
Fr
ac
ke
n
C
hr
is
N
1
03
5
B
la
ck
Jo
hn
Y
9
50
1
K
ap
la
n
B
ru
ce
N
1
12
3
W
ill
ia
m
s
M
ar
y
N
4
25
4
B
ai
le
y
R
ya
n
Y
2
24
1
Jo
ne
s
C
hr
is
N
5
92
7
Le
e
D
ia
ne
N
C
us
to
m
er
Pa
ym
en
t
Pa
ym
en
t
Ty
pe
D
es
c
M
C
M
as
te
rc
ar
d
V
IS
A
V
is
a
A
M
EX
A
m
er
ic
an
E
xp
re
ss
Pa
ym
en
t T
yp
e
R
ef
er
en
ti
al
I
nt
eg
ri
ty
:
■
A
ll
pa
ym
en
t t
yp
e
va
lu
es
in
O
rd
er
m
us
t e
xi
st
fi
rs
t i
n
th
e
Pa
ym
en
t T
yp
e
ta
bl
e
■
A
ll
C
us
t I
D
v
al
ue
s
in
o
rd
er
m
us
t e
xi
st
fi
rs
t i
n
th
e
C
us
to
m
er
ta
bl
e
O
rd
er
C
us
t
Pa
ym
en
t
N
um
be
r
D
at
e
ID
A
m
ou
nt
Ta
x
To
ta
l
Ty
pe
23
4
11
/2
3/
00
22
42
$
90
.0
0
$5
.8
5
$
95
.8
5
M
C
23
5
11
/2
3/
00
95
00
$
12
.0
0
$0
.6
0
$
12
.6
0
V
IS
A
23
6
11
/2
3/
00
15
56
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
23
7
11
/2
3/
00
22
42
$
75
.0
0
$4
.8
8
$
79
.8
8
A
M
EX
23
8
11
/2
3/
00
22
42
$
60
.0
0
$3
.9
0
$
63
.9
0
M
C
23
9
11
/2
3/
00
10
35
$
90
.0
0
$4
.5
0
$
94
.5
0
A
M
EX
24
0
11
/2
3/
00
95
01
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
24
1
11
/2
3/
00
11
23
$ 1
20
.0
0
$9
.6
0
$ 1
29
.6
0
M
C
24
2
11
/2
4/
00
95
00
$
60
.0
0
$3
.0
0
$
63
.0
0
V
IS
A
24
3
11
/2
4/
00
42
54
$
90
.0
0
$4
.5
0
$
94
.5
0
V
IS
A
24
4
11
/2
4/
00
95
00
$
24
.0
0
$1
.2
0
$
25
.2
0
V
IS
A
24
5
11
/2
4/
00
22
42
$
12
.0
0
$0
.7
8
$
12
.7
8
A
M
EX
24
6
11
/2
4/
00
42
54
$
20
.0
0
$1
.0
0
$
21
.0
0
M
C
24
7
11
/2
4/
00
22
41
$
50
.0
0
$2
.5
0
$
52
.5
0
V
IS
A
24
8
11
/2
4/
00
42
54
$
12
.0
0
$0
.6
0
$
12
.6
0
A
M
EX
24
9
11
/2
4/
00
59
27
$
50
.0
0
$2
.5
0
$
52
.5
0
A
M
EX
25
0
11
/2
4/
00
22
42
$
12
.0
0
$0
.7
8
$
12
.7
8
M
C
25
1
11
/2
4/
00
95
00
$
15
.0
0
$0
.7
5
$
15
.7
5
M
C
25
2
11
/2
4/
00
22
42
$ 1
32
.0
0
$8
.5
8
$ 1
40
.5
8
M
C
25
3
11
/2
4/
00
22
42
$
72
.0
0
$4
.6
8
$
76
.6
8
A
M
EX
O
rd
er
C
us
t I
D
is
a
fo
re
ig
n
ke
y
in
O
rd
er
Pa
ym
en
t t
yp
e
is
a
fo
re
ig
n
ke
y
in
O
rd
er
O
rd
er
N
um
be
r
ID
is
th
e
pr
im
ar
y
ke
y
of
th
e
O
rd
er
ta
bl
e
C
us
t I
D
is
th
e
pr
im
ar
y
ke
y
of
C
us
to
m
er
ta
bl
e.
Pa
ym
en
t t
yp
e
is
th
e
pr
im
ar
y
ke
y
of
th
e
Pa
ym
en
t T
yp
e
ta
bl
e
331
332 Chapter 9 Data Management Layer Design
Tables have a set number of columns and a variable number of rows that contain
occurrences of data. Structured query language (SQL) is the standard language for access-
ing the data in the tables. SQL operates on complete tables, as opposed to the individual
rows in the tables. Th us, a query written in SQL is applied to all the rows in a table all at
once, which is diff erent from a lot of programming languages, which manipulate data row
by row. When queries must include information from more than one table, the tables fi rst
are joined based on their primary key and foreign key relationships and treated as if they
were one large table. Examples of RDBMS soft ware are Microsoft SQL Server, Oracle, DB2,
and MySQL.
To use a RDBMS to store objects, objects must be converted so that they can be stored in
a table. From a design perspective, this entails mapping a UML class diagram to a relational
database schema.
Object-Relational Databases
Object-relational database management systems (ORDBMSs) are relational database manage-
ment systems with extensions to handle the storage of objects in the relational table structure.
Th is is typically done through the use of user-defi ned types. For example, an attribute in a
table could have a data type of map, which would support storing a map. Th is is an example
of a complex data type. In pure RDBMSs, attributes are limited to simple or atomic data types,
such as integers, fl oats, or chars.
ORDBMSs, because they are simply extensions to their RDBMS counterparts, also
have very good support for the typical data management operations that business has come
to expect from RDBMSs, including an easy-to-use query language (SQL), authorization,
concurrency-control, and recovery facilities. However, because SQL was designed to handle
only simple data types, it too has been extended to handle complex object data. Currently,
vendors deal with this issue in diff erent manners. For example, DB2, Informix, and Oracle
all have extensions that provide some level of support for objects.
Many of the ORDBMSs on the market still do not support all of the object-oriented
features that can appear in an object-oriented design (e.g., inheritance). As described in
Chapter 8, one of the problems in supporting inheritance is that inheritance support is
language dependent. For example, the way Smalltalk supports inheritance is different
from C11’s’ approach, which is diff erent from Java’s approach. Th us, vendors currently
must support many different versions of inheritance, one for each object-oriented lan-
guage, or decide on a specifi c version and force developers to map their object-oriented
design (and implementation) to their approach. Like RDBMSs, a mapping from a UML
class diagram to an object-relational database schema is required.
Object-Oriented Databases
Th e next type of database management system that we describe is the object-oriented database
management systems (OODBMS). Th ere have been two primary approaches to supporting
object persistence within the OODBMS community: adding persistence extensions to an
object-oriented programming language and creating an entirely separate database manage-
ment system.
With an OODBMS, collections of objects are associated with an extent. An extent is sim-
ply the set of instances associated with a particular class (i.e., it is the equivalent of a table in a
RDBMS). Technically speaking, each instance of a class has a unique identifi er assigned to it
by the OODBMS: the Object ID. However, from a practical point of view, it is still a good idea
to have a semantically meaningful primary key (even though from an OODBMS perspective
this is unnecessary). Referential integrity is still very important. In an OODBMS, from the
user’s perspective, it looks as if the object is contained within the other object. However, the
Object Persistence Formats 333
OODBMS actually keeps track of these relationships through the use of the Object ID, and
therefore foreign keys are not technically necessary.6
OODBMSs provide support for some form of inheritance. However, as already discussed,
inheritance tends to be language dependent. Currently, most OODBMSs are tied closely
to either a particular object-oriented programming language (OOPL) or a set of OOPLs.
Originally, most OODBMSs supported either Smalltalk or C11. Today, many of the commer-
cially available OODBMSs provide support for C11, Java, and Smalltalk.
OODBMSs also support the idea of repeating groups (fi elds) or multivalued attributes.
Th ese are supported through the use of attribute sets and relationships sets. RDBMSs do not
explicitly allow multivalued attributes or repeating groups. Th is is considered to be a viola-
tion of the fi rst normal form (discussed later in this chapter) for relational databases. Some
ORDBMSs do support repeating groups and multivalued attributes.
Until recently, OODBMSs have mainly been used to support multimedia applications
or systems that involve complex data (e.g., graphics, video, sound). Application areas, such
as computer-aided design and manufacturing (CAD/CAM), fi nancial services, geographic
information systems, health care, telecommunications, and transportation, have been the
most receptive to OODBMSs. Th ey are also becoming popular technologies for supporting
electronic commerce, online catalogs, and large Web multimedia applications. Examples of
pure OODBMSs include Gemstone, Objectivity, db4o, and Versant.
Although pure OODBMS exist, most organizations currently invest in ORDBMS
technology. Th e market for OODBMS is expected to grow, but its ORDBMS and RDBMS
counterparts dwarf it. One reason for this situation is that there are many more experienced
developers and tools in the RDBMS arena. Furthermore, relational users fi nd that using an
OODBMS comes with a fairly steep learning curve.
NoSQL Data Stores7
NoSQL data stores are the newest type of object persistence available. Depending on whom
you talk to, NoSQL either stands for No SQL or Not Only SQL. Regardless, the data stores
that are described as NoSQL typically do not support SQL. Currently, there is no standard
for NoSQL data stores. Most NoSQL data stores were created to address problems associated
with storing large amounts of distributed data in RDBMSs. NoSQL data stores tend to support
very fast queries. However, when it comes to updating, NoSQL data stores normally do not
support a locking mechanism, and consequently, all copies of a piece of data are not required
to be consistent at all times. Instead they tend to support an eventually consistent based
model. So it is technically possible to have diff erent values for diff erent copies of the same
object stored in diff erent locations in a distributed system. Depending on the application, this
could cause problems for decision makers. Th erefore, their applicability is limited and are
not applicable to most traditional business transaction processing systems. Some of the better
known NoSQL data stores include Google’s Big Table, Amazon’s Dynamo, Apache’s HBase,
Apache’s CouchDB, and Apache/Facebook’s Cassandra. Th ere are many diff erent types of
NoSQL data stores, including key-value stores, document stores, column-oriented stores,
and object databases. Besides object databases, which are either ORDBMSs or OODBMSs
(see previous sections), we describe each NoSQL data store type below.
6 Depending on the storage and updating requirements, it usually is a good idea to use a foreign key in addition to
the Object ID. Th e Object ID has no semantic meaning. Th erefore, in the case of needing to rebuild relationships
between objects, Object IDs are diffi cult to validate. Foreign keys, by contrast, should have some meaning outside of
the DBMS.
7 For a more complete description of NoSQL data stores see Pramod J. Sadalage and Martin Fowler, NoSQL Distilled:
A Brief Guide to the Emerging World of Polyglot Persistence (Upper Saddle River, NJ: Addison Wesley, 2013).
334 Chapter 9 Data Management Layer Design
Key-value data stores essentially provide a distributed index (primary key) to where a BLOB
(binary large object) is stored. A BLOB treats a set of attributes as one large object. A good
example of this type of NoSQL data store is Amazon’s Dynamo. Dynamo provides support
for many of the core services for Amazon. Obviously, as one of the largest e-commerce sites
in the world, Amazon needed a solution for object persistence that was scalable, distributable,
and reliable. Typical RDBMS-based solutions would not work for some of these applications.
Applications that typically use key-value data stores are Web-based shopping carts, product
catalogs, and bestseller lists. Th ese types of applications do not require updating the underlying
data. For example, you do not update the title of a book in your shopping cart when you are
making a purchase at Amazon. Given the scale and distributed nature of this type of system,
there are bound to be many failures across the system. Being fault tolerant and temporarily sac-
rifi cing some consistency across all copies of an object is a reasonable trade-off .
Document data stores, as the name suggests, are built around the idea of documents. Th e
idea of document databases has been around for a long time. One of the early systems that
used this approach was Lotus Notes. Th ese types of stores are considered to be schema free.
By that we mean there is no detailed design of the database. A good example of an applica-
tion that would benefi t from this type of approach is a business card database. In a relational
database, multiple tables would need to be designed. In a document data store, the design
is done more in a “just in time” manner. As new business cards are input into the system,
attributes not previously included are simply added to the evolving design. Previously entered
business cards would simply not have those attributes associated with them. One major dif-
ference between key-value data stores and document data stores is that the “document” has
structure and can be easily searched based on the non-key attributes contained in the docu-
ment, whereas the key-value data store simply treats the “value” as one big monolithic object.
Apache’s CouchDB is a good example of this type of data store.
Columnar data stores organize the data into columns instead of rows. However,
there seems to be some confusion as to what this actually implies. In the fi rst approach
to columnar data stores, the rows represent the attributes and the columns represent the
objects. In contrast, relational databases represent attributes in columns and represent
objects in rows. Th is type of columnar data store is very eff ective in business intelligence,
data mining, and data warehousing applications where the data are fairly static and many
computations are performed over a single or a small subset of the available attributes. In
comparison to a relational database where you would have to select a set of attributes from
all rows, with this type of data store, you would simply have to select a set of rows. Th is
should be a lot faster than with a relational database. A few good examples of this type of
columnar data store include Oracle’s Retail Predictive Application Server, HP’s Vertica,
and SAP’s Sybase IQ. Th e second approach to columnar data stores, which includes
Apache’s HBase, Apache/Facebook’s Cassandra, and Google’s BigTable, is designed to
handle very large data sets (petabytes of data) that can be accessed as if the data are stored
in columns. However, in this case, the data are actually stored in a three-dimensional map
composed of object ID, attribute name, timestamp, and value instead of using columns
and rows. Th is approach is highly scalable and distributable. Th ese types of data stores
support social applications such as Twitter and Facebook and support search applications
such as Google Maps, Earth, and Analytics.
Given the popularity of social computing, business intelligence, data mining, data ware-
housing, e-commerce, and their need for highly scalable, distributable, and reliable data stor-
age, NoSQL data stores are an area that should be considered as part of an object persistence
solution. However, given the overall diversity and complexity of NoSQL data stores and their
limited applicability to traditional business applications, we do not consider them any further
in this text.
Object Persistence Formats 335
Selecting an Object Persistence Format
Each of the fi le and database storage formats that have been presented has its strengths and
weaknesses, and no one format is inherently better than the others. In fact, sometimes a pro-
ject team chooses multiple formats (e.g., a relational database for one, a fi le for another, and
an object-oriented database for a third). Th us, it is important to understand the strengths and
weaknesses of each format and when to use each one. Figure 9-4 presents a summary of the
characteristics of each and the characteristics that can help identify when each type of format
is more appropriate.
Major Strengths and Weaknesses Th e major strengths of fi les include the following: Some
support for sequential and random access fi les is normally part of an OOPL, fi les can be
designed to be very effi cient, and they are a good alternative for temporary or short-term
storage. However, all fi le manipulation must be done through the OOPL. Files do not have
any form of access control beyond that of the underlying operating system. Finally, in most
cases, if fi les are used for permanent storage, redundant data most likely will result. Th is can
cause many update anomalies.
FIGURE 9-4 Comparison of Object Persistence Formats
Sequential and Object Relational Object-Oriented
Random Access Files Relational DBMS DBMS DBMS NoSQL data store
Major
Strengths
Major
Weaknesses
Data Types
Supported
Types of
Application
Systems
Supported
Existing
Storage
Formats
Future
Needs
Usually part of an
object-oriented
programming
language
Files can be
designed for fast
performance
Good for short-term
data storage
Redundant data
Data must be updated
using programs, i.e.,
no manipulation or
query language
No access control
Simple and Complex
Transaction processing
Organization
dependent
Poor future
prospects
Leader in the
database market
Can handle diverse
data needs
Cannot handle
complex data
No support for
object orientation
Impedance mismatch
between tables and
objects
Simple
Transaction
processing and
decision making
Organization
dependent
Good future
prospects
Able to handle
complex data
Direct support for
object orientation
Technology is still
maturing
Skills are hard to fi nd
Simple and Complex
Transaction
processing and
decision making
Organization
dependent
Good future
prospects
Able to handle
complex data
Technology is still
maturing
Skills are hard
to fi nd
Simple and Complex
Primarily decision
making
Organization
dependent
Good future
prospects
Based on established,
proven technology,
e.g., SQL
Able to handle
complex data
Limited support for
object orientation
Impedance mismatch
between tables and
objects
Simple and Complex
Transaction processing
and decision making
Organization
dependent
Good future
prospects
336 Chapter 9 Data Management Layer Design
RDBMSs bring with them proven commercial technology. They are the leaders in the
DBMS market. Furthermore, they can handle very diverse data needs. However, they can-
not handle complex data types, such as images. Therefore, all objects must be converted
to a form that can be stored in tables composed of atomic or simple data. They provide
no support for object orientation. This lack of support causes an impedance mismatch
between the objects contained in the OOPL and the data stored in the tables. An imped-
ance mismatch refers to the amount of work done by both the developer and DBMS and
the potential information loss that can occur when converting objects to a form that can
be stored in tables.
Because ORDBMSs are typically object-oriented extensions to RDBMSs, they inherit the
strengths of RDBMSs. Th ey are based on established technologies, such as SQL, and unlike
their predecessors, they can handle complex data types. However, they provide only limited
support for object orientation. Th e level of support varies among the vendors; therefore,
ORDBMSs also suff er from the impedance mismatch problem.
OODBMSs support complex data types and have the advantage of directly support-
ing object orientation. Th erefore, they do not suff er from the impedance mismatch that
the previous DBMSs do. However, the OODBMS community is still maturing. Th erefore,
this technology might still be too risky for some fi rms. Th e other major problems with
OODBMS are the lack of skilled labor and the perceived steep learning curve of the RDBMS
community. NoSQL data stores support complex data types. However, they can suff er from
some forms of impedance mismatch. Th e primary problems with NoSQL data stores are
their lack of maturity and the lack of skilled labor who knows how to eff ectively use them.
Data Types Supported Th e fi rst issue is the type of data that will need to be stored in the
system. Most applications need to store simple data types, such as text, dates, and numbers.
All fi les and DBMSs are equipped to handle this kind of data. Th e best choice for simple
data storage, however, is usually the RDBMS because the technology has matured over time
and has continuously improved to handle simple data very eff ectively.
Increasingly, applications are incorporating complex data, such as video, images, or audio.
ORDBMSs, OODBMSs, or NoSQL data stores are best able to handle data of this type. Complex
data stored as objects can be manipulated much faster than with other storage formats.
Type of Application System Th ere are many diff erent kinds of application systems that
can be developed. Transaction-processing systems are designed to accept and process many
simultaneous requests (e.g., order entry, distribution, payroll). In transaction-processing
systems, the data are continuously updated by a large number of users, and the queries that
these systems require typically are predefi ned or targeted at a small subset of records (e.g.,
List the orders that were backordered today or What products did customer #1234 order on
May 12, 2001?).
Another set of application systems is the set designed to support decision making, such
as decision support systems (DSS), management information systems (MIS), executive infor-
mation systems (EIS), and expert systems (ES). Th ese decision-making support systems are
built to support users who need to examine large amounts of read-only historical data. Th e
questions that they ask are oft en ad hoc, and include hundreds or thousands of records at a
time (e.g., List all customers in the West region who purchased a product costing more than
$500 at least three times, or What products had increased sales in the summer months that
have not been classifi ed as summer merchandise?).
Transaction-processing systems and DSSs thus have very diff erent data storage needs.
Transaction-processing systems need data storage formats that are tuned for a lot of data
updates and fast retrieval of predefi ned, specifi c questions. Files, relational databases,
Mapping Problem Domain Objects to Object Persistence Formats 337
object-relational databases, and object-oriented databases can all support these kinds of
requirements. By contrast, systems to support decision making are usually only reading
data (not updating it), oft en in ad hoc ways. Th e best choices for these systems usually are
RDBMSs because these formats can be confi gured specially for needs that may be unclear
and less apt to change the data. However, depending on the type of data needed to support
the decision-making application, RDBMSs may not be appropriate. In that case, ORDBMS,
OODBMS, or a NoSQL data store may be the better solution.
Existing Storage Formats The storage format should be selected primarily on the basis
of the kind of data and application system being developed. However, project teams
should consider the existing storage formats in the organization when making design
decisions. In this way, they can better understand the technical skills that already exist
and how steep the learning curve will be when the storage format is adopted. For exam-
ple, a company that is familiar with RDBMS will have little problem adopting a relational
database for the project, whereas an OODBMS or a NoSQL data store might require
substantial developer training.
Future Needs Not only should a project team consider the storage technology within the
company, but it should also be aware of current trends and technologies that are being used by
other organizations. A large number of installations of a specifi c type of storage format suggest
that skills and products are available to support the format. Th erefore, the selection of that for-
mat is safe. For example, it would probably be easier and less expensive to fi nd RDBMS exper-
tise when implementing a system than to fi nd help with an OODBMS or a NoSQL data store.
Other Miscellaneous Criteria Other criteria that should be considered include cost,
licensing issues, concurrency control, ease of use, security and access controls, version
management, storage management, lock management, query management, language
bindings, and APIs. We also should consider performance issues, such as cache manage-
ment, insertion, deletion, retrieval, and updating of complex objects. Finally, the level of
support for object orientation (such as objects, single inheritance, multiple inheritance,
polymorphism, encapsulation and information hiding, methods, multivalued attributes,
repeating groups) is critical.
MAPPING PROBLEM DOMAIN OBJECTS TO OBJECT
PERSISTENCE FORMATS
Th ere are many diff erent formats from which to choose to support object persistence. Each
of the formats can have some conversion requirements. Regardless of the object persistence
format chosen, we suggest supporting primary keys and foreign keys by adding them to the
problem domain classes at this point. However, this does imply that some additional pro-
cessing will be required. Th e developer has to set the value for the foreign key when adding
the relationship to an object. From a practical perspective, fi le formats are used mostly for
temporary storage. Th us, we do not consider them further.
We also recommend that data management functionality specifi cs, such as retrieval and
updating of data from the object storage, be included only in classes contained in the data man-
agement layer. Th is will ensure that the data management classes are dependent on the problem
domain classes and not vice versa. Th is allows the design of problem domain classes to be inde-
pendent of any specifi c object persistence environment, thus increasing their portability and their
potential for reuse. Like our previous recommendation, this also implies additional processing.
338 Chapter 9 Data Management Layer Design
Mapping Problem Domain Objects to an OODBMS Format
If we support object persistence with an OODBMS, the mappings between the problem domain
objects and the OODBMS tend to be fairly straightforward. As a starting point, we suggest that
each concrete problem domain class should have a corresponding object persistence class in
the OODBMS. Th ere will also be a data access and manipulation (DAM) class (described later
in this chapter) that contains the functionality required to manage the interaction between the
object persistence class and the problem domain layer. For example, using the appointment
system example from the previous chapters, the Patient class is associated with an OODBMS
class (see Figure 9-5). Th e Patient class essentially will be unchanged from analysis. Th e
Patient-OODBMS class will be a new class that is dependent on the Patient class, whereas the
Patient-DAM class will be a new class that depends on both the Patient class and the Patient-
OODBMS class. Th e Patient-DAM class must be able to read from and write to the OODBMS.
Otherwise, it will not be able to store and retrieve instances of the Patient class. Even though
this does add overhead to the installation of the system, it allows the problem domain class
to be independent of the OODBMS being used. If at a later time another OODBMS or object
persistence format is adopted, only the DAM classes will have to be modifi ed. Th is approach
increases both the portability and the potential for reuse of the problem domain classes.
Even though we are implementing the DAM layer using an OODBMS, a mapping from
the problem domain layer to the OODBMS classes in the data access and management layer
may be required. If multiple inheritance is used in the problem domain but not supported by
the OODBMS, then the multiple inheritance must be factored out of the OODBMS classes.
For each case of multiple inheritance (i.e., more than one superclass), the following rules can
be used to factor out the multiple inheritance eff ects in the design of the OODBMS classes.8
PD Layer
DM Layer
AppointmentPatient
Appointment-OODBMSPatient-OODBMS
Appointment-DAMPatient-DAM
FIGURE 9-5
Appointment
System Problem-
Domain and DM
Layers
8 Th e rules presented in this section are based on material in Ali Bahrami, Object-Oriented Systems Development Using
the Unifi ed Modeling Language (New York: McGraw-Hill, 1999); Michael Blaha and William Premerlani, Object-
Oriented Modeling and Design for Database Applications (Upper Saddle River, NJ: Prentice Hall, 1998); Akmal B.
Chaudri and Roberto Zicari, Succeeding with Object Databases: A Practical Look at Today’s Implementations with Java
and XML (New York: Wiley, 2001); Peter Coad and Edward Yourdon, Object-Oriented Design (Upper Saddle River, NJ:
Yourdon Press, 1991); Paul R. Read, Jr., Developing Applications with Java and UML (Boston: Addison-Wesley, 2002).
Mapping Problem Domain Objects to Object Persistence Formats 339
Rule 1a: Add a column(s) to the OODBMS class(es) that represents the subclass(es)
that will contain an Object ID of the instance stored in the OODBMS class that
represents the “additional” superclass(es). Th is is similar in concept to a foreign
key in an RDBMS. Th e multiplicity of this new association from the subclass
to the “superclass” should be 1..1. Add a column(s) to the OODBMS class(es)
that represents the superclass(es) that will contain an Object ID of the instance
stored in the OODBMS class that represents the subclass(es). If the superclasses
are concrete, that is, they can be instantiated themselves, then the multiplicity
from the superclass to the subclass is 0..1, otherwise, it is 1..1. An exclusive-or
(XOR) constraint must be added between the associations. Do this for each
“additional” superclass.
or
Rule 1b: Flatten the inheritance hierarchy of the OODBMS classes by copying the
attributes and methods of the additional OODBMS superclass(es) down to
all of the OODBMS subclasses and remove the additional superclass from the
design.9
Th ese multiple inheritance rules are very similar to those described in Chapter 8.
Figure 9-6 demonstrates the application of these rules. Th e right side of the fi gure portrays
the same problem domain classes that were in Chapter 8: Airplane, Car, Boat, FlyingCar,
and AmphibiousCar. FlyingCar inherits from both Airplane and Car, and AmphibiousCar
inherits from both Car and Boat. Figure 9-6a portrays the mapping of multiple inheritance
relationships into a single inheritance-based OODBMS using Rule 1a. Assuming that Car is
concrete, we apply Rule 1a to the Problem Domain classes, and we end up with the OODBMS
classes on the left side of Part a, where we have:
■ Added a column (attribute) to FlyingCar-OODBMS that represents an association
with Car-OODBMS;
■ Added a column (attribute) to AmphibiousCar-OODBMS that represents an
association with Car-OODBMS;
■ Added a pair of columns (attributes) to Car-OODBMS that represents an
association with FlyingCar-OODBMS and AmphibiousCar-OODBMS and for
completeness sake;
■ Added associations between AmphibiousCar-OODBMS and Car-OODBMS
and FlyingCar-OODBMS and Car-OODBMS that have the correct multiplicities
and the XOR constraint explicitly shown.
We also display the dependency relationships from the OODBMS classes to the problem
domain classes. Furthermore, we illustrate the fact that the association between FlyingCar-
OODBMS and Car-OODBMS and the association between AmphibiousCar-OODBMS and
Car-OODBMS are based on the original factored-out inheritance relationships in the problem
domain classes by showing dependency relationships from the associations to the inheritance
relationships.
On the other hand, if we apply Rule 1b to map the Problem Domain classes to a single-
inheritance-based OODBMS, we end up with the mapping in Figure 9-6b, where all the attrib-
utes of Car have been copied into the FlyingCar-OODBMS and AmphibiousCar-OODBMS
classes. In this latter case, you may have to deal with the eff ects of inheritance confl icts (see
Chapter 8).
9 It is also a good idea to document this modifi cation in the design so that in the future, modifi cations to the design
can be easily maintained.
B
oa
t-
O
O
D
B
M
S
-w
ei
gh
t
-l
en
gt
h
O
O
D
B
M
S
C
la
ss
es
Pr
ob
le
m
D
om
ai
n
C
la
ss
es
A
m
ph
ib
io
us
C
ar
-O
O
D
B
M
S
-m
fg
-y
r
-C
ar
-O
O
D
B
M
S
B
oa
t-
O
O
D
B
M
S
-w
ei
gh
t
-l
en
gt
h
Fl
yi
ng
C
ar
-O
O
D
B
M
S
-m
fg
-y
r
-C
ar
-O
O
D
B
M
S
C
ar
-O
O
D
B
M
S
-n
um
be
ro
fD
oo
rs
-r
eg
N
o
-F
ly
in
gC
ar
-O
O
D
B
M
S
-A
m
ph
ib
io
us
C
ar
-O
O
D
B
M
S
A
m
ph
ib
io
us
C
ar
-O
O
D
B
M
S
-n
um
be
rO
fD
oo
rs
-r
eg
N
o
-m
fg
-y
r
A
ir
pl
an
e-
O
O
D
B
M
S
-e
ng
in
eT
yp
e
-f
ue
lT
yp
e
Fl
yi
ng
C
ar
-O
O
D
B
M
S
-n
um
be
rO
fD
oo
rs
-r
eg
N
o
-m
fg
-y
r
0.
.*
1.
.1
1.
.1
0.
.*
(a
)
(b
)
A
ir
pl
an
e-
O
O
D
B
M
S
-E
ng
in
eT
yp
e
-F
ue
lT
yp
e
A
ir
pl
an
e
-E
ng
in
eT
yp
e
-F
ue
lT
yp
e
{X
O
R
}
Fl
yi
ng
C
ar
-m
fg
-y
r
A
ir
pl
an
e
-e
ng
in
eT
yp
e
-f
ue
lT
yp
e
Fl
yi
ng
C
ar
-m
fg
-y
r
A
m
ph
ib
io
us
C
ar
-m
fg
-y
r
B
oa
t
-w
ei
gh
t
-l
en
gt
h
C
ar
-n
um
be
rO
fD
oo
rs
-r
eg
N
o
A
m
ph
ib
io
us
C
ar
-m
fg
-y
r
C
ar
-n
um
be
rO
fD
oo
rs
-r
eg
N
o
B
oa
t
-w
ei
gh
t
-l
en
gt
h
FI
G
U
R
E
9
-6
M
ap
p
in
g
Pr
o
b
le
m
D
o
m
ai
n
O
b
je
ct
s
to
a
S
in
gl
e
In
h
er
it
an
ce
-B
as
ed
O
O
D
B
M
S
340
Mapping Problem Domain Objects to Object Persistence Formats 341
Th e advantage of Rule 1a is that all problem domain classes identifi ed during analysis are
preserved in the database. Th is allows maximum fl exibility of maintenance of the design of the
data management layer. However, Rule 1a increases the amount of message passing required
in the system, and it has added processing requirements involving the XOR constraint, thus
reducing the overall effi ciency of the design. Our recommendation is to limit Rule 1a to be
applied only when dealing with “extra” superclasses that are concrete because they have an
independent existence in the problem domain. Use Rule 1b when they are abstract because
they do not have an independent existence from the subclass.
In either case, additional processing will be required. In the fi rst case, cascading of deletes
will work, not only from the individual object to all its elements but also from the superclass
instances to all the subclass instances. In the second case, there will be a lot of copying and
pasting of the structure of the superclass to the subclasses. In the case that a modifi cation of
the structure of the superclass is required, the modifi cation must be cascaded to all of the sub-
classes. However, multiple inheritance is rare in most business problems. In most situations,
the preceding rules will never be necessary.
When instantiating problem domain objects from OODBMS objects, additional process-
ing will also be required. Th e additional processing will be in the retrieval of the OODBMS
objects and taking their elements to create a problem domain object. Also, when storing the
problem domain object, the conversion to a set of OODBMS objects is required. Basically
speaking, any time that an interaction takes place between the OODBMS and the system, if
multiple inheritance is involved and the OODBMS supports only single inheritance, a con-
version between the two formats will be required.
Mapping Problem Domain Objects to an ORDBMS Format
If we support object persistence with an ORDBMS, then the mapping from the problem
domain objects to the data management objects is much more involved. Depending on
the level of support for object orientation, diff erent mapping rules are necessary. For our
purposes, we assume that the ORDBMS supports Object IDs, multivalued attributes, and
stored procedures. However, we assume that the ORDBMS does not provide any support
for inheritance. Based on these assumptions, Figure 9-7 lists a set of rules that can be used to
design the mapping from the Problem Domain objects to the tables of the ORDBMS-based
data management layer.
First, all concrete Problem Domain classes must be mapped to the tables in the ORDBMS.
For example in Figure 9-8, the Patient class has been mapped to Patient-ORDBMS table.
Notice that the Participant class has also been mapped to an ORDBMS table. Even though
the Participant class is abstract, this mapping was done because in the complete class diagram
(see Figure 7-15), the Participant class had multiple direct subclasses (Employee and Patient).
Second, single-valued attributes should be mapped to columns in the ORDBMS tables.
Again, referring to Figure 9-8, we see that the amount attribute of the Patient class has been
included in the Patient Table class.
Th ird, depending on the level of support of stored procedures, the methods and derived
attributes should be mapped either to stored procedures or program modules.
Fourth, single-valued (one-to-one) aggregation and association relationships should be
mapped to a column that can store an Object ID. Th is should be done for both sides of the
relationship.
Fift h, multivalued attributes should be mapped to columns that can contain a set of
values. For example in Figure 9-8, the insurance carrier attribute in the Patient class may
contain multiple values because a patient may have more than one insurance carrier. Th us
in the Patient table, a multiplicity has been added to the insurance carrier attribute to por-
tray this fact.
342 Chapter 9 Data Management Layer Design
Rule 1: Map all concrete Problem Domain classes to the ORDBMS tables. Also, if an abstract problem domain class has multiple
direct subclasses, map the abstract class to an ORDBMS table.
Rule 2: Map single-valued attributes to columns of the ORDBMS tables.
Rule 3: Map methods and derived attributes to stored procedures or to program modules.
Rule 4: Map single-valued aggregation and association relationships to a column that can store an Object ID. Do this for both sides
of the relationship.
Rule 5: Map multivalued attributes to a column that can contain a set of values.
Rule 6: Map repeating groups of attributes to a new table and create a one-to-many association from the original table to the
new one.
Rule 7: Map multivalued aggregation and association relationships to a column that can store a set of Object IDs. Do this for both
sides of the relationship.
Rule 8: For aggregation and association relationships of mixed type (one-to-many or many-to-one), on the single-valued side (1..1
or 0..1) of the relationship, add a column that can store a set of Object IDs. The values contained in this new column will be the
Object IDs from the instances of the class on the multivalued side. On the multivalued side (1..* or 0..*), add a column that can
store a single Object ID that will contain the value of the instance of the class on the single-valued side.
For generalization/inheritance relationships:
Rule 9a: Add a column(s) to the table(s) that represents the subclass(es) that will contain an Object ID of the instance stored in the
table that represents the superclass. This is similar in concept to a foreign key in an RDBMS. The multiplicity of this new associa-
tion from the subclass to the “superclass” should be 1..1. Add a column(s) to the table(s) that represents the superclass(es) that will
contain an Object ID of the instance stored in the table that represents the subclass(es). If the superclasses are concrete, that is, they
can be instantiated themselves, then the multiplicity from the superclass to the subclass is 0..1, otherwise, it is 1..1. An exclusive-or
(XOR) constraint must be added between the associations. Do this for each superclass.
or
Rule 9b: Flatten the inheritance hierarchy by copying the superclass attributes down to all of the subclasses and remove the super-
class from the design.*
*It is also a good idea to document this modifi cation in the design so that in the future, modifi cations to the design can be maintained easily.
FIGURE 9-7 Schema for Mapping Problem Domain Objects to ORDBMS
Th e sixth mapping rule addresses repeating groups of attributes in a problem domain
object. In this case, the repeating group of attributes should be used to create a new table in
the ORDBMS. It can imply a missing class in the problem domain layer. Normally, when
a set of attributes repeats together as a group, it implies a new class. Finally, we must create a
one-to-many association from the original table to the new one.
Th e seventh rule supports mapping multivalued (many-to-many) aggregation and
association relationships to columns that can store a set of Object IDs. Basically, this is a
combination of the fourth and fi ft h rules. Like the fourth rule, this should be done for both
sides of the relationships. For example in Figure 9-8, the Symptom table has a multivalued
attribute (Patients) that can contain multiple Object IDs to Patient Table objects, and
Patient table has a multivalued attribute (Symptoms) that can contain multiple Object IDs
to Symptom Table objects.
Th e eighth rule combines the intentions of Rules 4 and 7. In this case, the rule maps
one-to-many and many-to-one relationships. On the single-valued side (1..1 or 0..1) of the
relationship, a column that can store a set of Object IDs from the table on the multivalued side
(1..* or 0..*) of the relationship should be added. On the multivalued side, a column should
be added to the table that can store an Object ID from an instance stored in the table on the
single-valued side of the relationship. For example, in Figure 9-8, the Patient table has a mul-
tivalued attribute (Appts) that can contain multiple Object IDs to Appointment Table objects,
whereas the Appointment table has a single-valued attribute (Patient) that can contain an
Object ID to a Patient Table object.
Th e ninth, and fi nal, rule deals with the lack of support for generalization and inher-
itance. In this case, there are two diff erent approaches. Th ese approaches are virtually
Mapping Problem Domain Objects to Object Persistence Formats 343
ORDBMS Tables Problem Domain Classes
Participant Table
-lastname[1..1]
-firstname[1..1]
-address[1..1]
-phone[1..1]
-birthdate[1..1]
-SubClassObjects[1..1]
Patient Table
-amount[1..1]
-Participant[1..1]
-Appts[0..*]
-Symptoms[1..*]
-Insurance carrier[0..*]
-Primary Insurance Carrier[0..*]
Symptom Table
-name[1..1]
-Patients[0..*]
Appointment Table
-Patient[1..1]
-time[1..1]
-date[1..1]
-reason[1..1]
1..1
1..1
0..*
0..*
0..* 0..*
0..*
1
1
1..* 1..*
0..*
1
1
suffers
schedules
Appointment
-time
-date
-reason
+cancel without notice()
+ primary
insurance
carrier
Patient
-amount
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
Participant
-lastname
-firstname
-address
-phone
-birthdate
-/ age
Symptom
-name
FIGURE 9-8 Example of Mapping Problem Domain Objects to ORDBMS Schema
identical to the rules described with the preceding OODBMS object persistence formats. For
example in Figure 9-8, the Patient table contains an attribute (Participant) that can contain
an Object ID for a Participant Table object, and the Participant table contains an attribute
(SubClassObjects) that contains an Object ID for an object, in this case, stored in the Patient
table. In the other case, the inheritance hierarchy is fl attened.
Of course, additional processing is required any time an interaction takes place between
the database and the system. Every time an object must be created or retrieved from the
344 Chapter 9 Data Management Layer Design
database, updated, or deleted, the ORDBMS object(s) must be converted to the problem
domain object, or vice versa. Th e only other choice is to modify the problem domain
objects. However, such a modifi cation can cause problems between the problem domain
layer and the physical architecture and human–computer interface layers. Generally speak-
ing, the cost of conversion between the ORDBMS and the problem domain layer should
be more than off set by the savings in development time associated with the interaction
between the problem domain and physical architecture and human–computer interaction
layers and the ease of maintenance of a semantically clean problem domain layer.
Mapping Problem Domain Objects to a RDBMS Format
If we support object persistence with an RDBMS, then the mapping from the problem
domain objects to the RDBMS tables is similar to the mapping to an ORDBMS. However, the
assumptions made for an ORDBMS are no longer valid. Figure 9-9 lists a set of rules that can
be used to design the mapping from the problem domain objects to the RDBMS-based data
management layer tables.
Th e fi rst four rules are basically the same set of rules used to map problem domain
objects to ORDBMS-based data management objects. First, all concrete problem domain
classes must be mapped to tables in the RDBMS. Second, single-valued attributes should be
mapped to columns in the RDBMS table. Th ird, methods should be mapped to either stored
procedures or program modules, depending on the complexity of the method. Fourth, sin-
gle-valued (one-to-one) aggregation and association relationships are mapped to columns
that can store the foreign keys of the related tables. Th is should be done for both sides of the
relationship. For example in Figure 9-10, we needed to include tables in the RDBMS for the
Participant, Patient, Symptom, and Appointment classes.
Rule 1: Map all concrete-problem domain classes to the RDBMS tables. Also, if an abstract Problem Domain class has multiple
direct subclasses, map the abstract class to a RDBMS table.
Rule 2: Map single-valued attributes to columns of the tables.
Rule 3: Map methods to stored procedures or to program modules.
Rule 4: Map single-valued aggregation and association relationships to a column that can store the key of the related table, i.e., add
a foreign key to the table. Do this for both sides of the relationship.
Rule 5: Map multivalued attributes and repeating groups to new tables and create a one-to-many association from the original table
to the new ones.
Rule 6: Map multivalued aggregation and association relationships to a new associative table that relates the two original tables
together. Copy the primary key from both original tables to the new associative table, i.e., add foreign keys to the table.
Rule 7: For aggregation and association relationships of mixed type, copy the primary key from the single-valued side (1..1 or 0..1)
of the relationship to a new column in the table on the multivalued side (1..* or 0..*) of the relationship that can store the key of the
related table, i.e., add a foreign key to the table on the multivalued side of the relationship.
For generalization/inheritance relationships:
Rule 8a: Ensure that the primary key of the subclass instance is the same as the primary key of the superclass. The multiplicity of this
new association from the subclass to the “superclass” should be 1..1. If the superclasses are concrete, that is, they can be instanti-
ated themselves, then the multiplicity from the superclass to the subclass is 0..1, otherwise, it is 1..1. Furthermore, an
exclusive-or (XOR) constraint must be added between the associations. Do this for each superclass.
or
Rule 8b: Flatten the inheritance hierarchy by copying the superclass attributes down to all of the subclasses and remove the
superclass from the design.*
* It is also a good idea to document this modifi cation in the design so that in the future, modifi cations to the design can be
maintained easily.
FIGURE 9-9 Schema for Mapping Problem Domain Objects to RDBMS
Mapping Problem Domain Objects to Object Persistence Formats 345
RDBMS Tables Problem Domain Classes
Participant Table
-lastname[1..1]
-firstname[1..1]
-address[1..1]
-phone[1..1]
-birthdate[1..1]
-participantNumber[1..1]
Patient Table
-amount[1..1]
-participantNumber[1..1]
-primaryInsuranceCarrier[0..1]
Symptom Table
-name[1..1]
Suffer Table
-participantNumber[1..1]
-name[1..1]
Insurance Carrier Table
-name[1..1]
-participantNumber[1..1]
Appointment Table
-time[1..1]
-date[1..1]
-reason[1..1]
-participantNumber[1..1]
1..1
1..1
1..1
1..1
1..1
1..1
0..*
0..* 0..*
0..*
1..*
0..*
1
1
1..*
0..*
+ primary
insurance
carrier
suffers
schedules
Appointment
-time
-date
-reason
+cancel without notice()Patient
-amount
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
Participant
-lastname
-firstname
-address
-phone
-birthdate
-/ age
Symptom
-name
FIGURE 9-10 Example of Mapping Problem Domain Objects to RDBMS Schema
346 Chapter 9 Data Management Layer Design
Th e fi fth rule addresses multivalued attributes and repeating groups of attributes in a
problem domain object. In these cases, the attributes should be used to create new tables
in the RDBMS. As in the ORDBMS mappings, repeating groups of attributes can imply
missing classes in the Problem Domain layer. In that case, a new problem domain class
may be required. Finally, we should create a one-to-many or zero-to-many association
from the original table to the new one. For example, in Figure 9-10, we needed to create a
new table for insurance carrier because it was possible for a patient to have more than one
insurance carrier.
Th e sixth rule supports mapping multivalued (many-to-many) aggregation and asso-
ciation relationships to a new table that relates the two original tables. In this case, the new
table should contain foreign keys back to the original tables. For example, in Figure 9-10, we
needed to create a new table that represents the suff er association between the Patient and
Symptom problem domain classes.
Th e seventh rule addresses one-to-many and many-to-one relationships. With these
types of relationships, the multivalued side (0..* or 1..*) should be mapped to a column
in its table that can store a foreign key back to the single-valued side (0..1 or 1..1). It is
possible that we have already taken care of this situation because we earlier recommended
inclusion of both primary and foreign key attributes in the problem domain classes. In the
case of Figure 9-10, we had already added the primary key from the Patient class to the
Appointment class as a foreign key (see participantNumber). However, in the case of the
refl exive relationship, primary insurance carrier, associated with the Patient class, we need
to add a new attribute (primaryInsuranceCarrier) to be able to store the relationship.
Th e eighth, and fi nal, rule deals with the lack of support for generalization and inher-
itance. As in the case of an ORDBMS, there are two diff erent approaches. Th ese approaches
are virtually identical to the rules described with OODBMS and ORDBMS object persistence
formats given earlier. Th e fi rst approach is to add a column to each table that represents a
subclass for each of the concrete superclasses of the subclass. Essentially, this ensures that
the primary key of the subclass is the same as the primary key for the superclass. If we had
previously added the primary and foreign keys to the problem domain objects, as we recom-
mended, then we do not have to do anything else. Th e primary keys of the tables will be used
to rejoin the instances stored in the tables that represent each of the pieces of the problem
domain object. Conversely, the inheritance hierarchy can be fl attened and the rules (Rules 1
through 7) can be reapplied.
As in the case of the ORDBMS approach, additional processing will be required any time
that an interaction takes place between the database and the system. Every time an object
must be created, retrieved from the database, updated, or deleted, the mapping between the
problem domain and the RDBMS must be used to convert between the two diff erent formats.
In this case, a great deal of additional processing will be required.
OPTIMIZING RDBMS-BASED OBJECT STORAGE
Once the object persistence format is selected, the second step is to optimize the object per-
sistence for processing effi ciency. Th e methods of optimization vary based on the format that
you select; however, the basic concepts remain the same. Once you understand how to opti-
mize a particular type of object persistence, you will have some idea as to how to approach the
optimization of other formats. Th is section focuses on the optimization of the most popular
storage format: relational databases.
Th ere are two primary dimensions in which to optimize a relational database: for storage
effi ciency and for speed of access. Unfortunately, these two goals oft en confl ict because the
Optimizing RDBMS-Based Object Storage 347
best design for access speed may take up a great deal of storage space as compared to other,
less-speedy designs. Ultimately, the project team will go through a series of trade-off s until the
ideal balance is reached.
Optimizing Storage Effi ciency
Th e most effi cient tables in a relational database in terms of storage space have no redundant
data and very few null values. Th e presence of null values suggests that space is being wasted
(and more data to store means higher data storage hardware costs). For example, the table
in Figure 9-11 repeats customer information, such as name and state, each time a customer
places an order, and it contains many null values in the product-related columns. Th ese nulls
occur whenever a customer places an order for fewer than three items (the maximum number
on an order).
In addition to wasting space, redundancy and null values also allow more room for error
and increase the likelihood that problems will arise with the integrity of the data. What if
customer 1035 moved from Maryland to Georgia? In the case of Figure 9-11, a program must
be written to ensure that all instances of that customer are updated to show Georgia as the
new state of residence. If some of the instances are overlooked, then the table will contain an
update anomaly, whereby some of the records contain the correctly updated value for state
and other records contain the old information.
Nulls threaten data integrity because they are diffi cult to interpret. A blank value in the
Order table’s product fi elds could mean the customer did not want more than one or two
products on his or her order, the operator forgot to enter in all three products on the order,
or the customer canceled part of the order and the products were deleted by the operator. It
is impossible to be sure of the actual meaning of the nulls.
For both these reasons—wasted storage space and data integrity threats—project teams
should remove redundancy and nulls from the table. During design, the class diagram is
used to examine the design of the RDBMS tables (e.g., see Figure 9-10) and to optimize it
for storage effi ciency. If you follow the modeling instructions and guidelines that were pre-
sented in Chapter 5, you will have little trouble creating a design that is highly optimized in
this way because a well-formed logical data model does not contain redundancy or many
null values.
Sometimes, however, a project team needs to start with a model that was poorly con-
structed or with one that was created for fi les or a nonrelational type of format. In these cases,
the project team should follow a series of steps that serve to check the model for storage
effi ciency. Th ese steps make up a process called normalization.10 Normalization is a process
whereby a series of rules are applied to the RDBMS tables to assess the effi ciency of the tables
(see Figure 9-12). Th ese rules help analysts identify tables that are not represented correctly.
Here, we describe three normalization rules that are applied regularly in practice. Figure 9-11
shows a model in 0 Normal Form, which is an unnormalized model before the normalization
rules have been applied.
A model is in fi rst normal form (1NF) if it does not lead to multivalued fi elds, fi elds
that allow a set of values to be stored, or repeating fi elds, which are fi elds that repeat within
a table to capture multiple values. Th e rule for 1NF says that all tables must contain the
same number of columns (i.e., fi elds) and that all the columns must contain a single value.
Notice that the model in Figure 9-11 violates 1NF because it causes product number,
10 Normalization also can be performed on the problem domain layer (see Chapter 8). However, the normalization
process should be used on the problem domain layer only to uncover missing classes. Otherwise, optimizations that
have nothing to do with the semantics of the problem domain can creep into the problem domain layer.
R
ed
un
da
nt
D
at
a
N
ul
l C
el
ls
Sa
m
pl
e
R
ec
or
ds
:
O
rd
er
-O
rd
er
N
um
be
r
: u
ns
ig
ne
d
lo
ng
-D
at
e
: D
at
e
-C
us
t I
D
:
un
si
gn
ed
lo
ng
-L
as
t N
am
e
: S
tr
in
g
-F
ir
st
N
am
e
: S
tr
in
g
-S
ta
te
:
St
ri
ng
-T
ax
R
at
e
: fl
oa
t
-P
ro
du
ct
1
N
um
be
r
: u
ns
ig
ne
d
lo
ng
-P
ro
du
ct
1
D
es
c.
:
St
ri
ng
-P
ro
du
ct
1
P
ri
ce
:
do
ub
le
-P
ro
du
ct
1
Q
ty
. :
u
ns
ig
ne
d
lo
ng
-P
ro
du
ct
2
N
um
be
r
: u
ns
ig
ne
d
lo
ng
-P
ro
du
ct
2
D
es
c.
:
St
ri
ng
-P
ro
du
ct
2
P
ri
ce
:
do
ub
le
-P
ro
du
ct
2
Q
ty
. :
u
ns
ig
ne
d
lo
ng
-P
ro
du
ct
3
N
um
be
r
: u
ns
ig
ne
d
lo
ng
-P
ro
du
ct
3
D
es
c.
:
St
ri
ng
-P
ro
du
ct
3
P
ri
ce
:
do
ub
le
-P
ro
du
ct
3
Q
ty
. :
u
ns
ig
ne
d
lo
ng
O
rd
er
C
us
t
La
st
Fi
rs
t
Ta
x
Pr
od
. 1
Pr
od
. 1
Pr
od
. 1
Pr
od
. 1
Pr
od
. 2
Pr
od
. 2
Pr
od
. 2
Pr
od
. 2
Pr
od
. 3
Pr
od
. 3
P
ro
d.
3
Pr
od
. 3
N
um
be
r
D
at
e
ID
N
am
e
N
am
e
St
at
e
R
at
e
N
um
be
r
D
es
c.
Pr
ic
e
Q
ty
.
N
um
be
r
D
es
c.
Pr
ic
e
Q
ty
.
N
um
be
r
D
es
c.
P
ri
ce
Q
ty
.
2
39
11
/2
3/
00
10
35
B
la
ck
Jo
hn
M
D
0.
05
55
5
C
he
es
e
Tr
ay
$4
5.
00
2
2
60
11
/2
4/
00
10
35
B
la
ck
Jo
hn
M
D
0.
05
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
1
2
73
11
/2
7/
00
10
35
B
la
ck
Jo
hn
M
D
0.
05
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
2
41
11
/2
3/
00
11
23
W
ill
ia
m
s
M
ar
y
C
A
0.
08
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
2
2
62
11
/2
4/
00
11
23
W
ill
ia
m
s
M
ar
y
C
A
0.
08
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
2
2
87
11
/2
7/
00
11
23
W
ill
ia
m
s
M
ar
y
C
A
0.
08
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
2
2
90
11
/3
0/
00
11
23
W
ill
ia
m
s
M
ar
y
C
A
0.
08
55
5
C
he
es
e
Tr
ay
$4
5.
00
3
2
34
11
/2
3/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
55
5
C
he
es
e
Tr
ay
$4
5.
00
2
2
37
11
/2
3/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
11
1
W
in
e
G
ui
de
$1
5.
00
1
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
1
2
38
11
/2
3/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
1
2
45
11
/2
4/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
2
50
11
/2
4/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
2
52
11
/2
4/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
2
2
53
11
/2
4/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
44
4
W
in
e
G
ift
P
ac
k
$6
0.
00
1
2
97
11
/3
0/
00
22
42
D
eB
er
ry
A
nn
D
C
0.
06
5
33
3
Ja
m
s
&
Je
lli
es
$2
0.
00
2
2
43
11
/2
4/
00
42
54
B
ai
le
y
R
ya
n
M
D
0.
05
55
5
C
he
es
e
Tr
ay
$4
5.
00
2
2
46
11
/2
4/
00
42
54
B
ai
le
y
R
ya
n
M
D
0.
05
33
3
Ja
m
s
&
Je
lli
es
$2
0.
00
3
2
48
11
/2
4/
00
42
54
B
ai
le
y
R
ya
n
M
D
0.
05
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
33
3
Ja
m
s
&
Je
lli
es
$2
0.
00
2
11
1
W
in
e
G
ui
de
$
15
.0
0
1
2
35
11
/2
3/
00
95
00
C
hi
n
A
pr
il
K
S
0.
05
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
1
2
42
11
/2
3/
00
95
00
C
hi
n
A
pr
il
K
S
0.
05
33
3
Ja
m
s
&
Je
lli
es
$2
0.
00
3
2
44
11
/2
4/
00
95
00
C
hi
n
A
pr
il
K
S
0.
05
22
2
B
ot
tle
O
pe
ne
r
$1
2.
00
2
2
51
11
/2
4/
00
95
00
C
hi
n
A
pr
il
K
S
0.
05
11
1
W
in
e
G
ui
de
$1
5.
00
2
FI
G
U
R
E
9
-1
1
O
p
ti
m
iz
in
g
St
o
ra
ge
348
description, price, and quantity to repeat three times for each order in the table. Th e result-
ing table has many records that contain nulls in the product-related columns, and orders
are limited to three products because there is no room to store information for more.
A much more effi cient design (and one that conforms to 1NF) leads to a separate
table to hold the repeating information; to do this, we create a separate table on the model
to capture product order information. A zero-to-many relationship would then exist
between the two tables. As shown in Figure 9-13, the new design eliminates nulls from
the Order table and supports an unlimited number of products that can be associated
with an order.
Second normal form (2NF) requires fi rst that the data model is in 1NF and second that the
data model leads to tables containing fi elds that depend on a whole primary key. Th is means
that the primary key value for each record can determine the value for all the other fi elds in
the record. Sometimes fi elds depend on only part of the primary key (i.e., partial dependency),
and these fi elds belong in another table.
For example, in the new Product Order table that was created in Figure 9-13, the primary
key is a combination of the order number and product number, but the product description
and price attributes are dependent only upon product number. In other words, by knowing
product number, we can identify the product description and price. However, knowledge of
the order number and product number is required to identify the quantity. To rectify this
violation of 2NF, a table is created to store product information, and the description and
price attributes are moved into the new table. Now, product description is stored only once
for each instance of a product number as opposed to many times (every time a product is
placed on an order).
Do any tables have repeating fields? Do some
records have a different number of columns
from other records?
Yes: Remove the repeating fields. Add a new
table that contains the fields that repeat.
No: The data model is in 1NF
0 Normal Form
Is the primary key made up of more than one
field? If so, do any fields depend on only a part
of the primary key?
Yes: Remove the partial dependency. Add a
new table that contains the fields that are
partially dependent.
No: The data model is in 2NF
Do any fields depend on another nonprimary
key field?
Yes: Remove the transitive dependency.
Add a new table that contains the fields
that are transitively dependent.
No: The data model is in 3NF
Third Normal Form
Second Normal Form
First Normal Form
FIGURE 9-12
The Steps of
Normalization
Optimizing RDBMS-Based Object Storage 349
350 Chapter 9 Data Management Layer Design
Order
-Order Number : unsigned long
-Date : Date
-Cust ID : unsigned long
-Last Name : String
-First Name : String
-State : String
-Tax Rate : float
Product Order
-Order Number : unsigned long
-Product Number : String
-Product Desc : String
-Product Price : double
-Product Qty : unsigned long
Revised Model:
0..* 1..*
Note: Order Number will serve as part
of the primary key of Product Order
Note: Product Number will serve as part
of the primary key of Product Order
Note: Order Number also will serve as a
foreign key in Product Order
Note: Order Number will serve as part
of the primary key of Order
Note: Cust ID also will serve as part of
the primary key of Order
(a)
Sample Records:
Order Table
Order 248 has 3 products
Order 237 has 2 products
Order Cust Last First Tax
Number Date ID Name Name State Rate
239 11/23/00 1035 Black John MD 0.05
260 11/24/00 1035 Black John MD 0.05
273 11/27/00 1035 Black John MD 0.05
241 11/23/00 1123 Williams Mary CA 0.08
262 11/24/00 1123 Williams Mary CA 0.08
287 11/27/00 1123 Williams Mary CA 0.08
290 11/30/00 1123 Williams Mary CA 0.08
234 11/23/00 2242 DeBerry Ann DC 0.065
237 11/23/00 2242 DeBerry Ann DC 0.065
238 11/23/00 2242 DeBerry Ann DC 0.065
245 11/24/00 2242 DeBerry Ann DC 0.065
250 11/24/00 2242 DeBerry Ann DC 0.065
252 11/24/00 2242 DeBerry Ann DC 0.065
253 11/24/00 2242 DeBerry Ann DC 0.065
297 11/30/00 2242 DeBerry Ann DC 0.065
243 11/24/00 4254 Bailey Ryan MD 0.05
246 11/24/00 4254 Bailey Ryan MD 0.05
248 11/24/00 4254 Bailey Ryan MD 0.05
235 11/23/00 9500 Chin April KS 0.05
242 11/23/00 9500 Chin April KS 0.05
244 11/24/00 9500 Chin April KS 0.05
251 11/24/00 9500 Chin April KS 0.05
Product Order Table
Order Product Product Product Product
Number Number Desc Price Qty
239 555 Cheese Tray $45.00 2
260 444 Wine Gift Pack $60.00 1
273 222 Bottle Opener $12.00 1
241 444 Wine Gift Pack $60.00 2
262 222 Bottle Opener $12.00 2
287 222 Bottle Opener $12.00 2
290 555 Cheese Tray $45.00 3
234 555 Cheese Tray $45.00 2
237 111 Wine Guide $15.00 1
237 444 Wine Gift Pack $60.00 1
238 444 Wine Gift Pack $60.00 1
245 222 Bottle Opener $12.00 1
250 222 Bottle Opener $12.00 1
252 222 Bottle Opener $12.00 1
252 444 Wine Gift Pack $60.00 2
253 222 Bottle Opener $12.00 1
253 444 Wine Gift Pack $60.00 1
297 333 Jams & Jellies $20.00 2
243 555 Cheese Tray $45.00 2
246 333 Jams & Jellies $20.00 3
248 222 Bottle Opener $12.00 1
248 333 Jams & Jellies $20.00 2
248 111 Wine Guide $15.00 1
235 222 Bottle Opener $12.00 1
242 333 Jams & Jellies $20.00 3
244 222 Bottle Opener $12.00 2
251 111 Wine Guide $15.00 2
(b)
FIGURE 9-13 1NF: Remove Repeating Fields
A second violation of 2NF occurs in the Order table: customer fi rst name and last name
depend only upon the customer ID, not the whole key (Cust ID and Order number). As a result,
every time the customer ID appears in the Order table, the names also appear. A much more
economical way of storing the data is to create a Customer table with the Customer ID as the
primary key and the other customer-related fi elds (i.e., last name and fi rst name) listed only
once within the appropriate record. Figure 9-14 illustrates how the model would look when
placed in 2NF.
Th ird normal form (3NF) occurs when a model is in both 1NF and 2NF and, in the result-
ing tables, none of the fi elds depend on nonprimary key fi elds (i.e., transitive dependency).
Figure 9-14 contains a violation of 3NF: Th e tax rate on the order depends upon the state to
which the order is being sent. Th e solution involves creating another table that contains state
abbreviations serving as the primary key and the tax rate as a regular fi eld. Figure 9-15 presents
the end results of applying the steps of normalization to the original model from Figure 9-11.
Optimizing Data Access Speed
Aft er you have optimized the design of the object storage for effi ciency, the end result is that
data are spread out across a number of tables. When data from multiple tables need to be
accessed or queried, the tables must be fi rst joined. For example, before a user can print out
a list of the customer names associated with orders, fi rst the Customer and Order tables need
to be joined, based on the customer number fi eld (see Figure 9-15). Only then can both the
order and customer information be included in the query’s output. Joins can take a lot of
time, especially if the tables are large or if many tables are involved.
Consider a system that stores information about 10,000 diff erent products, 25,000 cus-
tomers, and 100,000 orders, each averaging three products per order. If an analyst wanted
to investigate whether there were regional diff erences in music preferences, he or she would
need to combine all the tables to be able to look at products that have been ordered while
knowing the state of the customers placing the orders. A query of this information would
result in a huge table with 300,000 rows (i.e., the number of products that have been ordered)
and 11 columns (the total number of columns from all of the tables combined).
Th e project team can use several techniques to try to speed up access to the data, includ-
ing denormalization, clustering, and indexing.
Denormalization Aft er the object storage is optimized, the project team may decide that
increased data retrieval speed is more important than storage effi ciency or data update speed
and elect to denormalize or add redundancy back into the design. Denormalization reduces
the number of joins that need to be performed in a query, thus speeding up access. Figure 9-16
shows a denormalized model for customer orders. Th e customer last name was added back
into the Order table because the project team learned during analysis that queries about orders
usually require the customer last name fi eld. Instead of joining the Order table repeatedly to
the Customer table, the system now needs to access only the Order table because it contains all
of the relevant information needed to solve the music preference question posed above.
Denormalization is ideal in situations in which information is queried frequently but
updated rarely. However, due to the additional storage required and the potential update
anomalies, denormalization should be applied sparingly. Th ere are three cases in which you
may rely upon denormalization to reduce joins and improve performance. First, denormaliza-
tion can be applied in the case of look-up tables, which are tables that contain descriptions of
values (e.g., a table of product descriptions or a table of payment types). Because descriptions
of codes rarely change, it may be more effi cient to include the description along with its respec-
tive code in the main table to eliminate the need to join the look-up table each time a query is
performed (see Figure 9-17a).
Optimizing RDBMS-Based Object Storage 351
352 Chapter 9 Data Management Layer Design
Customer
-Cust ID : unsigned long
-Last Name : String
-First Name : String
Order
-Order Number : unsigned long
-Date : Date
-Cust ID : unsigned long
-State : String
-Tax Rate : float
Product Order
-Order Number : unsigned long
-Product Number : unsigned long
-Qty : unsigned long
Product
-Product Number : unsigned long
-Product Desc : String
-Price : double1..1 0..*
1..*0..*
Sample Records:
Customer Table
Cust Last First
ID Name Name
1035 Black John
1123 Williams Mary
2242 DeBerry Ann
4254 Bailey Ryan
9500 Chin April
Product Table
Product Product Product
Number Desc Price
111 Wine Guide $15.00
222 Bottle Opener $12.00
333 Jams & Jellies $20.00
444 Wine Gift Pack $60.00
555 Cheese Tray $45.00
Product Order Table
Order Product Product
Number Number Qty
239 555 2
260 444 1
273 222 1
241 444 2
262 222 2
287 222 2
290 555 3
234 555 2
237 111 1
237 444 1
238 444 1
245 222 1
250 222 1
252 222 1
252 444 2
253 222 1
253 444 1
297 333 2
243 555 2
246 333 3
248 222 1
248 333 2
248 111 1
235 222 1
242 333 3
244 222 2
251 111 2
Order Table
Order Cust
Number Date ID State
239 11/23/00 1035 MD
260 11/24/00 1035 MD
273 11/27/00 1035 MD
241 11/23/00 1123 CA
262 11/24/00 1123 CA
287 11/27/00 1123 CA
290 11/30/00 1123 CA
234 11/23/00 2242 DC
237 11/23/00 2242 DC
238 11/23/00 2242 DC
245 11/24/00 2242 DC
250 11/24/00 2242 DC
252 11/24/00 2242 DC
253 11/24/00 2242 DC
297 11/30/00 2242 DC
243 11/24/00 4254 MD
246 11/24/00 4254 MD
248 11/24/00 4254 MD
235 11/23/00 9500 KS
242 11/23/00 9500 KS
244 11/24/00 9500 KS
251 11/24/00 9500 KS
Note: Order Number will
serve as part of the
primary key of Product
Order.
Note: Order Number also
will serve as a foreign key
in Product Order.
Note: Product Number
will serve as part of the
primary key in Product
Order.
Note: Product Number
also will serve as a foreign
key in Product Order.
Note: Product Number will
serve as part of the primary
key of Product Order.
Note: Order Number will serve
as the primary key of Order.
Note: Cust ID will serve as a
foreign key in Order.
Note: Cust ID will serve
as the primary key of
Customer.
Product Desc and Price
was moved to the Product
table to eliminate
redundancy
Last Name and First
Name was moved
to the Customer
table to eliminate
redundancy
Tax
Rate
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.05
0.08
0.08
0.08
0.08
0.065
0.065
0.065
0.065
0.065
0.065
0.065
0.065
FIGURE 9-14 2NF Partial Dependencies Removed
Customer
-Cust ID : unsigned long
-Last Name : String
-First Name : String
State
-State : String
-Tax Rate : float
Order
-Order Number : unsigned long
-Date : Date
-Cust ID : unsigned long
-State : String
Product Order
-Order Number : unsigned long
-Product Number : unsigned long
-Qty : unsigned long
Product
-Product Number : unsigned long (idl)
-Product Desc : String
-Price : double1..1
0..*
1..1
0..*
1..*0..*
FIGURE 9-15 3NF Normalized Field
Second, one-to-one relationships are good candidates for denormalization. Although log-
ically two tables should be separated, from a practical standpoint the information from both
tables may regularly be accessed together. Th ink about an order and its shipping information.
Logically, it might make sense to separate the attributes related to shipping into a separate
table, but as a result the queries regarding shipping will probably always need a join to the
Order table. If the project team fi nds that certain shipping information, such as state and ship-
ping method, is needed when orders are accessed, they may decide to combine the tables or
include some shipping attributes in the Order table (see Figure 9-17b).
Th ird, at times it is more effi cient to include a parent entity’s attributes in its child entity on
the physical data model. For example, consider the Customer and Order tables in Figure 9-16,
which share a one-to-many relationship, with Customer as the parent and Order as the child.
FIGURE 9-16
Denormalized
Physical Data Model
Customer
-Cust ID (PK) : unsigned long
-Last Name : String
-First Name : String
Order
-Order Number (PK) : unsigned long
-Date : Date
-State (FK) : String
-Cust ID (FK) : unsigned long
-Customer Last Name : String
1..1 1..*
Last name is now in both classes
Optimizing RDBMS-Based Object Storage 353
354 Chapter 9 Data Management Layer Design
If queries regarding orders continuously require customer information, the most popular cus-
tomer fi elds can be placed in Order to reduce the required joins to the Customer table, as was
done with Customer Last Name.
Clustering Speed of access also is infl uenced by the way that the data are retrieved. Th ink
about shopping in a grocery store. If you have a list of items to buy but you are unfamiliar
with the store’s layout, you need to walk down every aisle to make sure that you don’t miss
anything from your list. Likewise, if records are arranged in no particular order (or in an
order that is irrelevant to your data needs), then any query of the records results in a table
scan in which the DBMS has to access every row in the table before retrieving the result set.
Table scans are the most ineffi cient of data retrieval methods.
One way to improve access speed is to reduce the number of times that the storage
medium needs to be accessed during a transaction. One method is to cluster records together
physically so that similar records are stored close together. With intrafi le clustering, like
records in the table are stored together in some way, such as in order by primary key or, in
the case of a grocery store, by item type. Th us, whenever a query looks for records, it can go
directly to the right spot on the disk (or other storage medium) because it knows in what
order the records are stored, just as we can walk directly to the bread aisle to pick up a loaf
of bread. Interfi le clustering combines records from more than one table that typically are
Shipment
-Shipment ID (PK) : unsigned long
-Shipment Street Address : String
-Shipment City : String
-Shipment State : String
-Shipment Zip Code : String
-Shipment Method : String
Order
-Order: Number (PK) : unsigned long
-Date : Date
-State (FK) : String
-Cust ID (FK) : unsigned long
-Customer Last Name : String
-Payment Type (FK) : unsigned long
-Payment Description : String
-Shipment ID (FK) : unsigned long
-Shipment State : String
-Shipment Method : String
Order
-Order : Number (PK) : unsigned long
-Date : Date
-State (FK) : String
-Cust ID (FK) : unsigned long
-Customer Last Name : String
-Payment Type (FK) : unsigned long
-Payment Description : String
1..1 1..1
1..1 0..*
Notice that the payment description field
appears in both Payment Type and Order.
(a)
Notice that the shipment state and shipment method
are included in both Shipment and Order.
(b)
Payment Type
-Payment Type (PK) : String
-Payment Description : String
FIGURE 9-17
Denormalization
Situations (FK, foreign
key; PK, primary key)
retrieved together. For example, if customer information is usually accessed with the related
order information, then the records from the two tables may be physically stored in a way
that preserves the customer-order relationship. Returning to the grocery store scenario, an
interfi le cluster would be similar to storing peanut butter, jelly, and bread next to each other
in the same aisle because they are usually purchased together, not because they are similar
types of items. Of course, each table can have only one clustering strategy because the records
can be arranged physically in only one way.
Indexing A familiar time saver is an index located in the back of a textbook, which points
directly to the page or pages that contain a topic of interest. Th ink of how long it would take
to fi nd all the times that relational database appears in this textbook without the index to rely
on! An index in data storage is like an index in the back of a textbook; it is a minitable that
contains values from one or more columns in a table and the location of the values within the
table. Instead of paging through the entire textbook, we can move directly to the right pages
and get the information we need. Indexes are one of the most important ways to improve
database performance. Whenever there are performance problems, the fi rst place to look is
an index.
A query can use an index to fi nd the locations of only those records that are included in
the query answer, and a table can have an unlimited number of indexes. Figure 9-18 shows
an index that orders records by payment type. A query that searches for all the customers
who used American Express can use this index to fi nd the locations of the records that con-
tain American Express as the payment type without having to scan the entire Order table.
Order
Order Cust Payment
Number Date ID Amount Tax Total Type
234 11/23/00 2242 $ 90.00 $5.85 $ 95.85 MC
235 11/23/00 9500 $ 12.00 $0.60 $ 12.60 VISA
236 11/23/00 1556 $ 50.00 $2.50 $ 52.50 VISA
237 11/23/00 2242 $ 75.00 $4.88 $ 79.88 AMEX
238 11/23/00 2242 $ 60.00 $3.90 $ 63.90 MC
239 11/23/00 1035 $ 90.00 $4.50 $ 94.50 AMEX
240 11/23/00 9501 $ 50.00 $2.50 $ 52.50 VISA
241 11/23/00 1123 $ 120.00 $9.60 $ 129.60 MC
242 11/24/00 9500 $ 60.00 $3.00 $ 63.00 VISA
243 11/24/00 4254 $ 90.00 $4.50 $ 94.50 VISA
244 11/24/00 9500 $ 24.00 $1.20 $ 25.20 VISA
245 11/24/00 2242 $ 12.00 $0.78 $ 12.78 AMEX
246 11/24/00 4254 $ 20.00 $1.00 $ 21.00 MC
247 11/24/00 2241 $ 50.00 $2.50 $ 52.50 VISA
248 11/24/00 4254 $ 12.00 $0.60 $ 12.60 AMEX
249 11/24/00 5927 $ 50.00 $2.50 $ 52.50 AMEX
250 11/24/00 2242 $ 12.00 $0.78 $ 12.78 MC
251 11/24/00 9500 $ 15.00 $0.75 $ 15.75 MC
252 11/24/00 2242 $ 132.00 $8.58 $ 140.58 MC
253 11/24/00 2242 $ 72.00 $4.68 $ 76.68 AMEX
Payment Type Index
Payment
Type Pointer
AMEX *
AMEX *
AMEX *
AMEX *
AMEX *
AMEX *
MC *
MC *
MC *
MC *
MC *
MC *
MC *
VISA *
VISA *
VISA *
VISA *
VISA *
VISA *
VISA *
VISA *
FIGURE 9-18 Payment Type Index
Optimizing RDBMS-Based Object Storage 355
356 Chapter 9 Data Management Layer Design
Project teams can make indexes perform even faster by placing them into the main
memory of the data storage hardware. Retrieving information directly from memory is
much faster than retrieving it from a hard disk—Th ink about how much faster it is to
retrieve a memorized phone number versus one that must be looked up in a phone book.
Similarly, when a database has an index in memory, it can locate records very, very quickly.
Of course, indexes require overhead in that they take up space on the storage medium.
Also, they need to be updated as records in tables are inserted, deleted, or changed. Th us,
although indexes lead to faster access to the data, they slow down the update process. In
general, we should create indexes sparingly for transaction systems or systems that require a
lot of updates, but we should apply indexes generously when designing systems for decision
support (see Figure 9-19).
Estimating Data Storage Size
Even if we have denormalized our physical data model, clustered records, and created indexes
appropriately, the system will perform poorly if the database server cannot handle its vol-
ume of data. Th erefore, one last way to plan for good performance is to apply volumetrics,
which means estimating the amount of data that the hardware will need to support. You can
incorporate your estimates into the database server hardware specifi cation to make sure that
the database hardware is suffi cient for the project’s needs. Th e size of the database is based
on the amount of raw data in the tables and the overhead requirements of the DBMS. To
estimate size, you will need to have a good understanding of the initial size of your database
as well as its expected growth rate over time.
Raw data refers to all the data that are stored within the tables of the database, and it is
calculated based on a bottom-up approach. First, write down the estimated average width
for each column (fi eld) in the table and sum the
values for a total record size (see Figure 9-20). For
example, if a variable-width Last Name column is
assigned a width of 20 characters, you can enter 13
as the average character width of the column. In
Figure 9-20, the estimated record size is 49.
Next, calculate the overhead for the table as
a percentage of each record. Overhead includes
the room needed by the DBMS to support such
functions as administrative actions and indexes,
and it should be assigned based on past experience,
recommendations from technology vendors, or
parameters that are built into soft ware that was
written to calculate volumetrics. For example, your
DBMS vendor might recommend that you allocate
30 percent of the records’ raw data size for over-
head storage space, creating a total record size of
63.7 in the Figure 9-20 example.
Use indexes sparingly for transaction systems.
Use many indexes to increase response times in decision support systems.
For each table, create a unique index that is based on the primary key.
For each table, create an index that is based on the foreign key to improve the performance of joins.
Create an index for fi elds that are used frequently for grouping, sorting, or criteria.
FIGURE 9-19
Guidelines for
Creating Indexes
FIGURE 9-20
Calculating
Volumetrics
Order Number 8
Date 7
Cust ID 4
Last Name 13
First Name 9
State 2
Amount 4
Tax Rate 2
Record Size 49
Overhead 30%
Total Record Size 63.7
Initial Table Size 50,000
Initial Table Volume 3,185,000
Growth Rate/Month 1,000
Table Volume @ 3 years 5,478,200
Field Average Size
Designing Data Access and Manipulation Classes 357
Finally, record the number of initial records that will be loaded into the table, as well as
the expected growth per month. Th is information should have been collected during analysis.
According to Figure 9-20, the initial space required by the fi rst table is 3,185,000, and future
sizes can be project based on the growth fi gure. Th ese steps are repeated for each table to get
a total size for the entire database.
Many CASE tools provide you with database-size information based on how you set up the
object persistence, and they calculate volumetrics estimates automatically. Ultimately, the size
of the database needs to be shared with the design team so that the proper technology can be
put in place to support the system’s data and potential performance problems can be addressed
long before they aff ect the success of the system.
DESIGNING DATA ACCESS AND MANIPULATION CLASSES
Th e fi nal step in developing the data management layer is to design the data access and
manipulation classes that act as a translator between the object persistence and the problem
domain objects. Th us, they should always be capable of at least reading and writing both the
object persistence and problem domain objects. As described earlier and in Chapter 8, the
object persistence classes are derived from the concrete problem domain classes, whereas
the data access and manipulation classes depend on both the object persistence and problem
domain classes.
Depending on the application, a simple rule to follow is that there should be one data
access and manipulation class for each concrete problem domain class. In some cases,
it might make sense to create data access and manipulation classes associated with the
human–computer interaction classes (see Chapter 10). However, this creates a depend-
ency from the data management layer to the human–computer interaction layer. Adding
this additional complexity to the design of the system normally is not recommended.
Returning to the ORDBMS solution for the Appointment system example (see
Figure 9-8), we see that we have four problem domain classes and four ORDBMS tables.
Following the previous rule, the DAM classes are rather simple. Th ey have to support only
a one-to-one translation between the concrete problem domain classes and the ORDBMS
tables (see Figure 9-21). Because the Participant problem domain class is an abstract class,
only three data access and manipulation classes are required: Patient-DAM, Symptom-
DAM, and Appointment-DAM. However, the process to create an instance of the Patient
problem domain class can be fairly complicated. Th e Patient-DAM class might have to be
able to retrieve information from all four ORDBMS tables. To accomplish this, the Patient-
DAM class retrieves the information from the Patient table. Using the Object-IDs stored
in the attribute values associated with the Participant, Appts, and Symptoms attributes, the
remaining information required to create an instance of Patient is easily retrieved by the
Patient-DAM class.
In the case of using an RDBMS to provide persistence, the data access and manipula-
tion classes tend to become more complex. For example, in the Appointment system, there
are still four problem domain classes, but, owing to the limitations of RDBMSs, we have to
support six RDBMS tables (see Figure 9-10). Th e data access and manipulation class for the
Appointment problem domain class and the Appointment RDBMS table is no diff erent from
those supported for the ORDBMS solution (see Figures 9-21 and 9-22). However, owing
to the multivalued attributes and relationships associated with the Patient and Symptom
problem domain classes, the mappings to the RDBMS tables were more complicated.
Consequently, the number of dependencies from the data access and manipulation classes
(Patient-DAM and Symptom-DAM) to the RDBMS tables (Patient table, Insurance Carrier
table, Suff er table, and the Symptom table) has increased. Furthermore, because the Patient
FI
G
U
R
E
9
-2
1
M
an
ag
in
g
Pr
o
b
le
m
D
o
m
ai
n
O
b
je
ct
s
to
O
R
D
B
M
S
U
si
n
g
D
A
M
C
la
ss
es
O
R
D
B
M
S
Ta
bl
es
Pr
ob
le
m
D
om
ai
n
C
la
ss
es
D
at
a
A
cc
es
s
an
d
M
an
ip
ul
at
io
n
C
la
ss
es
Pa
rt
ic
ip
an
t T
ab
le
-l
as
tn
am
e[
1.
.1
]
-fi
rs
tn
am
e[
1.
.1
]
-a
dd
re
ss
[1
..1
]
-p
ho
ne
[1
..1
]
-b
ir
th
da
te
[1
..1
]
-S
ub
C
la
ss
O
bj
ec
ts
[1
..1
]
Pa
ti
en
t-
D
A
M
+
R
ea
dP
at
ie
nt
Ta
bl
e(
)
+
W
ri
te
Pa
tie
nt
Ta
bl
e(
)
+
R
ea
dP
at
ie
nt
()
+
W
ri
te
Pa
tie
nt
()
Sy
m
pt
om
-D
A
M
+
R
ea
dS
ym
pt
om
Ta
bl
e(
)
+
W
ri
te
Sy
m
pt
om
Ta
bl
e(
)
+
R
ea
dS
ym
pt
om
()
+
W
ri
te
Sy
m
pt
om
()
A
pp
oi
nt
m
en
t-
D
A
M
+
R
ea
dA
pp
tT
ab
le
()
+
W
ri
te
A
pp
tT
ab
le
()
+
R
ea
dA
pp
t()
+
W
ri
te
A
pp
t()
Pa
ti
en
t T
ab
le
-a
m
ou
nt
[1
..1
]
-P
ar
tic
ip
an
t[
1.
.1
]
-A
pp
ts
[0
..*
]
-S
ym
pt
om
s[
1.
.*
]
-i
ns
ur
an
ce
c
ar
ri
er
[0
..*
]
-P
ri
m
ar
y
In
su
ra
nc
e
C
ar
ri
er
[0
..*
]
Sy
m
pt
om
T
ab
le
-n
am
e[
1.
.1
]
-P
at
ie
nt
s[
0.
.*
]
A
pp
oi
nt
m
en
t T
ab
le
-P
at
ie
nt
[1
..1
]
-t
im
e[
1.
.1
]
-d
at
e[
1.
.1
]
-r
ea
so
n[
1.
.1
]
1.
.1
1.
.1
0.
.*
0.
.*
0.
.*
0.
.*
0.
.*
1
1
1.
.*
1.
.*
0.
.*
1
1
su
ffe
rs
sc
he
du
le
s
A
pp
oi
nt
m
en
t
-t
im
e
-d
at
e
-r
ea
so
n
+c
an
ce
l w
ith
ou
t n
ot
ic
e(
)
Pa
ti
en
t
-a
m
ou
nt
-i
ns
ur
an
ce
c
ar
ri
er
+m
ak
e
ap
po
in
tm
en
t(
)
+c
al
cu
la
te
la
st
v
is
it(
)
+c
ha
ng
e
st
at
us
()
+p
ro
vi
de
s
m
ed
ic
al
h
is
to
ry
()
Pa
rt
ic
ip
an
t
-l
as
tn
am
e
-fi
rs
tn
am
e
-a
dd
re
ss
-p
ho
ne
-b
ir
th
da
te
-/
a
ge
Sy
m
pt
om
-n
am
e
+
p
ri
m
ar
y
in
su
ra
nc
e
ca
rr
ie
r
358
Designing Data Access and Manipulation Classes 359
problem domain class is associated with the other three problem domain classes, the actual
retrieval of all information necessary to create an instance of the Patient class could involve
joining information from all six RDBMS tables. To accomplish this, the Patient-DAM class
must fi rst retrieve information from the Patient table, Insurance Carrier table, Suff er table,
and the Appointment table. Because the primary keys of the Patient table and the Participant
table are identical, the Patient-DAM class can either directly retrieve the information from
the Participant table, or the information can be joined using the participantNumber attributes
of the two tables, which act as both primary and foreign keys. Finally, using the information
RDBMS Tables Problem Domain ClassesData Access and
Manipulation Classes
Participant Table
-lastname[1..1]
-firstname[1..1]
-address[1..1]
-phone[1..1]
-birthdate[1..1]
-participantNumber[1..1]
Patient-DAM
+ReadPatientTable()
+WritePatientTable()
+ReadInsuranceCarrierTable()
+WriteInsuranceCarrierTable()
+ReadSufferTable()
+WriteSufferTable()
+ReadApptTable()
+WriteApptTable()
+ReadPatient()
+WritePatient()
Symptom-DAM
+ReadSymptomTable()
+WriteSymptomTable()
+ReadSufferTable()
+WriteSufferTable()
+ReadSymptom()
+WriteSymptom()
Appointment-DAM
+ReadApptTable()
+WriteApptTable()
+ReadAppt()
+WriteAppt()
Patient Table
-amount[1..1]
-participantNumber[1..1]
-primaryInsuranceCarrier[0..1]
Appointment Table
-time[1..1]
-date[1..1]
-reason[1..1]
-personNumber[1..1]
1..1
1..1
1..1
0..* 0..*
0..*
1
1
1..1
1..*
Symptom Table
-name[1..1]
Suffer Table
-participantNumber[1..1]
-name[1..1]
Insurance Carrier Table
-name[1..1]
-participantNumber[1..1]
1..*
+ primary
insurance
carrier
0..*
0..*
1..1
1..1
1..1
0..*
Symptom
-name
Appointment
-time
-date
-reason
+cancel without notice()Patient
-amount
-insurance carrier
+make appointment()
+calculate last visit()
+change status()
+provides medical history()
Participant
-lastname
-firstname
-address
-phone
-birthdate
-/ age
suffers
schedules
FIGURE 9-22 Mapping Problem Domain Objects to RDBMS Using DAM Classes
360 Chapter 9 Data Management Layer Design
contained in the Suff er table, the information in the Symptom table can also be retrieved.
Obviously, the farther we get from the object-oriented problem domain class representation,
the more work must be performed. However, as in the case of the ORDBMS example, notice
that absolutely no modifi cations were made to the problem domain classes. Th erefore, the
data access and manipulation classes again have prevented data management functionality
from creeping into the problem domain classes.
One specifi c approach that has been suggested to support the implementation of
data access and manipulation classes is to use an object-relational mapping library such
as Hibernate.11 Hibernate, developed within the JBoss community, allows the mapping
of objects written in Java that are to be stored in an RDBMS. Instead of using an object-
oriented programming language to implement the data access and manipulation classes, with
Hibernate, they are implemented in XML fi les that contain the mapping. As in the above
approach, modeling the mapping in an XML fi le prevents the details on data access and
manipulation from sneaking into the problem domain representation.
NONFUNCTIONAL REQUIREMENTS AND DATA
MANAGEMENT LAYER DESIGN12
Recall that nonfunctional requirements refer to behavioral properties that the system must
have. Th ese properties include issues related to performance, security, ease of use, operational
environment, and reliability. In this text, we have grouped nonfunctional requirements into
four categories: operational, performance, security, and cultural and political requirements.
We describe each of these in relation to the data management layer.
Th e operational requirements for the data management layer include issues that deal with
the technology being used to support object persistence. However, the choice of the hardware
and operating system limits the choice of the technology and format of the object persistence
available. Th is is especially true when you consider mobile computing. Given the limited
memory and storage available on these devices, the choices to support object persistence
are limited. One possible choice to support object persistence that works both on Google’s
Android and Apple’s iOS-based platforms is SQLite. SQLite is a lightweight version of SQL
that supports RDBMS. However, there are many diff erent approaches to support object per-
sistence that are more platform dependent; for example, Android supports storing objects
with shared preferences (a key-value pair-based NoSQL approach), internal storage, on an
SD card, in a local cache, or on a remote system. Th is, in turn, determines which set of the
mapping rules described earlier will have to be used. Another operational requirement could
be the ability to import and export data using XML. Again, this could limit the object stores
under consideration.
Th e primary performance requirements that aff ect the data management layer are speed
and capacity. As described before, depending on the anticipated—and, aft erwards, actual—
usage patterns of the objects being stored, diff erent indexing and caching approaches may
be necessary. When considering distributing objects over a network, speed considerations
can cause objects to be replicated on diff erent nodes in the network. Th us, multiple copies
of the same object may be stored in diff erent locations on the network. Th is raises the issue
of update anomalies described before in conjunction with normalization. Depending on
the application being built, NoSQL data stores that support an eventually consistent update
11 For more information on Hibernate, see www.hibernate.org.
12 Because the vast majority of nonfunctional requirements aff ect the physical architecture layer, we provide
additional details in Chapter 11.
Verifying and Validating the Data Management Layer 361
model may be appropriate. Also, depending on the estimated size and growth of the system,
diff erent DBMSs may need to be considered. An additional requirement that can aff ect the
design of the data management layer deals with the availability of the objects being stored. It
might make sense to limit the availability to diff erent objects based on the time of day. For
example, one class of users may be allowed to access a set of objects only from 8 to 12 in the
morning and a second set of users may be able to access them only from 1 to 5 in the aft er-
noon. Th rough the DBMS, these types of restrictions could be set.
Th e security requirements deal primarily with access controls, encryption, and backup.
Th rough a modern DBMS, diff erent types of access can be set (e.g., Read, Update, or Delete)
granting access only to users (or class of users) who have been authorized. Furthermore,
access control can be set to guarantee that only users with “administrator” privileges are
allowed to modify the object storage schema or access controls. Encryption requirements
on this layer deal with whether the object should be stored in an encrypted format or not.
Even though encrypted objects are more secure than unencrypted objects, the process of
encrypting and decrypting the objects will slow down the system. Depending on the physical
architecture being used, the cost of encryption may be negligible. For example, if we plan on
encrypting the objects before transmitting them over a network, there may be no additional
cost of storing them in the encrypted format. Backup requirements deal with ensuring that
the objects are routinely copied and stored in case the object store becomes corrupted or
unusable. Having a backup copy made on a periodic basis and storing the updates that have
occurred since the last backup copy was made ensure that the updates are not lost and the
object store can be reconstituted by running the copies of the updates against the backup copy
to create a new current copy.
Th ere are few political and cultural requirements that can aff ect the data management layer.
Th ese include issues related to the expected number of characters that should be allocated for a
data fi eld, the format of a data fi eld, and the issues related to security. For example, how many
characters should be allocated for a last name fi eld that is part of an Employee object, what
format should a date be stored, or where will the data be physically located—diff erent parts of
the world have diff erent laws regarding the protection of data. Finally, there could be a corpo-
rate IT bias toward diff erent hardware and soft ware platforms. If so, this could limit the type
of object store available.
VERIFYING AND VALIDATING THE DATA
MANAGEMENT LAYER
Like the models on the problem domain layer, the specifi cations for the data management layer
need to be verifi ed and validated. By now, it might seem a little heavy handed to insist on more
verifying and validating. However, depending on the object persistence chosen, the changes that
have been applied to the design of the evolving system may be very substantial. Consequently,
it is crucial to thoroughly test the fi delity of the design again before the system is implemented.
Without thoroughly testing the data management layer, there is no guarantee that an effi cient
and eff ective system will be implemented. Verifying and validating the design of the data man-
agement layer fall into three basic groups.
First, we recommend verifying and validating any changes made to the problem domain
by performing walkthroughs of the modifi ed functional models (Chapter 4), structural
models (Chapter 5), and behavioral models (Chapter 6). Furthermore, all of the models
must be consistent and balanced (Chapter 7). And, if any problem domain class was mod-
ifi ed that was associated with a use-case scenario, that scenario should be tested again
through role-playing.
362 Chapter 9 Data Management Layer Design
Second, the dependency of the object persistence instances on the problem domain must be
enforced. For example, all invariants associated with a problem domain class (Chapter 8) need
to be verifi ed and validated. For example, if a name data fi eld is specifi ed in a problem domain
class as being thirty-fi ve characters long and as being a required fi eld, then similar constraints
must be enforced when the fi eld is stored.
Th ird, the design of the data access and manipulation classes need to be tested to ensure that
they are dependent on the problem domain classes and the object persistence format, not the
other way around. For example, in Figure 9-21, we see that the Patient-DAM class is dependent
on both the Patient problem domain class and the Patient table.
Once the system has been implemented, testing of the data management layer becomes
even more important. One issue that should be addressed is the testing of the nonfunctional
requirements. In this case, tests must be designed and performed for each of the nonfunc-
tional requirements. For example, for the performance requirements, load testing must be
performed to identify possible performance bottlenecks in the database. We will return to this
topic in Chapter 12.
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Ben Joseph, the data specialist, led the team in designing the model for the data management
layer for the fi rst phase of the Integrated Health Clinic Delivery System. Th eir fi rst step was
to choose the object persistence format that the system would use. Th en they mapped the
problem domain to the object persistence classes. Th ey also checked for optimization oppor-
tunities. Finally, they designed the Data Access and Manipulation (DAM) classes.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the diff erent types of object persistence formats.
Select the appropriate object persistence format based on its strengths and weaknesses.
Map a set of problem domain objects to an OODBMS format.
Map a set of problem domain objects to an ORDBMS format.
Map a set of problem domain objects to an RDBMS format.
Use normalization to minimize update anomalies and to increase storage effi ciency.
Describe the fi rst three normal forms.
Describe when to use denormalization, clustering, and indexing to increase the speed of data access.
Describe why denormalization, clustering, and indexing can slow down updating.
Apply volumetrics to estimate the amount of data storage required.
Create a set of data access and manipulation classes that act as a communication layer between the problem domain
layer and the actual object persistence used.
Describe why the operational and performance nonfunctional requirements of the object persistence format are
constrained by decisions made regarding the physical architecture layer.
Describe how the nonfunctional requirements of the object persistence format may infl uence the actual design of
the data management layer; these include both the object persistence format and the data access and manipulation
classes.
Understand how to verify and validate both the design of both the object persistence format and the data access and
manipulation classes.
Questions 363
KEY TERMS
Access control
Attribute sets
Audit fi le
Cluster
Column-oriented data
stores
Data access and manipula-
tion classes
Data management layer
Database
Database management
system (DBMS)
Decision support systems
(DSS)
Denormalization
Document data stores
End-user DBMS
Enterprise DBMS
Executive information
systems (EIS)
Expert system (ES)
Extent
File
First normal form (1NF)
Foreign key
Hardware and operating
system
History fi le
Impedance mismatch
Index
Interfi le clustering
Intrafi le clustering
Join
Key-value data stores
Linked list
Lookup fi le
Management information
system (MIS)
Master fi le
Multivalued attributes
(fi elds)
Normalization,
NoSQL data stores
Object ID
Object-oriented database
management system
(OODBMS)
Object-oriented
programming language
(OOPL)
Object persistence
Object-relational database
management system
(ORDBMS)
Operational requirements
Ordered sequential
access fi le
Overhead
Partial dependency
Performance requirements
Pointer
Political and cultural
requirements
Primary key
Problem domain classes
Random access fi les
Raw data
Referential integrity
Relational database
management system
(RDBMS)
Repeating groups (fi elds)
Second normal form
(2NF)
Security requirements
Sequential access fi les
Structured query language
(SQL)
Table scan
Th ird normal form (3NF)
Transaction fi le
Transaction-processing
system
Transitive dependency
Unordered sequential
access fi le
Update anomaly
Volumetrics
QUESTIONS
1. Describe the four steps in object persistence design.
2. How are a fi le and a database diff erent from each other?
3. What is the diff erence between an end-user database and
an enterprise database? Provide an example of each one.
4. What are the diff erences between sequential and ran-
dom access fi les?
5. Name fi ve types of fi les and describe the primary pur-
pose of each type.
6. What is the most popular kind of database today?
Provide three examples of products that are based on
this database technology.
7. What is referential integrity and how is it imple-
mented in an RDBMS?
8. List some of the diff erences between an ORDBMS and
an RDBMS.
9. What are the advantages of using an ORDBMS over
an RDBMS?
10. List some of the diff erences between an ORDBMS and
an OODBMS.
11. What are the advantages of using an ORDBMS over
an OODBMS?
12. What are the advantages of using an OODBMS over
an RDBMS?
13. What are the advantages of using an OODBMS over
an ORDBMS?
14. What are the factors in determining the type of
object persistence format that should be adopted for a
system? Why are these factors so important?
15. Why should you consider the storage formats that
already exist in an organization when deciding upon a
storage format for a new system?
16. When implementing the object persistence in an
ORDBMS, what types of issues must you address?
17. When implementing the object persistence in an
RDBMS, what types of issues must you address?
18. Name three ways null values can be interpreted in a
relational database. Why is this problematic?
19. What are the two dimensions in which to optimize a
relational database?
20. What is the purpose of normalization?
21. How does a model meet the requirements of third
normal form?
364 Chapter 9 Data Management Layer Design
EXERCISES
A. Using the Web or other resources, identify a product
that can be classifi ed as an end-user database and a
product that can be classifi ed as an enterprise data-
base. How are the products described and marketed?
What kinds of applications and users do they support?
In what kinds of situations would an organization
choose to implement an end-user database over an
enterprise database?
B. Visit a commercial website (e.g., Amazon.com). If fi les
were being used to store the data supporting the appli-
cation, what types of fi les would be needed? What
access type would be required? What data would they
contain?
C. Using the Web, review one of the following prod-
ucts. What are the main features and functions of
the soft ware? In what companies has the DBMS
been implemented, and for what purposes? Accord-
ing to the information that you found, what are
three strengths and weaknesses of the product?
1. Relational DBMS
2. Object-relational DBMS
3. Object-oriented DBMS
D. You have been given a fi le that contains the following
fi elds relating to CD information. Using the steps of
normalization, create a model that represents this fi le
in third normal form. Th e fi elds include:
Musical group name CD title 2
Musicians in group CD title 3
Date group was formed CD 1 length
Group’s agent CD 2 length
CD title 1 CD 3 length
Assumptions:
• Musicians in group contain a list of the members of
the people in the musical group.
• Musical groups can have more than one CD, so both
group name and CD title are needed to uniquely
identify a particular CD.
E. Jim Smith’s dealership sells Fords, Hondas, and
Toyotas. Th e dealership keeps information about each
car manufacturer with whom it deals so that the deal-
ership can get in touch with them easily. Th e dealer-
ship also keeps information about the models of cars
that it carries from each manufacturer. It keeps infor-
mation like list price, the price the dealership paid to
obtain the model, and the model name and series (e.g.,
Honda Civic LX). It also keeps information about all
sales that it has made (e.g., it records a buyer’s name,
the car bought, and the amount paid for the car). To
contact the buyers in the future, contact information
is also kept (e.g., address, phone number). Create a
class diagram for this situation. Apply the rules of nor-
malization to the class diagram to check the diagram
for processing effi ciency.
F. Describe how you would denormalize the model that
you created in exercise E. Draw the new class diagram
based on your suggested changes. How would perfor-
mance be aff ected by your suggestions?
G. Examine the model that you created in exercise F.
Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve the
performance of the database.
H. Calculate the size of the database that you created in
exercise F. Provide size estimates for the initial size of
22. Describe three situations that can be good candidates
for denormalization.
23. Describe several techniques that can improve perfor-
mance of a database.
24. What is the diff erence between interfi le and intrafi le
clustering? Why are they used?
25. What is an index and how can it improve the perfor-
mance of a system?
26. Describe what should be considered when estimating
the size of a database.
27. Why is it important to understand the initial and
projected size of a database during design?
28. What are some of the nonfunctional requirements that
can infl uence the design of the data management layer?
29. What are the key issues in deciding between using per-
fectly normalized databases and denormalized databases?
30. What is the primary purpose of the data access and
manipulation classes?
31. Why should the data access and manipulation classes
be dependent on the problem domain classes instead
of the other way around?
32. Why should the object persistence classes be depend-
ent on the problem domain classes instead of the other
way around?
Minicases 365
the database as well as for the database in one year’s
time. Assume that the dealership sells ten models of
cars from each manufacturer to approximately 20,000
customers a year. Th e system will be set up initially
with one year’s worth of data.
I. For the A Real Estate Inc. problem in Chapter 4
(exercises I, J, and K), Chapter 5 (exercises P and Q),
Chapter 6 (exercise D), Chapter 7 (exercise A), and
Chapter 8 (exercise A):
1. Apply the rules of normalization to the class diagram
to check the diagram for processing effi ciency.
2. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
J. For the A Video Store problem in Chapter 4 (exercises
L, M, and N), Chapter 5 (exercises R and S), Chapter
6 (exercise E), Chapter 7 (exercise B), and Chapter 8
(exercise B):
1. Apply the rules of normalization to the class
diagram to check the diagram for processing
effi ciency.
2. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
K. For the gym membership problem in Chapter 4 (exer-
cises O, P, and Q), Chapter 5 (exercises T and U),
Chapter 6 (exercise F), Chapter 7 (exercise C), and
Chapter 8 (exercise C):
1. Apply the rules of normalization to the class
diagram to check the diagram for processing
effi ciency.
2. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
L. For the Picnics R Us problem in Chapter 4 (exercises
R, S, and T), Chapter 5 (exercises V and W), Chapter
6 (exercise G), Chapter 7 (exercise D), and Chapter 8
(exercise D):
1. Apply the rules of normalization to the class diagram
to check the diagram for processing effi ciency.
2. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
M. For the Of-the-Month-Club problem in Chapter 4
(exercises U, V, and W), Chapter 5 (exercises X and
Y), Chapter 6 (exercise H), Chapter 7 (exercise E), and
Chapter 8 (exercise E):
1. Apply the rules of normalization to the class dia-
gram to check the diagram for processing effi ciency.
2. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
MINICASES
1. Th e system development team at the Wilcon Company
is working on developing a new customer order entry
system. In the process of designing the new system, the
team has identifi ed the following class and its attributes:
Inventory Order
Order Number (PK)
Order Date
Customer Name
Street Address
City
State
Zip
Customer Type
Initials
District Number
Region Number
1 to 22 occurrences of:
Item Name
Quantity Ordered
Item Unit
Quantity Shipped
Item Out
Quantity Received
a. State the rule that is applied to place a class in fi rst
normal form. Based on the above class, create a
class diagram that will be in 1NF.
b. State the rule that is applied to place a class into
second normal form. Revise the class diagram for
the Wilcon Company using the class and attributes
described (if necessary) to place it in 2NF.
c. State the rule that is applied to place a class into
third normal form. Revise the class diagram to
place it in 3NF.
d. When planning for the physical design of this data-
base, can you identify any likely situations where
the project team might choose to denormalize the
366 Chapter 9 Data Management Layer Design
class diagram? Aft er going through the work of
normalizing, why would this be considered?
2. In the new system under development for Holiday
Travel Vehicles, seven tables will be implemented in
the new relational database. Th ese tables are: New
Vehicle, Trade-in Vehicle, Sales Invoice, Customer,
Salesperson, Installed Option, and Option. Th e
expected average record size for these tables and the
initial record count per table are given here.
Average Initial Table
Table Name Record Size Size (records)
New Vehicle 65 characters 10,000
Trade-in Vehicle 48 characters 7,500
Sales Invoice 76 characters 16,000
Customer 61 characters 13,000
Salesperson 34 characters 100
Installed Option 16 characters 25,000
Option 28 characters 500
Perform a volumetrics analysis for the Holiday
Travel Vehicle system. Assume that the DBMS that
will be used to implement the system requires 35 per-
cent overhead to be factored into the estimates. Also,
assume a growth rate for the company of 10 percent
per year. Th e systems development team wants to
ensure that adequate hardware is obtained for the next
three years.
3. Refer to the Professional and Scientifi c Staff Manage-
ment (PSSM) minicase in Chapters 4, 6, 7, and 8.
a. Apply the rules of normalization to the class
diagram to check the diagram for processing
effi ciency.
b. Develop a clustering and indexing strategy for this
model. Describe how your strategy will improve
the performance of the database.
A user interface is the part of the system with which the users interact. From the user’s
point of view, the user interface is the system. It includes the screen displays that provide nav-
igation through the system, the screens and forms that capture data, and the reports that the
system produces (whether on paper, on the screen, or via some other medium). Th is chapter
introduces the basic principles and processes of interface design and discusses how to design
the interface structure and standards, navigation design, input design, and output design. Th e
chapter introduces the issues related to designing user interfaces for the mobile computing
environment and social media. It also introduces the issues that need to be considered when
designing user interfaces for a global audience. Finally, the chapter describes the eff ect of the
nonfunctional requirements on designing the human–computer interaction layer.
OBJECTIVES
■ Understand several fundamental user interface design principles.
■ Understand the process of user interface design.
■ Understand how to design the user interface structure.
■ Understand how to design the user interface standards.
■ Understand commonly used principles and techniques for navigation design.
■ Understand commonly used principles and techniques for input design.
■ Understand commonly used principles and techniques for output design.
■ Be able to design a user interface.
■ Understand the eff ect of nonfunctional requirements on the human–computer
interaction layer.
IINTRODUCTION
Interface design is the process of defi ning how a system will interact with external entities (e.g.,
customers, suppliers, other systems). In this chapter, we focus on the design of user interfaces,
but it is also important to remember that there are sometimes system interfaces, which exchange
information with other systems. System interfaces are typically designed as part of a systems
integration eff ort. Th ey are defi ned in general terms as part of the physical architecture and data
management layers. Th e human–computer interaction layer defi nes the way in which the users
interact with the system and the nature of the inputs and outputs that the system accepts and
produces.
Up until now, the entire development process has been focused on getting the problem
domain layer and its storage on the data management layer right. However, from the user’s point
of view, the user interface on the human–computer interaction layer is the system. Users do
not really care about how the problem domain objects are stored. But, they do care about how
367
C H A P T E R 1 0
Human–Computer Interaction
Layer Design
368 Chapter 10 Human–Computer Interaction Layer Design
they can use the system to support them in their activities. Based on our layered based design
approach, the user interface of the human–computer interaction layer is independent of the data
management layer. But it is dependent on both the problem domain and physical architecture
layers. Depending on the type of device that the human–computer interaction layer is deployed
on will set both opportunities and constraints as to what user interface features can be included.
For example, deploying the human computer interaction layer on both a smartphone and a
desktop computer will cause two diff erent user interfaces to be designed.
Even though there are command-line user interfaces (e.g., Terminal on Mac OSX), we are
only focusing on graphical user interfaces (GUI) that use windows, menus, icons, etc.1 Today,
GUI-based interfaces are the most common type of interfaces that we use.2 Regardless of the
underlying hardware being used, a GUI-based user interface comprises three fundamental
parts. Th e fi rst is the navigation mechanism, the way in which the user gives instructions to
the system and tells it what to do (e.g., buttons, menus). Th e second is the input mechanism,
the way in which the system captures information (e.g., forms for adding new customers). Th e
third is the output mechanism, the way in which the system provides information to the user or
to other systems (e.g., reports, Web pages). Each of these is conceptually diff erent, but they are
closely intertwined. All GUI-based displays contain navigation mechanisms, and most contain
input and output mechanisms. Th erefore, navigation design, input design, and output design are
tightly coupled and must be performed in an incremental and iterative manner.
In this chapter, even though we focus primarily on designing user interfaces that run in
a laptop or desktop type of environment, we also provide general guidelines for mobile com-
puting. We also address some of the unique issues you face when deploying the user interface
in social applications, such as FacebookTM and TwitterTM; in advanced technology interfaces,
such as 3D augmented and virtual reality applications; and fi nally, issues related to going
global with the user interface.
PRINCIPLES FOR USER INTERFACE DESIGN
In many ways, user interface design is an art. Th e goal is to make the interface pleasing to the
eye and simple to use while minimizing the eff ort the users need to accomplish their work.
Th e system is never an end in itself; it is merely a means to accomplish the business of the
organization.
We have found that the greatest problem facing experienced designers is using space eff ec-
tively. Simply put, oft en there is much more information that needs to be presented on a screen
or report or form than will fi t comfortably. Analysts must balance the need for simplicity and
pleasant appearance against the need to present the information across multiple pages or screens,
which decreases simplicity. In this section, we discuss some fundamental interface design princi-
ples, which are common for navigation design, input design, and output design3 (see Figure 10-1).
1 Many people attribute the origin of GUI interfaces to Apple or Microsoft . Some people know that Microsoft copied
from Apple, which, in turn, “borrowed” the whole idea from a system developed at the Xerox Palo Alto Research
Center (PARC) in the 1970s. Very few know that the Xerox system was based on a system developed by Doug
E nglebart of Stanford that was fi rst demonstrated at the Western Computer Conference in 1968. Around the same
time, he also invented the mouse, desktop video conferencing, groupware, and a host of other things we now take for
granted. Doug is a legend in the computer science community and has won too many awards to count but is relatively
unknown by the general public.
2 A set of good books on GUI design include Jennifer Tidwell, Designing Interfaces, 2nd Ed. (Sebastopol, CA: O’Reilly
Media, 2010); Ben Shneiderman, Designing the User Interface: Strategies for Eff ective Human–Computer Interaction,
3rd Ed. (Reading, MA: Addison-Wesley, 1998); Alan Cooper, About Face 3: Th e Essentials of Interaction Design
(Indianapolis, IN: Wiley, 2007).
3 A good book on the design of interfaces is Susan Weinschenk, Pamela Jamar, and Sarah Yeo, GUI Design Essentials
(New York: Wiley, 1997).
Principles for User Interface Design 369
Layout
Th e fi rst element of design is the basic layout of the screen, form, or report. Most soft ware
designed for personal computers follows the standard Windows or Macintosh approach
for screen design. Th e screen is divided into three boxes. Th e top box is the navigation
area, through which the user issues commands to navigate through the system. Th e bot-
tom box is the status area, which displays information about what the user is doing. Th e
middle—and largest—box is used to display reports and present forms for data entry.
Th is use of multiple layout areas also applies to inputs and outputs. Data areas on reports
and forms are oft en subdivided into subareas, each of which is used for a diff erent type of
information. Th ese areas are almost always rectangular, although sometimes space constraints
require odd shapes. Nonetheless, the margins on the edges of the screen should be consistent.
Each of the areas within the report or form is designed to hold diff erent information. For exam-
ple, on an order form (or order report), one part may be used for customer information (e.g.,
name, address), one part for information about the order in general (e.g., date, payment infor-
mation), and one part for the order details (e.g., how many units of which items at what price
each). Each area is self-contained so that information in one area does not run into another.
Th e areas and information within areas should have a natural intuitive fl ow to minimize
the users’ movement from one area to the next. People in Westernized nations (e.g., United
States, Canada, Mexico) tend to read left -to-right, top-to-bottom, so related information
should be placed so that it is used in this order (e.g., address lines, followed by city, state or
province, and then ZIP code or postal code). Sometimes the sequence is in chronological
order, or from the general to the specifi c, or from most frequently to least frequently used.
In any event, before the areas are placed on a form or report, the analyst should have a clear
understanding of what arrangement makes the most sense for how the form or report will
be used. Th e fl ow between sections should also be consistent, whether horizontal or vertical.
Ideally, the areas will remain consistent in size, shape, and placement for the forms used to
enter information (whether paper or on screen) and the reports used to present it.
Content Awareness
Content awareness refers to the ability of an interface to make the user aware of the informa-
tion it contains with the least amount of eff ort on the user’s part. All parts of the interface,
Layout The interface should be a series of areas on the screen that are used
consistently for different purposes—for example, a top area for commands and
navigation, a middle area for information to be input or output, and a bottom
area for status information.
Content Awareness Users should always be aware of where they are in the system and what
information is being displayed.
Aesthetics Interfaces should be functional and inviting to users through careful use of
white space, colors, and fonts. There is often a trade-off between including
enough white space to make the interface look pleasing without losing so much
space that important information does not fi t on the screen.
User Experience Although ease of use and ease of learning often lead to similar design decisions,
sometimes there is a trade-off between the two. Novice or infrequent users of
software prefer ease of learning, whereas frequent users prefer ease of use.
Consistency Consistency in interface design enables users to predict what will happen
before they perform a function. It is one of the most important elements in ease
of learning, ease of use, and aesthetics.
Minimal User Effort The interface should be simple to use. Most designers plan on having no
more than three mouse clicks from the starting menu until users perform work.
Principle Description
FIGURE 10-1
Principles of User
Interface Design
370 Chapter 10 Human–Computer Interaction Layer Design
whether navigation, input, or output, should provide as much content awareness as possi-
ble, but it is particularly important for forms or reports that are used quickly or irregularly.
Content awareness applies to the interface in general. All interfaces should have titles (on the
screen frame, for example). Menus should show where the user is and, if possible, where the
user came from to get there.
Content awareness also applies to the areas within forms and reports. All areas should be
clear and well-defi ned so that it is diffi cult for the user to become confused about the infor-
mation in any area. Th en users can quickly locate the part of the form or report that is likely
to contain the information they need. Sometimes the areas are marked by lines, colors, or
headings; in other cases, the areas are only implied.
Content awareness also applies to the fi elds within each area. Fields are the individual
elements of data that are input or output. Th e fi eld labels that identify the fi elds on the
interface should be short and specifi c—objectives that oft en confl ict. Th ere should be no
uncertainty about the format of information within fi elds, whether for entry or display.
For example, a date of 10/5/15 is diff erent depending on whether you are in the United
States (October 5, 2015) or in Canada (May 10, 2015). Any fi elds for which there is the
possibility of uncertainty or multiple interpretations should provide explicit explanations.
Content awareness also applies to the information that a form or report contains. In general,
all forms and reports should contain a preparation date (i.e., the date printed or the date com-
pleted) so that the age of information is obvious. Likewise, all printed forms and soft ware should
provide version numbers so that users, analysts, and programmers can identify outdated materials.
Aesthetics
Aesthetics refers to designing interfaces that are pleasing to the eye. Interfaces do not have to
be works of art, but they do need to be functional and inviting to use. In most cases, less is
more, meaning that a simple, minimalist design is the best.
Space is usually at a premium on forms and reports, and oft en there is the temp tation
to squeeze as much information as possible onto a page or a screen. Unfortunately, this can
make a form or report so unpleasant that users do not want to use it. In general, all forms
and reports need a minimum amount of white space that is intentionally left blank.
In general, novice or infrequent users of an interface, whether on a screen or on paper,
prefer interfaces with low density, oft en one with a density of less than 50 percent (i.e., less than
50 percent of the interface occupied by information). More-experienced users prefer higher
densities, sometimes approaching 90 percent occupied, because they know where information
is located and high densities reduce the amount of physical movement through the interface.
Th e design of text is equally important. As a general rule, all text should be in the same font
and about the same size. Fonts should be no smaller than 8 points, but 10 points is oft en pre-
ferred, particularly if the interface will be used by older people. Changes in font and size are used
to indicate changes in the type of information that is presented (e.g., headings, status indicators).
In general, italics and underlining should be avoided because they make text harder to read.
Serif fonts (i.e., those having letters with serifs, or tails, such as Times Roman) are the most
readable for printed reports, particularly for small letters. Sans serif fonts (i.e., those without
serifs, such as Helvetica or Arial) are the most readable for computer screens and are oft en used
for headings in printed reports. Never use all capital letters, except possibly for titles.
Color and patterns should be used carefully and sparingly and only when they serve a
purpose. (About 10 percent of men are color blind, so the improper use of color can impair
their ability to read color text.) A quick trip around the Web will demonstrate the prob-
lems caused by indiscriminate use of colors and patterns. Remember, the goal is pleasant
readability, not art; color and patterns should be used to strengthen the message, not over-
whelm it. Color is best used to separate and categorize items, such as showing the diff erence
Principles for User Interface Design 371
between headings and regular text, or to highlight important information. Th erefore, colors
with high contrast should be used (e.g., black and white). In general, black text on a white
background is the most readable, and blue on red is the least readable. Also, when it comes
to the proper use of color, cultural issues come into play. We discuss this later in the chapter.
User Experience
User experience can essentially be broken down into two levels: those with experience and
those without. Interfaces should be designed for both types of users. Novice users usually are
most concerned with ease of learning—how quickly they can learn new systems. Expert users
are usually most concerned with ease of use—how quickly they can use the system once they
have learned how to use it. Oft en these two are complementary and lead to similar design
decisions, but sometimes there are trade-off s. Novices, for example, oft en prefer menus that
show all available system functions, because these promote ease of learning. Experts, on
the other hand, sometimes prefer fewer menus organized around the most commonly used
functions.
Systems that will end up being used by many people on a daily basis are more likely to
have a majority of expert users (e.g., order-entry systems). Although interfaces should try to
balance ease of use and ease of learning, these types of systems should put more emphasis on
ease of use rather than ease of learning. Users should be able to access the commonly used
functions quickly, with few keystrokes or a small number of menu selections.
In many other systems (e.g., decision-support systems), most people remain occasional
users for the lifetime of the system. In this case, greater emphasis may be placed on ease of
learning rather than ease of use.
Ease of use and ease of learning oft en go hand-in-hand—but sometimes they don’t.
Research shows that expert and novice users have diff erent requirements and behavior pat-
terns in some cases. For example, novices virtually never look at the bottom area of a screen
that presents status information, whereas experts refer to the status bar when they need
information. Most systems should be designed to support frequent users, except for systems
designed to be used infrequently or when many new users or occasional users are expected.
Likewise, systems that contain functionality that is used only occasionally must contain a
highly intuitive interface or an interface that contains clear, explicit guidance regarding its
use. Th e balance of quick access to commonly used and well-known functions and guidance
through new and less-well-known functions is challenging to the interface designer, and this
balance oft en requires elegant solutions.
Consistency
Consistency in design is probably the single most important factor in making a system simple to
use because it enables users to predict what will happen. When interfaces are consistent, users
can interact with one part of the system and then know how to interact with the rest, aside from
elements unique to those parts. Consistency usually refers to the interface within one computer
system, so that all parts of the same system work in the same way. Ideally, the system should
also be consistent with other computer systems in the organization and with commercial
soft ware that is used. Many soft ware development tools support consistent system interfaces
by providing standard interface objects (e.g., list boxes, pull-down menus, and radio buttons).
Consistency occurs at many diff erent levels. Consistency in the navigation controls con-
veys how actions in the system should be performed. For example, using the same icon or
command to change an item clearly communicates how changes are made throughout the
system. Consistency in terminology is also important. Th is refers to using the same words
for elements on forms and reports (e.g., not customer in one place and client in another).
We also believe that consistency in report and form design is important, although a study
372 Chapter 10 Human–Computer Interaction Layer Design
suggests that being too consistent can cause problems.4 When reports and forms are very
similar except for very minor changes in titles, users sometimes mistakenly use the wrong
form and either enter incorrect data or misinterpret its information. Th e implication for
design is to make the reports and forms similar but give them some distinctive elements
(e.g., color, size of titles) that enable users to immediately detect diff erences.
Minimizing User Eff ort
Interfaces should be designed to minimize the amount of eff ort needed to accomplish tasks.
Th is means using the fewest possible mouse clicks or keystrokes to move from one part of the
system to another. Most interface designers follow the three-clicks rule: Users should be able
to go from the start or main menu of a system to the information or action they want in no
more than three mouse clicks or three keystrokes. However, with regard to this point, you
need to be aware of Krug’s principles (discussed later).
USER INTERFACE DESIGN PROCESS
User interface design is a use-case driven, incremental, and iterative process. Analysts oft en
move back and forth between the diff erent parts (navigation, input, and output) of the user inter-
face, rather than proceeding sequentially from one part to another part. Given that the design
process is use case driven, the analysts begin the user interface design process by examining the
use cases (see Chapter 4) and their associated sequence diagrams (see Chapter 6) developed in
analysis. Analysts then typically set down with users to develop use scenarios that describe com-
monly employed patterns of actions the users will perform so that the interface enables users
to quickly and smoothly perform these scenarios. In some cases, additional requirements are
uncovered. Depending on the importance of the newly uncovered requirements, this can cause
the problem domain layer to be modifi ed, which in turn can cause the data management layer
to be modifi ed. However, many times, these new requirements can be delayed until the next
iteration of the system. With agile approaches, user interface design and requirements mode-
ling is so intertwined that new requirements are uncovered on a regular basis. Consequently,
depending on the stability of the modeling of the problem domain, user interface design could
occur concurrently with functional modeling. Even though functional and behavioral modeling
is associated with the analysis workfl ow and user interface design is associated with the design
workfl ow, the level of activity associated with the two workfl ows overlaps (see Figures 1-15 and
1-16). As such, performing user interface design along side of functional and behavioral mode-
ling is compatible with both the Unifi ed Process and the Enhanced Unifi ed Process.
Once a basic set of use scenarios have been developed, the actual user interface is designed.
As we stated earlier, all GUI-based user interfaces comprise three parts: navigation, input, and
output. To some degree, all three parts tend to be designed together. Consequently, the user
interface design process tends to follow a prototyping style of development (see Chapter 1)
wherein the analyst and user will incrementally build a design by iterating across all three parts
of the user interface using diff erent design tools. For example, when designing the structure
of the navigation, a windows navigation diagram (WND) is very useful; when designing the
layout of the user interface, a windows layout diagram is very useful; and when attempting to
try and tie the navigation, input, and output designs together, storyboards, and user interface
prototypes are very useful. Another useful idea when developing a user interface is to have a
set of accepted interface standards that can be used across multiple applications. For example,
a standard set of menus, icons, and user interface templates simplify the entire design of the
human computer interaction layer. Once the basic design has been completed for a specifi c
use case, then the essential use case developed in functional modeling should be converted to
4 John Satzinger and Lorne Olfman, “User Interface Consistency Across End-User Application: Th e Eff ects of Mental
Models,” Journal of Management Information Systems (Spring 1998): 167–193.
User Interface Design Process 373
a real use case that, along with the other tools used to design the interface, can be used as a
basis for documentation, training, and testing. Finally, the individual interfaces are subjected
to interface evaluation to determine if they are satisfactory and how they can be improved.
Interface evaluations almost always identify improvements, so the interface design pro-
cess is repeated in a cyclical process until no new improvements are identifi ed. In practice,
most analysts interact closely with the users during the interface design process so that users
have many chances to see the interface as it evolves, rather than waiting for one overall inter-
face evaluation at the end of the interface design process. It is better for all concerned (both
analysts and users) if changes are identifi ed sooner rather than later. For example, if the inter-
face structure or standards need improvements, it is better to identify changes before most of
the screens that use the standards have been designed.5
Use Scenario Development
A use scenario is an outline of the steps that the users perform to accomplish some part of their
work. A use scenario is one path through an essential use case. For example, Figure 10-2 shows
Appointment System
Patient
New Patient
Old Patient
Update Patient
Information
Make Old Patient
Appt
Make New Patient
Appt
Make Payment
Arrangements
Create New Patient
Manage
Appointments
<
>
<
>
*
*
*
*
<>
<>
Produce Schedules
Management
Doctor
Record
Availability
Manage
Schedule
<>
<>
*
*
*
*
FIGURE 10-2 Appointment System Use-Case Diagram (see Figures 4-21 and 7-11)
5 A good source for more information on user interface evaluation is Deborah Hix and H. Rex Hartson, Developing
User Interfaces, Ensuring Usability Th rough Product & Process (New York: Wiley, 1993).
374 Chapter 10 Human–Computer Interaction Layer Design
the use-case diagram for the Appointment System. Th is fi gure shows that the Create New Patient
use case is distinct from the Make Payment Arrangements use case. We model these two use
cases separately because they represent separate processes that are used by the Make New Patient
Appt use case.
Th e use-case diagram was designed to model all possible uses of the system—its
complete functionality or all possible paths through the use case at a fairly high level of
abstraction. In one use scenario, a patient makes a request with the receptionist regarding
an appointment with the doctor. Th e receptionist looks up the patient and checks to see
if the patient has any bills to be paid. Th e receptionist then asks the patient whether he
or she wants to set up a new appointment, cancel an existing appointment, or change an
existing appointment. If the patient wants to make a new appointment, the receptionist
asks the patient for some suggested appointment times, which the receptionist matches
against potential times available. Th e receptionist fi nally creates a new appointment (see
Figures 6-1 and 6-10).
In another use scenario, a patient simply wants to cancel an appointment. In this case,
the receptionist looks up the patient and checks to see if the patient has any bills to be paid.
Th e receptionist then asks the patient for the time of the appointment to be canceled. Finally,
the receptionist deletes the appointment.
Use scenarios are presented in a simple narrative description that is tied to the essential
use cases developed during analysis (see Chapter 4). Figure 10-3 shows the two use scenarios
just described. Th e key point with using use cases for interface design is not to document all
possible use scenarios within a use case. Th e goal is to document two or three of the most
common use scenarios so that the interface can be designed to enable the most common uses
to be performed simply and easily.
Use scenario: Existing Patient Cancels
Appointment
1. Patient requests appointment (1) and gives
the receptionist his or her name and address (2).
2. The receptionist looks up the patient (3)
and determines whether the patient has
changed any information (3 & 4).
3. The receptionist then asks the patient
whether he or she is going to set up a new
appointment, change an appointment, or
delete an appointment (5).
4. The receptionist asks the patient for the
appointment time to be canceled (S-2, 1).
5. The receptionist finds and deletes the
appointment (S-2, 2).
6. The receptionist informs the patient that
his or her appointment time was canceled (6).
Use scenario: Existing Patient Makes
New Appointment
1. Patient requests appointment (1) and gives
the receptionist his or her name and address (2).
The numbers in parentheses refer to specific events in the essential use case.
2. The receptionist looks up the patient (3)
and determines whether the patient has
changed any information (3 & 4).
3. The receptionist then asks the patient
whether he or she is going to set up a new
appointment, change an appointment, or
delete an appointment (5).
4. The receptionist asks the patient for a list
of potential appointment times (S-1, 1).
5. The receptionist matches the potential
appointment times with the available
times and schedules the appointment
(S-1, 2).
6. The receptionist informs the patient of
his or her appointment time (6).
FIGURE 10-3 Use Scenarios
User Interface Design Process 375
Navigation Structure Design
Th e navigation structure defi nes the basic components of the interface and how they work
together to provide functionality to users. A windows navigation diagram (WND)6 is used to
show how all the screens, forms, and reports used by the system are related and how the user
moves from one to another. Most systems have several WNDs, one for each major part of
the system.
A WND is very similar to a behavioral state machine (see Chapter 6), in that they both
model state changes. A behavioral state machine typically models the state changes of an
object, whereas a WND models the state changes of the user interface. In a WND, each state
of the user interface is represented as a box. A box typically corresponds to a user interface
component, such as a window, form, button, or report. For example, in Figure 10-4, there are
fi ve separate states: Client Menu, Find Client Form, Add Client Form, Client List, and Client
Information Report.
Transitions are modeled as either a single-headed or double-headed arrow. A single-
headed arrow indicates that a return to the calling state is not required, whereas a double-
headed arrow represents a required return. For example in Figure 10-4, the transition from
the Client Menu state to the Find Client Form state does not require a return. Th e arrows are
labeled with the action that causes the user interface to move from one state to another. For
example, in Figure 10-4, to move from the Client Menu state to the Find Client Form state,
the user must click the Find Client Button on the Client Menu.
Th e last item to be described in a WND is the stereotype. A stereotype is modeled as a
text item enclosed within guillemets or angle brackets (<>). Th e stereotype represents the
type of user interface component of a box on the diagram. For example, the Client Menu is a
window, whereas Find Client Form is a form.
Th e basic navigation structure of an interface follows the basic structure of the business
process itself, as defi ned in the use cases and behavioral model. Th e analyst starts with the
essential use cases and develops the fundamental fl ow of control of the system as it moves
from object to object. Th e analyst then examines the use scenarios to see how well the WND
6 A WND is based on the behavioral state machine and object diagrams [see Meilir Page-Jones, Fundamentals of
Object-Oriented Design in UML (New York: Dorset House, 2000)].
Click Find Client
Button
Click Find
Client Button
Cl
ic
k
Fi
nd
Cl
ie
nt
B
ut
to
n
Click List
Clients
Button
<>
Client Information Report
<>
Client List
Click Add ClientButton
<>
Add Client
Form
<
<>
Find Client
Form
<
<>
Client Menu
<
<
Click
Hyperlink
FIGURE 10-4 Sample WND
376 Chapter 10 Human–Computer Interaction Layer Design
supports them. Quite oft en, the use scenarios identify paths through the WND that are more
complicated than they should be. Th e analyst then reworks the WND to simplify the ability of
the interface to support the use scenarios, sometimes by making major changes to the menu
structure, sometimes by adding shortcuts.
Interface Standards Design
Interface standards are the basic design elements that are common across the individual
screens, forms, and reports within the system. Depending on the application, there may be
several sets of interface standards for diff erent parts of the system (e.g., one for Web screens,
one for mobile screens, one for paper reports, one for input forms). For example, the part
of the system used by data entry operators might mirror other data entry applications in the
company, whereas a Web interface for displaying information from the same system might
adhere to some standardized Web format. Likewise, each individual interface might not
contain all of the elements in the standards (e.g., a report screen might not have an edit capa-
bility), and they might contain additional characteristics beyond the standard ones, but the
standards serve as the touchstone that ensures the interfaces are consistent across the system.
Th e following sections discuss some of the main areas in which interface standards should be
considered: metaphors, objects, actions, icons, and templates.
Interface Metaphor First of all, the analysts must develop the fundamental interface
metaphor(s) that defi nes how the interface will work. An interface metaphor is a concept
from the real world that is used as a model for the computer system. Th e metaphor helps the
user understand the system and enables the user to predict what features the interface might
provide, even without actually using the system. Sometimes systems have one metaphor,
whereas in other cases there are several metaphors in diff erent parts of the system.
Oft en, the metaphor is explicit. Quicken, for example, uses a checkbook metaphor for
its interface, even to the point of having the users type information into an on-screen form
that looks like a real check. In other cases, the metaphor is implicit or unstated, but it is there,
nonetheless. Many Windows systems use the paper form or table as a metaphor.
In some cases, the metaphor is so obvious that it requires no thought. For example, most
online stores use a shopping cart metaphor to temporarily store the items that the customer
is considering purchasing. In other cases, a metaphor is hard to identify. In general, it is
better not to force a metaphor that really doesn’t fi t a system, because an ill-fi tting meta-
phor will confuse users by promoting incorrect assumptions.
Interface Templates An interface template defi nes the general appearance of all screens in
the information system and the paper-based forms and reports that are used. Th e template
design, for example, specifi es the basic layout of the screens (e.g., where the navigation
area(s), status area, and form/report area(s) will be placed) and the color scheme(s) that
will be applied. It defi nes whether windows will replace one another on the screen or will
cascade over the top of each other. Th e template defi nes a standard placement and order
for common interface actions (e.g., File Edit View rather than File View Edit). In short,
the template draws together the other major interface design elements: metaphors, objects,
actions, and icons.
Interface Objects Th e template specifi es the names that the interface will use for the major
interface objects, the fundamental building blocks of the system, such as the classes. In many
cases, the object names are straightforward, such as calling the shopping cart the “shopping
cart.” In other cases, it is not so simple. For example, Amazon.com sells much more than
books. In some cases, the user might not know whether he or she is looking for a book,
User Interface Design Process 377
CD, DVD, or Kindle download. In those cases, the user can use a catchall search item: All
Departments. In the case that the user knows the type of item that he or she wants to buy,
the user can limit the search by specifying more-specifi c types of search items, such as Apps
for Android, Books, Kindle Store, or Music. Obviously, the object names should be easily
understood and help promote the interface metaphor.
In general, in cases of disagreements between the users and the analysts over names,
whether for objects or actions (discussed later), the users should win. A more understandable
name always beats a more precise or more accurate one.
Interface Actions Th e template also specifi es the navigation and command language style
(e.g., menus) and grammar (e.g., object-action order; see the navigation design section later
in this chapter). It gives names to the most commonly used interface actions in the navigation
design (e.g., buy versus purchase or modify versus change).
Interface Icons Th e interface objects and actions and their status (e.g., deleted or overdrawn)
may be represented by interface icons. Icons are pictures that appear on command buttons as
well as in reports and forms to highlight important information. Icon design is very challenging
because it means developing a simple picture less than half the size of a postage stamp that needs
to convey an oft en-complex meaning. Th e simplest and best approach is to simply adopt icons
developed by others (e.g., a blank page to indicate create a new fi le, a diskette to indicate save).
Th is has the advantage of quick icon development, and the icons might already be well under-
stood by users because they have seen them in other soft ware.
Commands are actions that are especially diffi cult to represent with icons because they are
in motion, not static. Many icons have become well known from widespread use, but icons are
not as well understood as fi rst believed. Use of icons can sometimes cause more confusion than
insight. Icon meanings become clearer with use, but sometimes a picture is not worth even one
word; when in doubt, use a word, not a picture.
Interface Design Prototyping
An interface design prototype is a mock-up or a simulation of a computer screen, form, or
report. A prototype is prepared for each interface in the system to show the users and the
programmers how the system will perform. In the “old days,” an interface design prototype
was usually specifi ed on a paper form that showed what would be displayed on each part of
the screen. Paper forms are still used today, but more and more interface design prototypes
are being built using computer tools instead of paper. Th e four most common approaches
to interface design prototyping are storyboards, windows layout diagrams, and language
prototypes.
Windows Layout Diagram A windows layout diagram is simply a picture that resembles the
actual user interface that the user will gradually receive. Typically, it is created using a tool such
as Microsoft ’s Visio. Using this type of tool, the designer can quickly drag and drop the user
interface components onto the canvas to lay out the design of the user interface. For example,
Figure 10-5 portrays a simple windows layout diagram. Even though there is no executable
capability associated with a windows layout diagram, it does allow the user to quickly get a feel
for the look of the user interface that will be delivered.
Storyboard At its simplest, an interface design prototype is a paper-based storyboard. Th e
storyboard shows hand-drawn pictures of what the screens will look like and how they fl ow
from one screen to another, in the same way a storyboard for a cartoon shows how the action
will fl ow from one scene to the next (see Figure 10-6). Storyboards are the simplest technique
378 Chapter 10 Human–Computer Interaction Layer Design
because all they require is paper (oft en a fl ip chart) and a pen—and someone with some artistic
ability. Storyboards also combine both the navigation information of the windows navigation
diagram and to some degree the layout information of the windows layout diagram. However,
with today’s graphics tools, the designer can work eff ectively with a set of users to design both
FIGURE 10-5
Sample Windows
Layout Diagram
Add a Client
First Name: ____________ Last Name: ____________
Address: _______________________________
_______________________________
City: __________________________________
State: ________ Zip Code: ________
Client Menu
Add Client
Find Client
List Clients
Client Information
First Name: Pat Last Name: Smith
Address: 1234 Anywhere St.
Apt 56
City: Somethingville
State: CA Zip Code: 90211
Find a Client
(Type in information to search on)
First Name: ____________ Last Name: ____________
Address: _______________________________
_______________________________
City: __________________________________
State: ________ Zip Code: ________
Client List
(Click on a client for more information)
Adams, Clare
Adams, John
Baker, Robin
FIGURE 10-6 Sample Storyboard
User Interface Design Process 379
the look and feel of the evolving system without having to actually implement anything, by com-
bining the windows layout diagrams with the windows navigation diagram into a single better
storyboard type of diagram (see Figure 10-7).
User Interface Prototypes With today’s programming environments, such as Visual Studio
and NetBeans, it is fairly easy to develop executable prototypes (see Figure 10-8) of the user
interface that would allow the user to be able to interact with the user interface by clicking
on buttons and entering pretend data into forms (but because there is no system behind
the pages, the data are never processed). Th e diff erent parts of the user interface are linked
together so that as the user clicks on buttons, the requested part of the system appears. Th ese
executable prototypes take longer to develop than windows navigation diagrams, windows
layout diagrams, and storyboards but have the distinct advantage of showing exactly what
the screens will look like. Th e user does not have to guess about the shape or position of the
elements on the screen. However, one of the potential issues that can arise when developing
user interface prototypes is that the user’s expectations of when the systems will be com-
pleted can become unrealistic. To actually connect the prototype up to the problem domain
such that the system actually works is not a trivial problem. So, user expectations need to be
carefully managed. Otherwise, a system that meets all of its specifi cations could end up being
considered a failure.
Selecting the Appropriate Techniques Projects oft en use a combination of diff erent inter-
face design prototyping techniques for diff erent parts of the system. Storyboarding is the
fastest and least expensive but provides the least amount of detail. Windows layout diagrams
provide more of a feel that the user will experience, while remaining fairly inexpensive
to develop. User interface prototypes are the slowest, most expensive, and most detailed
approach. Th erefore, storyboarding is used for parts of the system in which the interface
is well understood and when more-expensive prototypes are thought to be unnecessary.
However, in most cases it is probably worth the additional cost of developing windows lay-
out diagrams in addition to storyboards. User interface prototypes are used for parts of the
system that are critical, yet not well understood.
FIGURE 10-7
Sample Combined
Windows Naviga-
tion and Layout
Diagrams
380 Chapter 10 Human–Computer Interaction Layer Design
FIGURE 10-8
Sample User
Interface Prototype
Interface Evaluation7
Th e objective of interface evaluation is to understand how to improve the interface design
before the system is complete. Most interface designers intentionally or unintentionally
design an interface that meets their personal preferences, which might or might not match
7 Verifying and validation approaches, in general, were described in Chapters 4 through 7. Also, further approaches
to testing the evolving system are described in Chapter 12. In this section, we describe approaches that have been
customized to the human–computer interaction layer.
User Interface Design Process 381
the preferences of the users. Th e key message, therefore, is to have as many people as possible
evaluate the interface, and the more users the better. Most experts recommend involving at
least ten potential users in the evaluation process.
Many organizations save interface evaluation for the very last step in the systems
development before the system is installed. Ideally, however, interface evaluation should be
performed while the system is being designed—before it is built—so that any major design
problems can be identifi ed and corrected before the time and cost of programming have been
spent on a weak design. It is not uncommon for the system to undergo one or two major
changes aft er the users see the fi rst interface design prototype because they identify problems
that are overlooked by the project team.
As with interface design prototyping, interface evaluation can take many diff erent forms,
each with diff erent costs and diff erent amounts of detail. Four common approaches are heu-
ristic evaluation, walkthrough evaluation, interactive evaluation, and formal usability testing.
As with interface design prototyping, the diff erent parts of a system can be evaluated using
diff erent techniques.
Heuristic Evaluation A heuristic evaluation examines the interface by comparing it to a set
of heuristics or principles for interface design. Th e project team develops a checklist of inter-
face design principles—from the list at the start of this chapter, for example, as well as the list
of principles in the navigation, input, and output design sections later in this chapter. At least
three members of the project team then individually work through the interface design pro-
totype, examining every interface to ensure that it satisfi es each design principle on a formal
checklist. Aft er each has gone through the prototype separately, they meet as a team to discuss
their evaluations and identify specifi c improvements that are required.
Walkthrough Evaluation An interface design walkthrough evaluation is a meeting con-
ducted with the users who ultimately have to operate the system. Th e project team presents
the prototype to the users and walks them through the various parts of the interface. Th e
project team shows the storyboard and windows layout diagrams or actually demonstrates
the user interface prototype and explains how the interface will be used. Th e users identify
improvements to each of the interfaces that are presented.
Interactive Evaluation With an interactive evaluation, the users themselves actually work
with the user interface prototype in a one-person session with member(s) of the project team
(an interactive evaluation cannot be used with a storyboard or windows layout diagrams). As
the user works with the prototype (oft en by going through the use scenarios, using the real
use cases described later in this chapter, or just navigating at will through the system), he or
she tells the project team member(s) what he or she likes and doesn’t like and what additional
information or functionality is needed. As the user interacts with the prototype, team mem-
ber(s) records the cases when he or she appears to be unsure of what to do, makes mistakes, or
misinterprets the meaning of an interface component. If the pattern of uncertainty, mistakes,
or misinterpretations reoccurs across several of the users participating in the evaluation, it is
a clear indication that those parts of the interface need improvement.
Formal Usability Testing Formal usability testing is commonly done with commercial soft –
ware products and products developed by large organizations that will be widely used through
the organization. As the name suggests, it is a very formal—almost scientifi c— process that can
be used only with language prototypes (and systems that have been completely built awaiting
installation or shipping).8 As with interactive evaluation, usability testing is done in one-person
8 A good source for usability testing is Jakob Nielsen and Robert Mack (eds.), Usability Inspection Methods (New York:
Wiley, 1994). See also www.useit.com/papers.
382 Chapter 10 Human–Computer Interaction Layer Design
sessions in which a user works directly with the soft ware. However, it is typically done in a
special lab equipped with video cameras and special soft ware that records every keystroke and
mouse operation so that they can be replayed to understand exactly what the user did.
Th e user is given a specifi c set of tasks to accomplish (usually the use scenarios), and aft er
some initial instructions, the project team’s members are not permitted to interact with the
user to provide assistance. Th e user must work with the soft ware without help, which can be
hard on the users if they become confused with the system. It is critical that users understand
that the goal is to test the interface, not their abilities, and if they are unable to complete the
task, the interface—not the user—has failed the test.
Formal usability testing is very expensive, because each one-user session can take one
to two days to analyze depending on the volume of detail collected in the computer logs and
videos. Sessions typically last one to two hours. Most usability testing involves fi ve to ten
users, because if there are fewer than fi ve users, the results depend too much on the specifi c
individual users who participated, and more than ten users are oft en too expensive to justify
(unless a large commercial soft ware developer is involved).
Common Sense Approach to User Interface Design
When you consider all of the above material, creating an eff ective user interface design
can be a daunting and very time-consuming task. An interesting book by Steve Krug,9 how-
ever, provides us with a set of guiding principles for Web usability. In this section, we adapt
his principles to general user interface design.
First, the user should never have to think about how to navigate the user interface. As Krug
puts it, “Don’t make me think.” Cognitively speaking, any time the user has to stop and fi gure
out how to use the user interface, the creator of the user interface has failed. Th at might seem
a little harsh, but it is true. From the user’s perspective, the user interface is the system. If the
developers have done their homework, the user interface should be intuitive to use. From a
practical perspective, we should study how the user really uses the system. Based on Krug’s
observations of users, he found that users do not read Web pages; instead, they tend to scan
them. As a general user interface design guideline, we suggest that you make it easy for
users to identify the diff erent parts of the user interface so that they simply scan the screen
to see the section of the interface that is applicable to the problem that they are solving.
Given the user’s tendency to simply scan the user interface, Krug suggests that we should
consider studying billboards for inspiration. Billboards are designed to be “read” at 70 mph
as you drive down the highway. Obviously, the most relevant information must catch your
attention for the billboard advertisement to work. He suggests that we should use the set of
conventions with which we are familiar. For example, when looking at a newspaper you
know that it is organized into diff erent sections. In the case of the Wall Street Journal, you
know that the front page acts as an index into the rest of the paper. Consequently, we should
look for conventions that we can employ to aid the user.
Second, he suggests that the number of clicks that a user must perform to complete the
task is somewhat irrelevant. Instead, building on his fi rst guiding principle, the important
thing is to design the user interface such that the choices (clicks) to be made are unambigu-
ous. Making a lot of obvious choices is a lot quicker and easier than a few vague and ambig-
uous ones. Consequently, don’t worry about the number of screens that the user must work
through. However, like any other rule, this can be taken to an extreme. Too many clicks is
still too many clicks. Th e overall goal is to minimize the user’s eff ort. Simply focus on making
it easier for the user to complete the task.
9 Steve Krug, Don’t Make Me Th ink: A Common Sense Approach to Web Usability, 2nd Ed. (Berkeley, CA: 2nd Ed.
New Riders, 2006).
Navigation Design 383
Th ird, minimize the number of words on the screen. Given that users scan the screen to
fi nd for what they are searching, make it easier by not cluttering the screen with lots of noise.
He suggests that in the case of Web interfaces, 50 percent to 75 percent of the words can be
eliminated without losing any information contained on the screen. Obviously, this may be
somewhat extreme, but it does suggest that following the KISS10 principle is critical when
designing eff ective user interfaces.
NAVIGATION DESIGN
Th e navigation component of the interface enables the user to enter commands to navigate
through the system and perform actions to enter and review information it contains. Th e
navigation component also presents messages to the user about the success or failure of his
or her actions. Th e goal of the navigation system is to make the system as simple as possible
to use. A good navigation component is one the user never really notices. It simply functions
the way the user expects, and thus the user gives it little thought. In other words, keep Krug’s
three guiding principles in mind as you work through the next three sections of the text.
Basic Principles
One of the hardest things about using a computer system is learning how to manipulate
the navigation controls to make the system do what you want. Analysts should assume that
users have not read the manual, have not attended training, and do not have external help
readily at hand. All controls should be clear and understandable and placed in an intuitive
location on the screen. Ideally, the controls should anticipate what the user will do and
simplify his or her eff orts. For example, many setup programs are designed so that for a
typical installation, the user can simply keep pressing the Next button.
Prevent Mistakes Th e fi rst principle of designing navigation controls is to prevent the user
from making mistakes. A mistake costs time and causes frustration. Worse still, a series of
mistakes can cause the user to discard the system. Mistakes can be reduced by labeling com-
mands and actions appropriately and by limiting choices. Too many choices can confuse the
user, particularly when the choices are similar and hard to describe in the short space availa-
ble on the screen. When there are many similar choices on a menu, consider creating a second
menu level or a series of options for basic commands.
Never display a command that cannot be used. Many Windows applications gray out
commands that cannot be used; they are displayed on pull-down menus in a very light-
colored font, but they cannot be selected. Th is shows that they are available but cannot be
used in the current context. It also keeps all menu items in the same place.
When the user is about to perform a critical function that is diffi cult or impossible to
undo (e.g., deleting a fi le), it is important to confi rm the action with the user (and make sure
the selection was not made by mistake). Having the user respond to a confi rmation message,
which explains what the user has requested and asks the user to confi rm that this action is
correct, usually does this.
Simplify Recovery from Mistakes No matter what the system designer does, users will
make mistakes. Th e system should make it as easy as possible to correct these errors. Ideally,
the system has an Undo button that makes mistakes easy to override; however, writing the
soft ware for such buttons can be very complicated.
10 Keep it simple, stupid!
384 Chapter 10 Human–Computer Interaction Layer Design
Use Consistent Grammar Order One of the most fundamental decisions is the grammar
order. Most commands require the user to specify an object (e.g., fi le, record, word), and the
action to be performed on that object (e.g., copy, delete). Th e interface can require the user to
fi rst choose the object and then the action (an object–action order) or fi rst choose the action
and then the object (an action–object order). Most Windows applications use an object–action
grammar order (e.g., think about copying a block of text in your word processor).
Th e grammar order should be consistent throughout the system, both at the data element
level and at the overall menu level. Experts debate about the advantages of one approach over
the other, but because most users are familiar with the object–action order, most systems
today are designed using that approach.
Types of Navigation Controls
Th ere are two traditional hardware devices that can be used to control the user interface: the
keyboard and a pointing device such as a mouse, trackball, or touch screen. Today, depending
on the hardware being used, voice recognition systems can also be used to control the user
interface. Th ere are three basic soft ware approaches for defi ning user commands: languages,
menus, and direct manipulation.
Languages With a command language, the user enters commands using a special language devel-
oped for the computer system (e.g., UNIX and SQL both use command languages). Command
languages sometimes provide greater fl exibility than other approaches because the user can com-
bine language elements in ways not predetermined by developers. However, they put a greater
burden on users because users must learn syntax and type commands rather than select from
a well-defi ned, limited number of choices. Systems today use command languages sparingly,
except in cases where there is an extremely large number of command combinations that make it
impractical to try to build all combinations into a menu (e.g., SQL queries for databases).
Natural language interfaces are designed to understand the user’s own language (e.g.,
English, French, Spanish). Th ese interfaces attempt to interpret what the user means, and
oft en they present back to the user a list of interpretations from which to choose. An example
of the use of natural language is Google’s search engine. Google’s search engine enables users
to use free-form text to search the Web for topics of interest.
Menus Th e most common type of navigation system today is the menu. A menu presents a user
with a list of choices, each of which can be selected. Menus are easier to learn than languages
because a limited number of available commands are presented to the user in an organized fash-
ion. Clicking on an item with a pointing device or pressing a key that matches the menu choice
(e.g., a function key) takes very little eff ort. Th erefore, menus are usually preferred to languages.
Menus need to be designed with care because the submenus behind a main menu are
hidden from users until they click on the menu item. It is better to make menus broad and
shallow (i.e., each menu containing many items with only one or two layers of menus) rather
than narrow and deep (i.e., each menu containing only a few items, but each leading to three
or more layers of menus). A broad and shallow menu presents the user with the most infor-
mation initially so that he or she can see many options and requires only a few mouse clicks or
keystrokes to perform an action. A narrow and deep menu makes users hunt for items hidden
behind menu items and requires many more clicks or keystrokes to perform an action.
Research suggests that in an ideal world, any one menu should contain no more than
eight items, and it should take no more than two mouse clicks or keystrokes from any menu
to perform an action (or three from the main menu that starts a system).11 However, analysts
sometimes must break this guideline in the design of complex systems by grouping menu
11 Kent L. Norman, Th e Psychology of Menu Selection (Norwood NJ: Ablex Publishing Corp., 1991).
Navigation Design 385
items separated by a horizontal line. Oft en menu items have hot keys that enable experienced
users to quickly invoke a command with keystrokes in lieu of a menu choice (e.g., on a
Windows machine, across many applications, Ctrl-F tends to invoke the Find command; on
a Mac, you use Command-F instead).
Menus should put together like items so that the user can intuitively guess what each
menu contains. Most designers recommend grouping menu items by interface objects
(e.g., customers, purchase orders, inventory) rather than by interface actions (e.g., new,
update, format), so that all actions pertaining to one object are in one menu, all actions
for another object are in a diff erent menu, and so on. However, this is highly dependent
on the specifi c interface. Some of the more common types of menus include menu bars,
drop-down menus, pop-up menus, tab menus, toolbars, and image maps (see Figure 10-9).
Direct Manipulation With direct manipulation, the user enters commands by working
directly with interface objects. For example, users can change the size of objects in Microsoft
PowerPoint by clicking on them and moving their sides, or they can move fi les in Windows
Explorer by dragging the fi lenames from one folder to another. Direct manipulation can be
simple, but it suff ers from two problems. First, users familiar with language- or menu-based
Type of Menu When to Use Notes
Menu bar
List of commands at the top of
the screen; always on-screen
Main menu for system Use the same organization as the operating system and other
packages (e.g., File, Edit, View).
Menu items are always one word, never two.
Menu items lead to other menus rather than perform action.
Never allow users to select actions they can’t perform
(instead, use grayed-out items).
Drop-down menu
Menu that drops down imme-
diately below another menu;
disappears after one use
Second-level menu, often from
menu bar
Menu items are often multiple words.
Avoid abbreviations.
Menu items perform action or lead to another cascading
drop-down menu, pop-up menu, or tab menu.
Pop-up menu
Menu that pops up and fl oats
over the screen; disappears
after one use
As a shortcut to commands for
experienced users
Pop-up menus often (not always) are invoked by a right click
in Windows-based systems.
These menus are often overlooked by novice users, so usually
they should duplicate functionality provided in other menus.
Tab menu
Multipage menu with one tab
for each page that pops up and
fl oats over the screen; remains
on-screen until closed
When user needs to change sev-
eral settings or perform several
related commands
Menu items should be short to fi t on the tab label.
Avoid more than one row of tabs, because clicking on a tab to
open it can change the order of the tabs and in virtually no
other case does selecting from a menu rearrange the menu
itself.
Tool bar
Menu of buttons (often with
icons) that remains on screen
until closed
As a shortcut to commands for
experienced users
All buttons on the same tool bar should be the same size.
If the labels vary dramatically in size, then use two different
sizes (small and large).
Buttons with icons should have a tool tip, an area that dis-
plays a text phrase explaining the button when the user
pauses the mouse over it.
Image map
Graphic image in which
certain areas are linked to
actions or other menus
Only when the graphic image
adds meaning to the menu
The image should convey meaning to show which parts
perform action when clicked.
Tool tips can be helpful.
FIGURE 10-9 Types of Menus
386 Chapter 10 Human–Computer Interaction Layer Design
interfaces don’t always expect it. Second, not all commands are intuitive. [How do you copy
(not move) fi les in Windows Explorer? On the Macintosh, why does moving a folder to the
trash delete the fi le if it is on the hard disk, but eject the DVD if the fi le is on a DVD?]
Messages
Messages are the way the system responds to a user and informs him or her of the status of the
interaction. Th ere are many diff erent types of messages, such as error messages, confi rmation
messages, acknowledgment messages, delay messages, and help messages (see Figure 10-10). In
general, messages should be clear, concise, and complete, which are sometimes confl icting
objectives. All messages should be grammatically correct and free of jargon and abbrevia-
tions (unless they are the users’ jargon and abbreviations). Avoid negatives because they can
be confusing (e.g., replace Are you sure you do not want to continue? with Do you want to
quit?). Likewise, avoid humor, because it wears off quickly aft er the same message appears
dozens of times.
Messages should require the user to acknowledge them (by clicking, for example), rather
than being displayed for a few seconds and then disappearing. Th e exceptions are messages
that inform the user of delays in processing, which should disappear once the delay has
passed. In general, messages are text, but sometimes, standard icons are used. For example,
Windows displays an hourglass when the system is busy. All messages should be carefully
Type of Messages When to Use Notes
Error message
Informs the user that he or she
has attempted to do something
to which the system cannot
respond
When the user does something
that is not permitted or not
possible
Always explain the reason and suggest corrective action.
Traditionally, error messages have been accompanied by a
beep, but many applications now omit it or permit users to
remove it.
Confi rmation message
Asks users to confi rm that they
really want to perform the
action they have selected
When user selects a potentially
dangerous choice, such as
deleting a fi le
Always explain the cause and suggest possible action.
Often include several choices other than OK and cancel.
Acknowledgment message
Informs the user that the sys-
tem has accomplished what it
was asked to do
Seldom or never. Users quickly
become annoyed with all the
unnecessary mouse clicks
Acknowledgment messages are typically included because
novice users often like to be reassured that an action has
taken place.
The best approach is to provide acknowledgment informa-
tion without a separate message on which the user must
click. For example, if the user is viewing items in a list and
adds one, then the updated list on the screen showing the
added item is suffi cient acknowledgment.
Delay message
Informs the user that the
computer system is working
properly
When an activity takes more
than seven seconds
Should permit the user to cancel the operation in case he or
she does not want to wait for its completion.
Should provide some indication of how long the delay
will last.
Help message
Provides additional informa-
tion about the system and its
components
In all systems Help information is organized by table of contents and/or
keyword search.
Context-sensitive help provides information that depends on
what the user was doing when help was requested.
Help messages and online documentation are discussed in
Chapter 12.
FIGURE 10-10 Types of Messages
Input Design 387
craft ed, but error and help messages require particular care. Messages (and especially error
messages) should always explain the problem in polite, succinct terms (e.g., what the user
did incorrectly) and explain corrective action as clearly and as explicitly as possible so that
the user knows exactly what needs to be done. In the case of complicated errors, the error
message should display what the user entered, suggest probable causes for the error, and pro-
pose possible user responses. When in doubt, provide either more information than the user
needs or the ability to get additional information. Error messages should provide a message
number. Message numbers are not intended for users, but their presence makes it simpler
for help desks and customer support lines to identify problems and help users because many
messages use similar wording.
Navigation Design Documentation
Th e design of the navigation for a system is done through the use of WNDs and real use cases.
Real use cases are derived from the essential use cases (see Chapter 4), use scenarios, and WNDs.
Recall that an essential use case is one that describes only the minimum essential issues necessary
to understand the required functionality. A real use case describes a specifi c set of steps that a user
performs to use a specifi c part of a system. Real use cases are implementation dependent (i.e., they
are detailed descriptions of how to use the system once it is implemented).
To evolve an essential use case into a real use case, two changes must be made. First, the
use-case type must be changed from essential to real. Second, all events must be specifi ed
in terms of the actual user interface. And, given the peculiarities of diff erent platforms, e.g.,
desktops, tablets, and smartphones, real-use cases will need to be developed for each plat-
form on which the use case is being deployed. Th erefore, the normal fl ow of events, subfl ows,
and alternative/exceptional fl ows must be modifi ed. Th e normal fl ow of events, subfl ows,
and alternative/exceptional fl ows for the real use case associated with the storyboard user
interface prototype given in Figure 10-6 is shown in Figure 10-11. For example, step 2 of the
normal fl ow of events states that “Th e System provides the Sales Rep with the Main Menu
for the System,” which allows the Sales Rep to interact with the Maintain Client List aspect
of the system.
INPUT DESIGN
Inputs facilitate the entry of data into the computer system, whether highly structured data,
such as order information (e.g., item numbers, quantities, costs) or unstructured information
(e.g., comments). Input design means designing the screens used to enter the information as
well as any forms on which users write or type information (e.g., timecards, expense claims).
Basic Principles
Th e goal of the input mechanism is to simply and easily capture accurate information for the
system. Th e fundamental principles for input design refl ect the nature of the inputs (whether
batch or online) and ways to simplify their collection.
Online versus Batch Processing Th ere are two general formats for entering inputs into a com-
puter system: online processing and batch processing. With online processing (sometimes called
transaction processing), each input item (e.g., a customer order, a purchase order) is entered
into the system individually, usually at the same time as the event or transaction prompting the
input. For example, when you check a book out from the library, buy an item at the store, or
make an airline reservation, the computer system that supports that process uses online pro-
cessing to immediately record the transaction in the appropriate database(s). Online processing
388 Chapter 10 Human–Computer Interaction Layer Design
Use-Case Name: Maintain Client List ID: 12 Importance Level: High
Primary Actor: Sales Rep Use-Case Type: Detail, Real
Stakeholders and Interests: Sales Rep – wants to add, f ind, or list clients
Brief Description: Th is use case describes how sales representatives can search and maintain the
client list.
Trigger: Patient calls and asks for a new appointment or asks to cancel or change an existing
appointment.
Type: External
Relationships:
Association: Sales Rep
Include:
Extend:
Generalization:
Normal Flow of Events:
1. Th e Sales Rep starts up the system.
2. Th e System provides the Sales Rep with the Main Menu for the System.
3. Th e System asks Sales Rep if he or she would like to Add a client. Find an existing Client, or to
List all existing clients.
If the Sales Rep wants to add a client, he or she clicks on the Add Client Link and execute S-1:
New Client.
If the Sales Rep wants to f ind a client, he or she clicks on the Find Client Link and execute S-2:
Find Client.
If the Sales Rep wants to list all clients, he or she clicks on the List Client Link and execute S-3:
List Clients.
4. Th e System returns the Sales Rep to the Main Menu of the System.
Subfl ows:
S-1: New Client
1. Th e System asks the Sales Rep for relevant information.
2. Th e Sales Rep types in the relevant information into the Form
3. Th e Sales Rep submits the information to the System.
S-2: Find Client
1. Th e System asks the Sales Rep for the search information.
2. Th e Sales Rep types in the search information into the Form
3. Th e Sales Rep submits the information to the System.
4. If the System f inds a single Client that meets the search information,
the System produces a Client Information report and returns the Sales Rep to the Main
Menu of the System
Else
If the System f inds a list of Clients that meet the search information. Th e System executes
S-3: List Clients.
S-3: List Clients
1. If this Subf low is executed from Step 3
Th e System creates a List of All clients
Else
Th e System creates a List of clients that matched the S-2: Find Client search criteria.
2. Th e Sales Rep selects a client.
3. Th e System produces a Client Information report.
Alternate/Exceptional Flows:
S-2 4a. Th e System produces an Error Message.
FIGURE 10-11
Real Use-Case
Example
is most commonly used when it is important to have real-time information about the business
process. For example, when you reserve an airline seat, the seat is no longer available for some-
one else to use.
With batch processing, all the inputs collected over some time period are gathered
together and entered into the system at one time in a batch. Some business processes
Input Design 389
naturally generate information in batches. For example, most hourly payrolls are done
using batch processing because time cards are gathered together in batches and processed
at once. Batch processing is also used for transaction processing systems that do not
require real-time information. For example, most stores send sales information to district
offi ces so that new replacement inventory can be ordered. Th is information can be sent in
real time as it is captured in the store so that the district offi ces are aware within a second
or two that a product is sold. If stores do not need this up-to-the- second real-time data,
they will collect sales data throughout the day and transmit it every evening in a batch
to the district offi ce. Th is batching simplifi es the data communications process and oft en
saves in communications costs, but it does mean that inventories are not accurate in real
time but rather are accurate only at the end of the day aft er the batch has been processed.
Capture Data at the Source Perhaps the most important principle of input design is to cap-
ture the data in an electronic format at its original source or as close to the original source as
possible. In the early days of computing, computer systems replaced traditional manual sys-
tems that operated on paper forms. As these business processes were automated, many of the
original paper forms remained, either because no one thought to replace them or because it
was too expensive to do so. Instead, the business process continued to contain manual forms
that were taken to the computer center in batches to be typed into the computer system by a
data entry operator.
Many business processes still operate this way today. For example, most organizations
have expense claim forms that are completed by hand and submitted to an accounting
department, which approves them and enters them into the system in batches. Th ere are
three problems with this approach. First, it is expensive because it duplicates work (the
form is fi lled out twice, once by hand, once by keyboard).12 Second, it increases processing
time because the paper forms must be physically moved through the process. Th ird, it
increases the cost and probability of error, because it separates the entry from the pro-
cessing of information; someone might misread the handwriting on the input form, data
may be entered incorrectly, or the original input could contain an error that invalidates
the information.
Most transaction-processing systems today are designed to capture data at its source.
Source data automation refers to using special hardware devices to automatically capture data
without requiring anyone to type it. Stores commonly use bar-code readers that automatically
scan products and enter data directly into the computer system. No intermediate formats
such as paper forms are used. Similar technologies include optical character recognition,
which can read printed numbers and text (e.g., on checks), magnetic stripe readers, which can
read information encoded on magnetic strip (e.g., credit cards), and smart cards, which con-
tain microprocessors, memory chips, and batteries (much like credit card–sized calculators).
As well as reducing the time and cost of data entry, these systems reduce errors because they
are far less likely to capture data incorrectly. Today, portable computers and scanners allow
data to be captured at the source even in mobile settings (e.g., air courier deliveries, use of
rental cars).
Th ese automatic systems are not capable of collecting a lot of information, so the next-
best option is to capture data immediately from the source using a trained entry operator.
Many airline and hotel reservations, loan applications, and catalog orders are recorded
directly into a computer system, while the customer provides the operator with answers to
12 Or, in the case of the University of Georgia, three times: fi rst by hand on an expense form, a second time when it is
typed onto a new form for the “offi cial” submission because the accounting department refuses handwritten forms,
and, fi nally, when it is typed into the accounting computer system.
390 Chapter 10 Human–Computer Interaction Layer Design
questions. Some systems eliminate the operator altogether and allow users to enter their own
data. For example, many universities no longer accept paper-based applications for admis-
sions; all applications are typed by students into electronic forms.
Th e forms for capturing information (on a screen, on paper, etc.) should support the data
source. Th at is, the order of the information on the form should match the natural fl ow of
information from the data source, and data-entry forms should match paper forms used to
initially capture the data.
Minimize Keystrokes Another important principle is to minimize keystrokes. Keystrokes
cost time and money, whether they are performed by a customer, user, or trained data-entry
operator. Th e system should never ask for information that can be obtained in another way
(e.g., by retrieving it from a database or by performing a calculation). Likewise, a system
should not require a user to type information that can be selected from a list; selecting reduces
errors and speeds entry.
In many cases, some fi elds have values that oft en recur. Th ese frequent values should
be used as the default value for the fi eld so that the user can simply accept the value and not
have to retype it time and time again. Examples of default values are the current date, the area
code held by the majority of a company’s customers, and a billing address, which is based on
the customer’s residence. Most systems permit changes to default values to handle data-entry
exceptions as they occur.
Types of Inputs
Each data item that has to be input is linked to a fi eld on the form into which its value is
typed. Each fi eld also has a fi eld label, which is the text beside, above, or below the fi eld that
tells the user what type of information belongs in the fi eld. Oft en the fi eld label is similar to
the name of the data element, but they do not have to have identical words. In some cases, a
fi eld displays a template over the entry box to show the user exactly how data should be typed.
Th ere are many diff erent types of inputs, in the same way that there are many diff erent types
of fi elds (see Figure 10-12).
Text As the name suggests, a text box is used to enter text. Text boxes can be defi ned to have a
fi xed length or can be scrollable and can accept a virtually unlimited amount of text. In either
case, boxes can contain single or multiple lines of textual information. We never use a text
box if we can use a selection box.
Text boxes should have fi eld labels placed to the left of the entry area, their size clearly
delimited by a box (or a set of underlines in a non-GUI interface). If there are multiple text
boxes, their fi eld labels and the left edges of their entry boxes should be aligned. Text boxes
should permit standard GUI functions, such as cut, copy, and paste.
Numbers A number box is used to enter numbers. Some soft ware can automatically format
numbers as they are entered, so that 3452478 becomes $34,524.78. Dates are a special form
of numbers that sometimes have their own type of number box. Never use a number box if
you can use a selection box.
Selection Box A selection box enables the user to select a value from a predefi ned list. Th e
items in the list should be arranged in some meaningful order, such as alphabetical for long
lists or in order of most frequently used. Th e default selection value should be chosen with
care. A selection box can be initialized as unselected. However, it is better to start with the
most commonly used item already selected.
Input Design 391
Input Validation
All data entered into the system need to be validated to ensure their accuracy. Input validation
(also called edit checks) can take many forms. Ideally, computer systems should not accept
data that fail any important validation check to prevent invalid information from entering
the system. However, this can be very diffi cult, and invalid data oft en slip past data-entry
operators and the users providing the information. It is up to the system to identify invalid
data and either make changes or notify someone who can resolve the information problem.
Th ere are six diff erent types of validation checks: completeness check, format check, range
check, check digit check, consistency check, and database check (see Figure 10-13). Every system
should use at least one validation check on all entered data and, ideally, performs all appro-
priate checks where possible.
Type of Box When to Use Notes
Check box
Presents a complete list of
choices, each with a square
box in front
When several items can be
selected from a list of items
Check boxes are not mutually exclusive.
Do not use negatives for box labels.
Check box labels should be placed in some logical order,
such as that defi ned by the business process, or failing that,
alphabetically or most commonly used fi rst.
Use no more than ten check boxes for any particular set of
options. If you need more boxes, group them into subcat-
egories.
Radio button
Presents a complete list of
mutually exclusive choices,
each with a circle in front
When only one item can be
selected from a set of mutually
exclusive items
Use no more than six radio buttons in any one list; if you
need more, use a drop-down list box.
If there are only two options, one check box is usually preferred
to two radio buttons, unless the options are not clear.
Avoid placing radio buttons close to check boxes to prevent
confusion between different selection lists.
On-screen list box
Presents a list of choices in
a box
Seldom or never—only if there
is insuffi cient room for check
boxes or radio buttons
This type of box can permit only one item to be selected (in
which case it is an ugly version of radio buttons).
This type of box can also permit many items to be selected
(in which case it is an ugly version of check boxes), but
users often fail to realize they can choose multiple items.
This type of box permits the list of items to be scrolled, thus
reducing the amount of screen space needed.
Drop-down list box
Displays selected item in one-
line box that opens to reveal
list of choices
When there is insuffi cient room
to display all choices
This type of box acts like radio buttons but is more compact.
This type of box hides choices from users until it is opened,
which can decrease ease of use; conversely, because it
shelters novice users from seldom-used choices, it can
improve ease of use.
This type of box simplifi es design if the number of choices is
unclear, because it takes only one line when closed.
Combo box
A special type of drop-down
list box that permits uses to
type as well as scroll the list
Shortcut for experienced users This type of box acts like drop-down list but is faster for
experienced users when the list of items is long.
Slider
Graphic scale with a sliding
pointer to select a number
Entering an approximate
numeric value from a large
continuous scale
The slider makes it diffi cult for the user to select a precise
number.
Some sliders also include a number box to enable the user
to enter a specifi c number.
FIGURE 10-12 Types of Selection Boxes
392 Chapter 10 Human–Computer Interaction Layer Design
OUTPUT DESIGN
Outputs are what the system produces, whether on the screen, on paper, or in other media,
such as the Web. Outputs are perhaps the most visible part of any system because a primary
reason for using an information system is to access the information that it produces.
Basic Principles
Th e goal of the output mechanism is to present information to users so that they can accu-
rately understand it with the least eff ort. Th e fundamental principles for output design refl ect
how the outputs are used and ways to make it simpler for users to understand them.
Type of Validation When to Use Notes
Completeness check
Ensures all required data have
been entered
When several fi elds must be
entered before the form can
be processed
If required information is missing, the form is returned to the
user unprocessed.
Format check
Ensures data are of the right
type (e.g., numeric) and in the
right format (e.g., month, day,
year)
When fi elds are numeric or
contain coded data
Ideally, numeric fi elds should not permit users to type text
data, but if this is not possible, the entered data must be
checked to ensure it is numeric.
Some fi elds use special codes or formats (e.g., license plates
with three letters and three numbers) that must be checked.
Range check
Ensures numeric data are
within correct minimum and
maximum values
With all numeric data, if
possible
A range check permits only numbers between correct values.
Such a system can also be used to screen data for “reasona-
bleness”—e.g., rejecting birthdates prior to 1880 because
people do not live to be a great deal over 100 years old
(most likely, 1980 was intended).
Check digit check
Check digits are added to
numeric codes
When numeric codes are used Check digits are numbers added to a code as a way of enabling
the system to quickly validate correctness. For example, U.S.
Social Security numbers and Canadian Social Insurance num-
bers assign only eight of the nine digits in the number. The
ninth number—the check digit—is calculated using a mathe-
matical formula from the fi rst eight numbers.
When the identifi cation number is typed into a computer sys-
tem, the system uses the formula and compares the result
with the check digit. If the numbers don’t match, then an
error has occurred.
Consistency checks
Ensure combinations of data
are valid
When data are related Data fi elds are often related. For example, someone’s birth year
should precede the year in which he or she was married.
Although it is impossible for the system to know which
data are incorrect, it can report the error to the user for
correction.
Database checks
Compare data against a data-
base (or fi le) to ensure they
are correct
When data are available to be
checked
Data are compared against information in a database (or fi le)
to ensure they are correct. For example, before an iden-
tifi cation number is accepted, the database is queried to
ensure that the number is valid.
Because database checks are more expensive than the other
types of checks (they require the system to do more work),
most systems perform the other checks fi rst and perform
database checks only after the data have passed the previ-
ous checks.
FIGURE 10-13 Types of Input Validation
Output Design 393
Understand Report Usage Th e fi rst principle in designing reports is to understand how
they are used. Reports can be used for many diff erent purposes. In some cases—but not
very oft en—reports are read cover to cover because all information is needed. In most
cases, reports are used to identify specifi c items or used as references to fi nd information,
so the order in which items are sorted on the report or grouped within categories is criti-
cal. Th is is particularly important for the design of electronic or Web-based reports. Web
reports that are intended to be read from start to fi nish should be presented in one long
scrollable page, whereas reports that are used primarily to fi nd specifi c information should
be broken into multiple pages, each with a separate link. Page numbers and the date on
which the report was prepared are also important for reference reports.
Th e frequency of the report can also play an important role in its design and distribution.
Real-time reports provide data that are accurate to the second or minute at which they were
produced (e.g., stock market quotes). Batch reports are those that report historical informa-
tion that may be months, days, or hours old, and they oft en provide additional information
beyond the reported information (e.g., totals, summaries, historical averages).
There are no inherent advantages to real-time reports over batch reports. The only
advantages lie in the time value of the information. If the information in a report is time
critical (e.g., stock prices, air-traffic control information), then real-time reports have
value. This is particularly important because real-time reports are often expensive to pro-
duce; unless they offer some clear business value, they might not be worth the extra cost.
Manage Information Load Most managers get too much information, not too little (i.e., the
information load that the manager must deal with is too great). Th e goal of a well-designed
report is to provide all the information needed to support the task for which it was designed.
Th is does not mean that the report needs to provide all the information available on the
subject—just what the users decide they need in order to perform their jobs. In some cases, this
can result in the production of several diff erent reports on the same topics for the same users
because they are used in diff erent ways. Th is is not a bad design.
For users in Westernized countries, the most important information should always be
presented fi rst in the top-left corner of the screen or paper report. Information should be
provided in a format that is usable without modifi cation. Th e user should not need to re-sort
the report’s information; instead critical information should be highlighted so that users can
fi nd it more easily amid a mass of data, or perform additional mathematical calculations.
Minimize Bias No analyst sets out to design a biased report. Th e problem with bias is that
it can be very subtle; analysts can introduce it unintentionally. Bias can be introduced by the
way lists of data are sorted because entries that appear fi rst in a list can receive more atten-
tion than those later in the list. Data are oft en sorted in alphabetical order, making those
entries starting with the letter A more prominent. Data can be sorted in chronological order
(or reverse chronological order), placing more emphasis on older (or most recent) entries.
Data may be sorted by numeric value, placing more emphasis on higher or lower values. For
example, consider a monthly sales report by state. Should the report be listed in alphabetical
order by state name, in descending order by the amount sold, or in some other order (e.g.,
geographic region)? Th ere are no easy answers to this, except to say that the order of pres-
entation should match the way the information is used.
Graphical displays and reports can present particularly challenging design issues.13 Th e
scale on the axes in graphs is particularly subject to bias. For most types of graphs, the scale
should always begin at zero; otherwise, comparisons among values can be misleading. For
13 Two of the best books on the design of charts and graphical displays are by Edward R. Tuft e, Th e Visual Display
of Quantitative Information, Envisioning Information (Cheshire, CT: Graphics Press, 2001) and Visual Explanations:
Images and Quantities, Evidence and Narrative (Cheshire, CT: Graphics Press, 1997). Another good book is by
William Cleveland, Visualizing Data (Summit, NJ: Hobart Press, 1993).
394 Chapter 10 Human–Computer Interaction Layer Design
example, have sales increased by very much since year 1 (see Figure 10-14a and b)? Th e num-
bers in both charts are the same, but the visual images the two present are quite diff erent. A
glance at Figure 10-14a would suggest only minor changes, whereas a glance at Figure 10-14b
might suggest that there have been some signifi cant increases. In fact, sales have increased by
a total of 15 percent over fi ve years, or 3 percent per year. Figure 10-14a presents the most
accurate picture; Figure 10-14b is biased because the scale starts very close to the lowest value
in the graph and misleads the eye into inferring that there have been major changes. You
should also beware of the so-called 3D eff ects. For example, the pie charts in Figures 10-14c
and d represent the same data; in fact the data itself are constant. However, owing to the “3D”
pie chart, the slices nearer the front look bigger.
Types of Outputs
Th ere are many diff erent types of reports, such as detail reports, summary reports, excep-
tion reports, turnaround documents, and graphs (see Figure 10-15). Classifying reports is
challenging because many reports have characteristics of several diff erent types. For example,
some detail reports also produce summary totals, making them summary reports.
Media
Many diff erent types of media are used to produce reports. Today, most organizations have
moved toward “printing” reports electronically. One popular format is Adobe’s pdf. Th ese
“reports” are stored in electronic format on fi le servers or Web servers so that users can
Sales
120
100
80
60
40
20
0
(a) Unbiased graph with scale starting at 0
1 2 43 5 6 7 8 9 10
Sales
110
106
108
104
102
100
98
96
94
(b) Biased graph with scale starting at 94
1 2 43 5 6 7 8 9 10
(d) Biased graph in 3D(c) Unbiased graph in 2D
FIGURE 10-14 Bias in Graphs
Mobile Computing and User Interface Design 395
easily access them. Oft en the reports are available in more predesigned formats than their
paper-based counterparts because the cost of producing and storing diff erent formats is
minimal. Electronic reports also can be produced on demand as needed, and they enable
the user to more easily search for certain words. Furthermore, electronic reports can
provide a means of supporting ad hoc reports, where users customize the contents of the
report at the time the report is generated. Some users still print the electronic report on
their own printers, but the reduced cost of electronic delivery over distance and the ease of
enabling more users to access the reports than when they were only in paper form usually
off set the cost of local printing.
MOBILE COMPUTING AND USER INTERFACE DESIGN14
From a user interface design perspective, going mobile is both exciting and challenging.
Obviously, with today’s smartphones, such as the DroidTM or iPhoneTM, there are many pos-
sibilities. However, just because these phones have the ability to surf the Web doesn’t mean
Type of Report When to Use Notes
Detail report
Lists detailed information
about all the items requested
When user needs full informa-
tion about the items
This report is usually produced only in response to a query
about items matching some criteria.
This report is usually read cover to cover to aid understand-
ing of one or more items in depth.
Summary report
Lists summary information
about all items
When user needs brief informa-
tion on many items
This report is usually produced only in response to a query
about items matching some criteria, but it can be a com-
plete database.
This report is usually read for the purpose of comparing sev-
eral items to each other.
The order in which items are sorted is important.
Turnaround document
Outputs that “turn around”
and become inputs
When a user (often a customer)
needs to return an output to
be processed
Turnaround documents are a special type of report that are
both outputs and inputs. For example, most bills sent to
consumers (e.g., credit-card bills) provide information
about the total amount owed and also contain a form that
consumers fi ll in and return with payment.
Graphs
Charts used in addition to and
instead of tables of numbers
When users need to compare
data among several items
Well-done graphs help users compare two or more items or
understand how one has changed over time.
Graphs are poor at helping users recognize precise numeric
values and should be replaced by or combined with tables
when precision is important.
Bar charts tend to be better than tables of numbers or
other types of charts when it comes to comparing values
between items (but avoid three-dimensional charts that
make comparisons diffi cult).
Line charts make it easier to compare values over time,
whereas scatter charts make it easier to fi nd clusters or
unusual data.
Pie charts show proportions or the relative shares of a whole.
FIGURE 10-15 Types of Reports
14 Obviously, in a short section we cannot cover all of the issues related to developing mobile applications. For
anyone who is seriously considering developing mobile applications, we recommend that you begin by looking at
books that deal with the specifi c devices on which you will be deploying your application. For example, Donn Felker,
AndroidTM Application Development for DummiesTM (Hoboken, NJ: Wiley, 2011); Neal Goldstein and Tony Bove,
iPhoneTM Application Development All-In-One for DummiesTM (Hoboken, NJ: Wiley, 2010); Neal Goldstein and
Tony Bove, iPadTM Application Development for DummiesTM (Hoboken, NJ: Wiley, 2010); Chris Stevens, Designing
for the iPadTM: Building Applications that Sell (Chichester, UK: Wiley, 2011).
396 Chapter 10 Human–Computer Interaction Layer Design
that a simple Web interface is the answer. Th ese devices have limited screen space and have
capabilities, such as touch screens and haptic feedback (such as vibration or pulses), which
regular computers do not. Consequently, you really need to focus on designing the interface
for the device and not simply porting the Web interface over to it. Furthermore, you need
to realize that a tablet, such as the iPadTM, is not a big smartphone; it is in its own category
with its own challenges and capabilities. Consequently, you really need to design the inter-
face for mobile devices from the ground up. In this section, we discuss some challenges and
provide some guidelines to develop eff ective mobile interfaces. However, before we begin,
you should realize that all of the material described previously is still applicable. It’s just that
when you are dealing with these devices, additional issues must be considered.
Tidwell15 identifi es six challenges that a mobile user interface designer must face. Th e
screen of a phone is tiny. Th ere simply is not a lot of “real estate” available to use. Not only
are the screens tiny, but they come in diff erent sizes. What works on one screen might not
work on another screen. Some screens have haptic abilities: Th ey respond to touch and orien-
tation, and in some case, they vibrate. Obviously, these abilities are not available on all mobile
devices. However, they do provide interesting possibilities for user interface design. Virtual
and actual physical keypads are tiny. Consequently, too much typing can be challenging
for the user to input the right information. People use their mobile devices, especially their
phones, in all kinds of environments. Th ey use them in dark places (like a poorly lit class-
room). Th ey use them in bright sunlight. Th ey use them in quiet places (like the library or
movie theater) and they use them in noisy places (such as at a football game). Th ese devices
are simply used everywhere today. Because these devices are used everywhere, the users can be
easily distracted from the device. For example, have you ever texted someone when you aren’t
supposed to be using your phone, like during class? Or, what about out on a date? In other
words, users are typically multitasking while using their phone. Th ey do not want to spend a
lot of energy on trying to navigate your mobile site or app. Consequently, Krug’s three design
principles described earlier are very important, especially his fi rst one: Don’t make me think!
Based on these challenges, Tidwell provides a set of suggestions that you should follow
in designing a user interface for these devices. First, given the mobile context, you really need
to focus on what the user needs and not what the user might want. In other words, you really
should go back to business process and functional modeling (Chapter 4). In this case, only
focus on the tasks that users need to perform when they are in the mobile context. Th is is
a good example of a nonfunctional requirement (mobile computing) aff ecting the possible
functional requirements.
Second, if you are porting an application or website to a mobile device, remove all
“fl uff ” from the site: Strip the site down to its bare essentials. If the user needs access to
the full site, be sure to provide a link to it in an obvious location. Alternatively, you could
provide a complete mobile version of the application or website to the user. Obviously,
the design of the user interface will be diff erent, but the functionality should be the same.
Th ird, whenever possible, take advantage of the unique capabilities built into these
devices. Some of the devices have GPS built in. Depending on your application, knowing
where the user is could change the results. In other cases, the device has an accelerometer that
allows the app to “know” the orientation of the device. Many of these devices have speech
recognition capabilities, cameras that can be used for scanning, touch screens that allow
sophisticated gestures to be used, and haptic feedback, such as bumps and vibrations. All of
these capabilities could prove useful in developing diff erent mobile applications.
Fourth, when considering a phone, you tend to have a limited width from which to work.
Consequently, you should try to linearize the content of the application (see Figure 10-16).
15 Jenifer Tidwell, Designing Interfaces: Patterns for Eff ective Design, 2nd Ed. (Sebastopol, CA: O’Reilly, 2010).
Mobile Computing and User Interface Design 397
By that we mean, take advantage of vertical scrolling and try to
minimize, if not eliminate, horizontal scrolling. It is simply more
natural for users to scroll up and down instead of left to right on
these devices.
Fift h, optimize your mobile application for the user. Th is
includes minimizing the number of times the device must inter-
act with a server to download or upload information with a
server. Not everyone has access to 3G, alone true 4G, networks.
In many cases, uploading and downloading are still very slow.
Optimization also includes the user’s interaction with the device.
Instead of using a lot of typing, scrolling, and taps on a touch
screen, consider using the speech recognition capability. It’s a lot
easier to speak slowly to a smartphone than it is to have to type a
lot into a virtual or physical keyboard.
Tidwell also provides a set of reusable patterns that have
been customized for mobile devices. Th ese include things such as
a vertical stack, fi lmstrip, and bottom navigation.16
In addition to the general suggestions that Tidwell provides, the whole area of interaction
must be designed. With traditional GUI-based interfaces, the interactions tended to be limited
to typing on a keyboard; using a mouse to click, scroll, or zoom in an interface; or using a
combination of keyboard and the mouse to rotate part of the content in the interface. Overall,
it is a fairly limited set of interactions. However, with the new mobile devices that have speech
recognition, voice generation, touch screens, haptic feedback (via vibration), accelerometers
that allow the device to know its orientation, and cameras that can be used for scanning input,
the number of options for designing the navigation and input and output parts of the user
interface have increased substantially. From our perspective, we are just now detecting the tip
of the proverbial iceberg of possibilities with these devices. However, the general prototyping
approach suggested earlier in the chapter works. Just the number of options to consider has
increased substantially.
When it comes to the navigation part of the user interface, the primary additional option is
the use of the touchscreen.17 In fact, there is an entire vocabulary when it comes to the way users
interact with touchscreens. Today, the designer needs to consider tapping, pinching, spreading,
fl icking, scrolling (one-fi nger vs. two-fi nger), and dragging to name a few. For example:
■ Tapping can be used to open or activate an app, to select an object in the interface, or
to stop an action, such as scrolling.
■ Pinching is used to shrink or zoom out.
■ Spreading is used to enlarge or zoom in.
■ Flicking can be used to move an object or for interacting with a slider to scroll.
■ Scrolling can be accomplished by using a single fi nger on a scroll bar or by using two
fi ngers on anywhere on the interface.
■ Move an object on the screen by placing a fi nger on the object and dragging it
to another location. Th is is similar to using a mouse to drag an object to another
location.
Diff erent mobile devices may implement each of these slightly diff erently. Obviously, the num-
ber of choices to support navigation has greatly expanded.
FIGURE 10-16
Linearization
Example, Wiley
Business Study
Center, used with
permission.
16 Tidwell also suggests that the Design for Mobile (patterns.design4mobile.com) pattern library provides many good
patterns to use when developing mobile applications.
17 A good reference to gestural design is Dan Saff er, Designing Gestural Interfaces (Sebastopol, CA: O’Reilly, 2008).
398 Chapter 10 Human–Computer Interaction Layer Design
For the input part of the user interface, the primary additional options to consider are the
use of the camera as a device to scan items; microphone as a speech input device; and accelero-
meter to detect acceleration, orientation, and vibrations. For example, the Amazon MobileTM
and the eBay Mobile’s Red Laser Barcode & QR ScannerTM provide a means to easily scan a
barcode of a product and quickly look the product up online to provide all kinds of information
on the product. Google MapsTM supports a voice recognition interface to fi nd a location. Also,
there are numerous apps that use the accelerometer to detect acceleration and vibration (e.g.,
Wavefront Labs’ Accelerometer Data ProTM) and to compute orientation (e.g., NH for Mobikats’
Easy Spirit LevelTM). Furthermore, many games available on mobile devices use the accelero-
meter as an input device.
Th e primary additional options to consider for the output part of the user interface
include voice generation and haptic feedback. For example, Google MapsTM and WazeTM
support a voice generation capability that provides driving directions. Additionally, virtu-
ally all smartphones support a vibrate option when the phone rings and when alerts or text
messages arrive.
SOCIAL MEDIA AND USER INTERFACE DESIGN18
Given the impact that FacebookTM and TwitterTM, e.g., the Arab uprising, have had in
today’s world, developing applications for social media has obviously come to the fore-
front. In many ways, mobile computing and social media have grown up together. Like
mobile computing, each social media platform has its own capabilities and challenges.
Social media platforms range from sites that allow you to simply upload material to them,
such as FlickrTM and YouTubeTM, to sites that support a virtual existence in the metaverse,
such as Second LifeTM. During your career, you might need to develop applications for a
specifi c social media platform, such as FacebookTM or TwitterTM, or possibly develop your
own social media site.
When developing your own social media presence, you must understand who is your
target audience. Is the audience employees of your fi rm, or is the audience outside of the fi rm?
In this section, we only focus on an external audience. Once you know who the audience is,
you need to know what they are saying about the fi rm. In many ways, social media is nothing
more than another channel for marketing the fi rm’s products and capabilities. Before you can
deploy a social media presence, you really need to understand what the users’ needs (desires)
are. In other words, back to requirements determination. In this case, the problem is that the
users are “out there” somewhere, so typical approaches to gathering requirements, such as
interviews and observation, don’t work. Instead, you need to hunt through the Web to root
out your requirements. Some of the more useful places to look are blogs or other social media
outlets that address issues that would be of interest to your fi rm. When all else fails, you can
always use a search engine such as GoogleTM. Regardless, you obviously have to understand
the functional requirements before you can design your social media presence.
Once you understand your functional requirements, you need to determine what type
of social media presence is necessary to address the requirements eff ectively. Each social
media platform has its own niche. Consequently, you might need to deploy many diff erent
applications across diff erent platforms to eff ectively meet the fi rm’s social media presence
requirements. Also, you must look at your social media site as a means for your fi rm to
build and maintain a positive image or brand. Th erefore, the social media site must contain
material that your potential customers want to consume. You must remember that the
18 Much of the material in this section has been based upon material from Jenifer Tidwell, Designing Interfaces:
Patterns for Eff ective Design, 2nd Ed. (Sebastopol, CA: O’Reilly, 2010).
Social Media and User Interface Design 399
underlying purpose of marketing is to “manufacture” wants and then “convert” the wants
into needs. Given that your social media site is eff ectively another marketing channel, your
site must be able to draw in new customers and to get current customers to regularly return.
In this section, we provide some general guidelines for developing your own social media
site so that both new customers visit and current customers return.19
First, you really need to post to your site regularly. If the content of the site becomes
stagnant, no one will want to visit. The content of the site should contain a mixture of
media: videos, podcasts, sound clips, and so on. The site’s material should include a mix-
ture of firm-driven material, material from customers, and links to relevant content that
is located on other sites. Also, be sure to include ways for visitors to join in a “conversa-
tion” with the firm, such as FacebookTM comments or TwitterTM Tweets.
Second, make sure that you understand the difference between push and pull
approaches. If the user must come to you to fi nd out something, then you are using a pull-
based approach. On the other hand, if you put the information out to the user, then you
are using a push-based approach. When it comes to social media, you really need to use
a combination of the approaches. For example, in FacebookTM if someone posts on your
wall or sends you a request, FacebookTM will send you an e-mail message to try and entice
you back to the FacebookTM site. Th e act of posting to your site was a pull-based action,
and the e-mail message sent to you is a push-based action. In a nutshell, you want to focus
on more of a push-based approach. You want your content to get to your customers in
as an eff ective manner as possible. You don’t want them to have to come looking for you.
Encourage them to opt in for update notifi cations to come to them in a form that they
prefer. Some might prefer e-mail notifi cations, and others might prefer you post to their
FacebookTM or TwitterTM accounts. Also, be sure to include links to your social media sites
on your home page. But be sure not to overwhelm the customer. Not every customer wants
to know every tidbit regarding the fi rm. Only give the customer what the customer wants.
Remember, Krug’s fi rst principle: Don’t make me think! A corollary to this principle for
social media would be: Don’t make me work! Make it easy for the customer to fi nd only
what they want (or maybe what we want them to want).
Th ird, be sure that your home page and your social media sites are all synced together
so that when one is updated, the other sites “know” about the update. Th is makes your job of
maintaining the diff erent sites much easier, and it allows your customers to have a consistent
experience across all sites. However, don’t overdo this. It is obvious that diff erent sites have
diff erent media and, potentially, diff erent audiences. You aren’t going to use FacebookTM in the
same way you would use TwitterTM, YouTubeTM, or a blog. Be sure to include crosslinks among
the diff erent sites. Th is enables your customer to easily navigate through your diff erent sites.
Fourth, enable the customers to share the great content that you have created. You
can include buttons that allow them to email the content to their closest “friends” or other
followers in their own social network. You also should provide a means to gather feedback
from your customers regarding your content. One way is to include the ability for cus-
tomers to make and share comments regarding your content. Another way is to provide a
voting or “like” mechanism to encourage the customer to become engaged with your site.
Fift h, be sure to design your sites so that not only your customers can easily fi nd the
material for which they are searching, but also search engines can fi nd the material. Search
engines are at least as likely to bring new customers to your sites as other customers. Design
19 Two good books devoted to developing applications for social media in general are Erin Malone, Designing Social
Interfaces (Sebastopol, CA: O’Reilly, 2009) and Gavin Bell, Building Social Web Applications: Establishing Community
at the Heart of Your Site (Sebastopol, CA: O’Reilly, 2009). Th ere are a couple of books devoted to two specifi c social
media platforms. Two good books are Jesse Stay, FacebookTM Application Development for DummiesTM (Hoboken,
NJ: Wiley, 2011) and Dusty Reagan, TwitterTM Application Development for DummiesTM (Hoboken, NJ: Wiley, 2010).
400 Chapter 10 Human–Computer Interaction Layer Design
the site so that once the customer lands on your site, he or she stays there for a while. One way
that you can accomplish this is by providing the customer with links to “related” material. If
you decided to include a voting or “like” mechanism, be sure to enable the customer to see
the “best” or, at least, the most popular material fi rst. Another possibility is to create a lead-
erboard that displays the most shared material. You need to leverage the information gained
by implementing the fourth guideline.
Sixth, one of the more diffi cult things to accomplish is to have your sites become a place
that your customers feel that they belong. You want your customers to feel that they are
members of something; you want to try to build a feeling of community. Th e more they feel
that they belong, the more likely they will recommend your site to their friends. One way to
accomplish this is to encourage employees, at least the “right” employees, to author their own
“independent” sites that discuss topics of interest to your customers. Th is will give a more
personal feel to the fi rm and possibly entice customers to stick around on the site longer.
Finally, in most cases, your customers visit your sites using a variety of hardware plat-
forms. Th e platforms range from the desktop to the notebook to the tablet to the smartphone.
Consequently, all of the material related to general user interface design and to mobile com-
puting is applicable. Because you have a global audience, you need to be sure to take into
account international and cultural issues in your design.
GAMES, MULTIDIMENSIONAL INFORMATION VISUALIZATIONS,
AND IMMERSIVE ENVIRONMENTS20
With the advent of games and multidimensional information visualizations being used in busi-
ness and the potential of applying immersive technologies, such as Google GlassTM and the
Oculus Rift TM, to solve business problems using augmented and virtual reality, the design of the
human computer interaction layer is becoming even more important in information systems
development. In many ways, user interface design for games, multidimensional information
visualizations, and immersive environments is very similar to designing a user interface for more
traditional application areas. However, in other ways, it is very diff erent.
Games, Gamifi cation, and User Interface Design
Games have been around for a very, very long time. Th ey have been very successful in many
diff erent areas because they are fun and engaging.21 When applying games to business situations,
there are two general approaches to consider: development of games that support business pro-
cesses and gamifi cation of business processes. Th e development of games to solve business prob-
lems is relatively new. Traditionally, they have been used primarily with academic simulations.
However, given the popularity of games in our culture, business games are being developed
and deployed to increase customer and employee engagement.22 Gamifi cation deals with applying
gaming mechanics to non-gaming situations. Gamifi cation has been used to redesign classrooms
20 Obviously, in a short section we cannot cover all of the issues related to developing games, multidimensional infor-
mation visualizations, or using immersive environments to solve business applications. However, in this section we
provide with an overview of the types of issues that you may run into when using these technologies. We also provide
pointers to many references that have been useful in our development eff orts.
21 Jane McGonigal, Reality Is Broken: Why Games Make Us Better and How Th ey Can Change the World (New York:
Penguin Books, 2011); James Paul Gee, Why Video Games Are Good for Your Soul (Champaign, IL: Common Ground
Publishing, 2005); Bernard Suits, Th e Grasshopper: Games, Life, and Utopia (Ontario, CA: Broadview Press, 2005).
22 Jesse Schell, Th e Art of Game Design: A Book of Lenses (Boca Raton, FL: CRC Press, 2008); Jon Radoff , Game On:
Energize Your Business with Social Media Games (Indianapolis, IN: Wiley, 2011); Bryon Reeves and J. Leighton Read,
Total Engagement: Using Games and Virtual Worlds to Change the Way People Work and Businesses Compete (Boston,
MA: Harvard Business Press, 2009).
Games, Multidimensional Information Visualizations, and Immersive Environments 401
and to support learning, and, like games, it too has been used to increase customer and employee
engagement.23
In both using games and gamifi cation, the secret to success deals with motivating the cus-
tomer and/or employee to remain engaged with the business process. Even though traditional
motivation approaches have worked to motivate employees in the past, due to the nature of the
changing types of work performed, they no longer function in an effi cient or eff ective manner
(see Chapter 2). Traditional approaches typically used a “stick and carrot” approach to moti-
vation. If an individual (child, student, or employee) did something undesired, he or she was
punished. On the other hand, if they did something desired, they were rewarded. In other words,
motivation was based entirely on extrinsic benefi ts. Games and gamifi cation focuses primarily
on intrinsic benefi ts; not extrinsic benefi ts.24 Typically, you play games because you enjoy the
game. When was the last time that you played a game because you had to and not wanted to?
Did you play it for money? Was it so that you could please someone else? Or, was it so that you
could be part of something larger than yourself? How enjoyable was it? Are you motivated to
play it again? Why? Chances are that money was not a suffi cient motivator to play it again. But,
playing a game for the fun of it will probably motivate you to play it over and over again. In fact,
depending on your gaming personality type, playing a game to please someone else or to be part
of something larger than yourself may motivate you to play it again.25
Given the success of business game development and the gamifi cation of business processes,
there are a few things that we can learn to improve user interfaces. One of the fi rst things is that
games are designed explicitly to be fun.26 Typically, when we design business information sys-
tems, one of the last things we think about is whether the system is fun to use or not. When was
the last time you considered using an accounting information system as being fun? However,
in this case, the fun component of a business information system deals specifi cally with how
engaging the user interface is.27 Th erefore, there are a few things that we can apply from games
to develop more engaging user interfaces.28
First, games are about creating a user experience. Obviously, when creating an experience,
the user interface designer must pay close attention to all of the issues that we have described
earlier. Otherwise, not only will the experience be light on the engagement factor, it could create
a negative experience instead of the positive one that we hope for.
Second, game experiences are all about the ideas and themes woven throughout the game.
In our case, the ideas of story telling (see Chapter 3) and use cases (see Chapter 4) provide a basis
to design and develop the user(s)’ engagement experience.
Th ird, game developers worry a lot about the player (user). Th is brings a better focus to the
roles (actors) that the users play in our system (see Chapter 3). When it comes to game design,
not only do we have to worry about the tasks in which the user will be engaged, but we also
23 Lee Sheldon, Th e Multiplayer Classroom: Designing Coursework as a Game (Boston, MA: Course Technology, 2012);
Rajat Paharia, Loyalty 3.0: How Big Data and Gamifi cation Are Revolutionizing Customer and Employee Engagement
(New York: McGraw-Hill, 2013); Gabe Zichermann and Joselin Linder, Th e Gamifi cation Revolution: How Leaders
Leverage Game Mechanics to Crush the Competition (New York: McGraw-Hill, 2013); Kris Duggin and Kate Shoup,
Business Gamifi cation for Dummies (Hoboken, NJ: Wiley, 2013).
24 Kevin Werbach and Dan Hunter, For the Win: How Game Th inking Can Revolutionize Your Business (Philadelphia,
AA: Wharton Digital Press, 2012).
25 Ralph Koster, A Th eory of Fun for Game Design, 2nd Ed. (Sebastopol, CA: O’Reilly Media, 2014); Jesse Schell, Th e
Art of Game Design: A Book of Lenses (Boca Raton, FL: CRC Press, 2008); Kevin Werbach and Dan Hunter, For the
Win: How Game Th inking Can Revolutionize Your Business (Philadelphia, AA: Wharton Digital Press, 2012).
26 Ralph Koster, A Th eory of Fun for Game Design, 2nd Ed. (Sebastopol, CA: O’Reilly Media, 2014).
27 Jon Radoff , Game On: Energize Your Business with Social Media Games (Indianapolis, IN: Wiley, 2011) and Bryon
Reeves and J. Leighton Read, Total Engagement: Using Games and Virtual Worlds to Change the Way People Work and
Businesses Compete (Boston, MA: Harvard Business Press, 2009).
28 Jesse Schell, Th e Art of Game Design: A Book of Lenses (Boca Raton, FL: CRC Press, 2008).
402 Chapter 10 Human–Computer Interaction Layer Design
have to start thinking about the individual psychological and cognitive diff erences among the
diff erent users.29 Th is applies both to diff erent types of customers and employees.
Fourth, game developers also tend to try and build a community around the game. In
this way, users have a built-in support mechanism. Schell30 suggests a set of tips to develop a
strong community that can be applied to general business information systems development.
He suggests that we should foster friendships by encouraging the users to talk with each other
about the system and we should try to create community property by having the users (and
developers) take joint responsibility for the system. However, one of his more relevant sug-
gestions is to support multiple levels of users based on their level of experience. By having the
system detect the level of expertise of the user in using the system, the system can introduce
features, such as short cuts, once a specifi c “level” has been reached. Th is is associated with
“leveling up” in games. Th is would help in addressing the trade-off s between ease of learning
and ease of use when developing a user interface. Th is could also encourage users to “buy in”
to the system.
Fift h, when it comes to successful game design, you must consider the aesthetics. As such,
without a focus on aesthetics, the experience that you want the customer or employee to incur
may be less than desirable. In fact, it could discourage them from returning to your site.
Multidimensional Information Visualization Design
Th ere have been many diff erent types of multidimensional information visualizations that have
been used in business.31 However, the diff erent types fall into two basic categories: multidimen-
sional information visualizations in 2D space and multidimensional information visualizations
in nonimmersive 3D space. Th ose visualizations that are displayed in 2D space include the basic
business charts and graphs you would fi nd in a spreadsheet or statistics package, e.g., heat maps,
maps, node-link diagrams, parallel coordinates, radar charts, scatterplots, and treemaps.32 Th e
primary issue related to the use of these types of charts and diagrams deals with the potential of
bias creeping into the display (see earlier in the chapter). However, when considering visualiza-
tions that are displayed in nonimmersive 3D space, additional issues are raised.33
Th e fi rst issue that comes up is the issue of being able to determine a specifi c value that is
being represented. For example, in Figure 10-17, which portrays a multidimensional surface
chart, it is virtually impossible to determine specifi c values represented in the 3D space. In this
case, there are four separate values being plotted: one for the X-axis, one for the Y-axis, one
for the Z-axis, and one that uses the color of the surface. Only the last value, due to the legend,
can be easily determined. Another example of this problem is portrayed in Figure 10-18 that
shows a multidimensional bar chart. Again, four separate values are depicted. Th ere are two
29 Chaomei Chen, Information Visualization and Virtual Environments (London: Springer-Verlag, 1999); Howard
Gardner, Frames of Mind: Th e Th eory of Multiple Intelligences (New York: Basic Books, 1983).
30 Jesse Schell, Th e Art of Game Design: A Book of Lenses (Boca Raton, FL: CRC Press, 2008).
31 David Tegarden, “Business Information Visualization,” Communications of the Association for Information Systems,
Vol. 1, Article 4 (1999) (http://aisel.aisnet.org/cais/vol1/iss1/4).
32 A set of good books that address these types of information visualizations are: Ben Fry, Visualizing Data (Sebas-
topol, CA: O’Reilly Media, 2008); Derek L. Hansen, Ben Shneiderman, and Marc A. Smith, Analyzing Social Media
Networks with NodeXL: Insights from a Connected World (Burlington, MA: Morgan Kaufmann, 2011); Nathan Yau,
Visualize Th is: Th e FlowingData Guide to Design, Visualization, and Statistics (Indianapolis, IN: Wiley, 2011); Nathan
Yau, Data Points: Visualization Th at Means Something (Indianapolis, IN: Wiley, 2013).
33 A set of good books that address these types of information visualizations are: Judith R. Brown, Rae Earnshaw, Mikael
Jern, and John Vince, Visualization: Using Computer Graphics to Explore Data and Present Information (New York:
Wiley, 1995); Robert Spence, Information Visualization (Harlow England: ACM Press, 2001); Chaomei Chen, Informa-
tion Visualization: Beyond the Horizon, 2nd Ed. (London: Springer-Verlag, 2004); Usama Fayyad, Georges Grinstein,
and Andreas Wierse (eds.), Information Visualization in Data Mining and Knowledge Discovery (San Francisco: Morgan
Kaufmann, 2002).
basic approaches used to address
this problem: being able to rotate
and zoom the visualization and pro-
viding a drill-down capability that
allows the specifi c values to be dis-
played (see Figure 10-19).
Another issue that comes up
when displaying data in 3D space
is occlusion; that is, when viewing
data in 3D, some of the visualiza-
tion may be covered up, hidden,
by other parts of the visualization.
For example, in Figure 10-18, the
values drawn at the “back” of the
visualization cannot be easily deter-
mined. In Figure 10-19, the negative
values are drawn below the surface
of the fl oor of the visualization. As
such, the negative values cannot be
seen. However, by rotating the vis-
ualization “up” and being able to
click on a specifi c value, the values
associated with that observation can
be drilled down into and displayed
in a semi-transparent window. In
Figure 10-20, the visualization dis-
plays the basic values on the fl oor.
With this visualization, in addition
to supporting displays on the walls,
the user can also use a slicing plane
that “cuts through” the visualiza-
tion to help better understand the
data being visualized. Th ese types of
visualizations have been used quite
extensively in supporting business
decision making.
Th ere are many more types of multidimensional information visualizations that have been
used in business, e.g., volumes, fl oors and walls, maps, and surfaces. Each has its own strengths,
weaknesses, and challenges. Furthermore, today there are many specialized tools that can be
used to aid in designing and developing these types of visualizations. However, the basic design
process is essentially the same user interface design process described earlier. You still have to
design the navigation controls, the input mechanisms, and the output. To begin with, you will
have to understand the underlying problem domain and the tasks that the user needs to per-
form. Next, you will need to choose the type of visualization to be designed based on the task
that the user needs supported. Th is is still an art form. Our recommendation is to sit down with
the user and go through diff erent types of information visualizations to try and determine which
of the information visualizations are reasonable. Th is decision should be based on whether the
mapping of the data to the visualization is “intuitive” from the user’s perspective or not. Also,
remember that just because you can implement a complex, multidimensional information visu-
alization does not mean that you should do it. In many cases, simple business charts and graph-
ics are more than suffi cient. Like game design, be sure to focus on aesthetics. If the visualization
FIGURE 10-17
Multidimensional
Surface Chart in
3D Space
FIGURE 10-18
Multidimensional
Bar Chart in
3D Space
FIGURE 10-19
Multidimensional
Bar Chart on
Floors and Walls in
3D Space
Games, Multidimensional Information Visualizations, and Immersive Environments 403
404 Chapter 10 Human–Computer Interaction Layer Design
is not pleasing to the eye of the user, then it
is probably the wrong visualization. Finally,
given that visualization design is more of an
art than a science, testing the visualization’s
eff ectiveness with many users is critical.
User Interface Design and
Immersive Environments
Augmented and virtual reality using immer-
sive technologies, such as Google GlassTM and
the Oculus Rift TM, is among the latest and exciting application areas being utilized to solve
business problems. Where virtual reality (VR) technologies completely immerse the user into
an artifi cial simulated digital environment, augmented reality (AR) technologies are used to aug-
ment or enhance the view of the real world. Th ere are both opportunities and challenges with
deploying both of these technologies.
AR has primarily been used in advertising, real world navigation, room design, games,
social networking, and medical applications.34 If you happen to watch the NFLTM on TV,
you have already experienced augmented reality: Th ink of the “fi rst down” line that magically
appears on the screen. Th e typical hardware used is a smartphone or tablet. One of the primary
challenges in AR is the ability to use the camera to see and the soft ware to interpret the real-
world landscape that is being augmented by the system. Th is is known as object recognition. Th is
especially is a problem when using AR browsers that “connects” the real world with content on
the Web. In this case, a browser, such as JunaioTM, must recognize where it is physically located
and add “links” to allow the user to look up additional information about the locations that it
sees. To do this, the browser must use the camera, the GPS, and the accelerometer of the smart-
phone. Another major issue with regard to AR is of a social nature. For example, using Google
GlassTM while talking with someone can raise issues about privacy, such as are you paying
attention to the other person or are you looking at the information being displayed and what
exactly is the information being displayed. Th ere are also apps available today that support facial
recognition, which obviously raises even more issues with privacy. However, there are all kinds
of possible benefi ts from using this technology, e.g., think about the advertising that takes place
in the movie Minority Report. Using Google GlassTM could allow you to see information that has
been personalized about the locations around you as you walk down the street.
A prototyping approach, similar to the user interface design approach described earlier is
used to design AR applications.35 All of the issues related to traditional user interface design,
game design, and multidimensional information visualization design are relevant and must be
addressed, e.g., imagine using a heads-up display while driving your car to provide information
about your location; occlusion becomes a real issue. Consequently, testing the AR system in
real-world situations is essential.
VR has been both overhyped and under utilized. Even though today VR has primarily been
associated with games, it has been used in business for a long time.36 It has been used in areas
such as derivatives trading, fi nancial risk management, industrial process control, marketing
FIGURE 10-20
Multidimensional
Bar Chart on Floor
with Line Graphs on
Walls in 3D Space
34 A good book to get a basic overview of augmented reality is Gregory Kipper and Joseph Rampolla, Augmented
Reality: An Emerging Technologies Guide to AR (Waltham, MA: Syngress, 2013).
35 Tony Mullen, Prototyping Augmented Reality (Indianapolis, IN: Syngress, 2011).
36 David Tegarden, “Business Information Visualization,” Communications of the Association for Information Systems,
Vol. 1, Article 4 (1999) (http://aisel.aisnet.org/cais/vol1/iss1/4); Alan Wexelblat (ed.), Virtual Reality: Applications
and Explorations (Boston, MA: Academic Press, 1993); Dimitris Chorafas and Heinrich Steinmann, Virtual Reality:
Practical Applications in Business and Industry (Englewood Cliff s, NJ: Prentice Hall, 1995); Robert Th ierauf, Virtual
Reality Systems for Business (Westport, CN: Quorum Books, 1995).
analysis, network modeling, operations management, organizational modeling, portfolio man-
agement, product design and manufacturing, room layout design, sensitivity analysis, simulated
meetings, stock market analysis, and training. However, from a user interface design perspective,
VR raises additional issues that need to be addressed.
It is believed that VR attains its power by captivating the user’s attention and inducing a
sense of immersion, the feeling of being present in the space being simulated. Th e challenge
of creating the feeling of immersion has two primary dimensions: sensory and aff ective. In
the sensory dimension, the combination of stimuli employed must be of suffi cient vividness
that an individual’s automatic perceptual processes are triggered, resulting in the simulation
being perceived as life-like. In the aff ective dimension, the user should be cast in an interactive,
exploratory role. When considering these dimensions, one should remember that virtual reality
resides in an individual’s consciousness and, therefore, the relative contribution of each of these
dimensions in creating a sense of immersion will vary across individuals. Th us, immersion is a
function of both technology and perceiver. Th is raises the issue of individual psychological and
cognitive diff erences.37
Interaction with a virtual world may take the form of wayfi nding through the virtual space,
rearranging existing, or creating new, 3D objects, or communicating with another agent (person
or automaton) sharing the same virtual space. Visitors to large virtual worlds are oft en unable
to comprehend the overall topological structure of the space. Th ey may wander aimlessly when
attempting to fi nd a particular location for the fi rst time and may subsequently have diffi culty
fi nding their way back to locations already visited. Wayfi nding tasks require the user to be able
to conceptualize the virtual space as a whole and to develop a cognitive map of it.38 A cognitive
map consists of not only spatial relationships, but also of auditory, sensory, and emotional
impressions. In games, a map is typically provided to help with understanding where one is in
the virtual space and from where one has come. Like the immersion issue, wayfi nding also raises
issues related to individual psychological and cognitive diff erences.
Th e last challenge with regard to using VR as a user interface platform deals with collabora-
tion. When considering multiuser, distributed VR systems, such as many of today’s video games,
occlusion can become a very large problem. Not only can objects in the VR space hide other VR
objects, they can hide other users. Also, it is possible for one user to see something of interest that
the other users may not. In this case, the issue of wayfi nding comes back up. For example, if user
A fi nds an interesting piece of information, then user A must communicate how to navigate to a
location in which the other users will be able to observe the fi nding. Th ere are multiple possibil-
ities here, including providing wayfi nding directions from a specifi ed “viewpoint,” “teleporting”
the other users from their individual current locations to the current location of user A, having
user A go fi nd the other users and bring them to the appropriate location, or having user A
simply “drive” all other users to the appropriate location by taking over their ability to navigate
through the visualization. Furthermore, once the other users are at the appropriate location, how
do they return to their previous location?
Obviously, designing eff ective and effi cient VR applications is very diffi cult.39 Again, the
overall design process is similar to the general user interface design process described ear-
lier. However, given the potential for VR to support business decision making by combining
37 Chaomei Chen, Information Visualization and Virtual Environments (London: Springer-Verlag, 1999); Howard
Gardner, Frames of Mind: Th e Th eory of Multiple Intelligences (New York: Basic Books, 1983).
38 Reginald Golledge (ed.), Wayfi nding Behavior: Cognitive Mapping and Other Spatial Processes (Baltimore, MD: Th e
John Hopkins University Press, 1999); Rob Kitchin and Scott Freundschuh, Cognitive Mapping: Past, Present and
Future (London: Routledge, 2000).
39 A recent book that tackles how to design a VR system is Ann Lantham Cudworth, Virtual World Design (Boca
Raton, FL: CRC Press, 2014).
Games, Multidimensional Information Visualizations, and Immersive Environments 405
406 Chapter 10 Human–Computer Interaction Layer Design
gaming and information visualization technologies into a single seamless distributed envi-
ronment and that the investment in specialized hardware and soft ware is dropping, VR could
provide large payoff s.
INTERNATIONAL AND CULTURAL ISSUES AND USER
INTERFACE DESIGN40
With the World Wide Web, virtually any fi rm can have a global presence. With this capabil-
ity, a fi rm must be cognizant of a set of international and cultural issues. Th ese issues include
multilingual requirements, color, and cultural diff erences.
Multilingual Requirements
Th e fi rst and most obvious diff erence between applications used in one region and those
designed for global use is language. Global applications oft en have multilingual requirements,
which means that they have to support users who speak diff erent languages and write using
non-English letters (e.g., those with accents, Cyrillic, Japanese). One of the most challenging
aspects in designing global systems is getting a good translation of the original language mes-
sages into a new language. Words oft en have similar meanings but can convey subtly diff erent
meanings when they are translated, so it is important to use translators skilled in translating
technical words. A few rules that you should follow are to:
■ Keep the writing short and simple. It is much easier to avoid mistranslations.41
■ Avoid humor, jargon, slang, clichés, puns, analogies, and metaphors. Th ese tend to
be too culturally specifi c. Consequently, the underlying point being made will most
likely be lost in translation.
■ Use good grammar. Be sure to punctuate everything correctly. Even though you
might be tempted to ignore grammar and punctuation rules to try to make a point, it
makes translating more diffi cult, especially for automated translation systems. Don’t
depend on automated spelling and grammar checkers to enforce this. At this time,
they simply aren’t good enough.
Another challenge is oft en screen space. In general, English-language messages usually take
20 percent to 30 percent fewer letters than their French or Spanish counterparts. Designing
global systems requires allocating more screen space to messages than might be used in the
English-language version.
Some systems are designed to handle multiple languages on the fl y so that users in dif-
ferent countries can use diff erent languages concurrently; that is, the same system supports
several diff erent languages simultaneously (a concurrent multilingual system). Other systems
contain separate parts that are written in each language and must be reinstalled before a
specifi c language can be used; that is, each language is provided by a diff erent version of the
system so that any one installation will use only one language (i.e., a discrete multilingual
40 A set of books that provide a good introduction to building information systems for a multicultural audience
include Elisa M. del Galdo and Jakob Nielsen, International User Interfaces (New York, NY: Wiley, 1996); Nitish
Singh and Arun Pereira, Th e Culturally Customized Web Site: Customizing Web Sites for the Global Marketplace
(Oxford, UK: Elsevier Butterworth Heinemann, 2005); John Yunker, Beyond Borders: Web Globalization Strategies
(Berkley, CA: New Riders, 2003).
41 However, even this does not guarantee good translations if you use an automatic translation facility. For example,
type the text “I would like my steak cooked rare” into babel fi sh (http://babelfi sh.yahoo.com/) and translate it to
Russian and back to English. You will get back “I wanted would be my rare welded [steykom] done”—not exactly
the most useful translation.
International and Cultural Issues and User Interface Design 407
system). Either approach can be eff ective, but this functionality must be designed into the
system well in advance of implementation.
Finally, one other consideration that must be considered is reading direction. In most
Western societies, readers read from left to right and top to bottom. Th is is not true for
many cultures. For example, in Arabic countries, readers typically read right to left and top
to bottom.
Color
To begin with, color is not black and white. Th e meaning associated with a color is totally
culturally dependent. In fact, black and white isn’t necessarily black and white; they could
be white and black. In most Western cultures, black is associated with death, mourning, and
grief or with respect and formality. For example, in the United States, we typically wear black
to a funeral, or you would expect to see religious leaders in black (think about the robes typ-
ically worn by a Catholic priest). In many Eastern cultures, on the other hand, white is asso-
ciated with death or the color of robes worn by religious leaders. In an example reported by
Singh and Pereira, when senior citizens in the United States and India were asked to “visualize
the following statement: A lady dressed in white, in a place of worship,” the results that came
back were as near to the opposite as one could get. In India, the lady would be a widow, but
in the United States she would be expected to be a bride.
Other colors that have meanings that are culturally driven include green, blue, red,
yellow, and purple. In the United States, red implies excitement, spice passion, sex, and
even anger; in Mexico, it indicates religion; in the United Kingdom, it indicates authority,
power, and government; in Scandinavian countries, it indicates strength; and in China, it
means communism, joy, and good luck. Blue is associated with holiness in Israel; cleanliness
in Scandinavia; love and truth in India; loyalty in Germany; and trust, justice, and “offi cial”
business in the United States. In Ireland, green signifi es nationalism and Catholicism, and in
the United States it denotes health, environmentalism, safety, greed, and envy. Green is a very
confusing color for Americans. In the Arab Middle East green is a sign of holiness, in France
it represents criminality, and in Malaysia it signifi es danger and disease. Yellow also has many
culturally dependent meanings. In the United States, it is associated with caution and cow-
ardice; in Scandinavia, warmth; in Germany, envy; and in India, commerce. Purple signifi es
death, nobility, or the Church in Latin America, the United States, and Italy, respectively.
Obviously, when building a website for a global audience, colors must be chosen carefully;
otherwise, unintentional messages will be sent.
Cultural Diff erences
Th e New York Times columnist Tom Friedman talks about the need for a fi rm to use its own
local capabilities as a basis for competitive advantage in a global market. He refers to this process
as glocalization. In some ways, when developing a website for an international audience, you
need to consider the opposite of glocalization. You need to think about what message needs
to be sent to a local culture from your global organization to achieve the business goals of the
fi rm. Consequently, you need to be able to understand the diff erent local cultures. Cultural
issues have been studied at both organizational and national levels. Diff erent researchers have
emphasized diff erent dimensions on which to focus our attention. In this section, we limit our
discussion to cultural issues that eff ect designing eff ective user interfaces. In particular, we only
address the research of Edward Hall and Geert Hofstede.42
42 See Geert Hofstede, Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations Across
Nations, 2nd Ed. (Th ousand Oaks, CA: Sage, 2001); Geert Hofstede, Gert Jan Hofstede, and Michael Minkov,
Cultures and Organizations: Soft ware of the Mind, 3rd Ed. (New York: McGraw-Hill, 2010); Edward T. Hall, Beyond
Culture (New York: Anchor Books, 1981).
408 Chapter 10 Human–Computer Interaction Layer Design
Hall identifi ed three dimensions that are directly relevant to user interface design:
speed of messages, context, and time. Th e speed of messages dimension deals with how fast
a member of a culture is expected to understand a message and how “deep” the content
of a typical message will be in a culture. Th e deeper the message content, the longer it
will take for a member of a culture to understand the message. For example, two diff er-
ent approaches to describe a historical event would be a news headline (fast and shallow)
and a documentary (slow and deep). According to Hall, diff erent cultures have diff erent
expectations of the content of and response to a message. Th is particular dimension has
implications for the content of the message contained in the user interface. Krug’s third
design principle turns out to be culturally driven. For a Western audience, minimizing the
number of words contained in a user interface makes sense. Westerners prefer to get to
the point as fast as possible. However, this is not true for Eastern cultures.43 Consequently,
for a fi rm like Amazon.com, providing detailed reviews and short excerpts from a book
provides support for a slow and deep culture, while providing bullet point types of com-
ments supports the fast and shallow culture. By providing both, Amazon.com addresses
both needs.
Th e second dimension, context, deals with the level of implicit information that is used
in the culture versus the information needing to be made explicit. In high-context cultures,
most information is known intrinsically and does not have to be made explicit. Th erefore, the
actual content of the message is fairly limited. However, in low-context cultures, everything
must be spelled out explicitly to avoid any ambiguity, and therefore the message needs to be
very detailed. You will fi nd this dimension causing problems when attempting to close a busi-
ness deal. In most Western societies, the lawyers want everything spelled out. In contrast, in
most Eastern societies, it may, in fact, be considered insulting to have to spell everything out.
From a website design perspective, Singh and Pereira point out that in a high-context culture,
focusing the design on aesthetics, politeness, and humility produces an eff ective website, but
in a low-context culture, things such as the terms and conditions of a purchase, the “rank”
of the product and fi rm, and the use of superlatives in describing the product and fi rm are
critical attributes of a successful website.
Hall’s third dimension, time, addresses how a culture deals with many diff erent things
going on simultaneously. In a polychronic time culture, members of the culture tend to do
many things at the same time but are easily distracted and view time commitments as very
fl exible. With monochronic time cultures, members of the culture solve many things by focus-
ing on one thing at a time, are single-minded, and consider time commitments as something
that is set in stone. When designing for a polychronic culture, the liberal use of “pop-up”
messages might be fun and engaging, while in a monochronic culture, pop-up messages sim-
ply annoy the user. In the past, Northern Hemisphere cultures have been monochronic and
Southern Hemisphere cultures have been polychronic. However, with the use of e-mail inter-
ruptions and text messaging, this could change over time. Regardless, allowing interruptions
to occur does in fact distract the users from their current task. Depending on the culture, this
could be a good or bad thing to support.
Hofstede also has identifi ed cultural dimensions that are relevant to the user interface.
Th ese include power distance, uncertainty avoidance, individualism versus collectivism, and
masculinity versus femininity. Th e fi rst dimension, power distance, addresses how the dis-
tribution of social power is dealt with in the culture. In cultures with a high power distance,
43 See Richarde E. Nisbett, Th e Geography of Th ought: How Asians and Westerners Th ink Diff erently … And Why
(New York: Free Press, 2003).
International and Cultural Issues and User Interface Design 409
members of the culture believe in the authority of the social hierarchy. In cultures with low
power distance, members of the culture believe that power should be more equally distrib-
uted. Consequently, in cultures with a high power distance, emphasis on the “greatness” of
the leaders of the fi rm, the use of “proper titles” for members of the fi rm, and the posting
of testimonials on behalf of the fi rm by “prominent” members of society is important.
International awards won by the fi rm, its members, or its products should also be posted
prominently on the website.
Th e second dimension, uncertainty avoidance, addresses to what degree a culture is
comfortable with uncertainty. In a culture with a high uncertainty avoidance, members avoid
taking risks, value tradition, and are much more comfortable in a rule-driven society. In cul-
tures that score high on uncertainty avoidance, more customer service needs to be provided,
more important “local” contacts need to be available, the fi rm’s and product’s history and
tradition need to be provided on the website, and, in the case of soft ware, the use of free trials
and downloads is critical. In other words, you need to build trust and reduce perceived risk
between the customer and the fi rm. Th is can be supported through product seals of approval
or the use of WebTrustTM and SysTrustTM certifi cations for the website.44 Merely translating
a website from a low uncertainty avoidance culture to a high uncertainty avoidance culture
is not suffi cient. You also need to point out relationships between the local culture and the
fi rm’s products.
Th e third dimension, individualism versus collectivism, is based on the level of empha-
sis the culture places on the individual or the collective, or group. In North America and
Europe, individualism is rewarded. However, in East Asia, it is believed that by focusing
on optimizing the group, the individual will be most successful. In other words, it is the
group that is the most important. In a collective society, presenting information on how
the fi rm “gives back” to the community; supports “member” clubs, “loyalty” programs,
and “chat” facilities; and provides links to “local” sites of interest are very important
characteristics for a website. In contrast, in an individualistic society, providing support
for personalization of the user’s experience with the website, emphasizing the uniqueness
of the products that the user is viewing, and emphasizing the privacy policy of the site
are critical.
Hofstede’s fourth dimension, masculinity versus femininity, does not mean how men
and women are treated by the culture. But, instead this dimension addresses how well mas-
culine and feminine characteristics are valued by the culture. For example, in a masculine
culture, characteristics such as being assertive, ambitious, aggressive, and competitive are
valued, whereas in a feminine culture, characteristics such as being encouraging, compas-
sionate, thoughtful, gentle, and cooperative are valued. In masculine cultures, a focus on the
eff ectiveness of the fi rm’s products is essential. Also, clearly separating male- and female-ori-
ented topics and placing them on diff erent sections of a website can be critical. According to
Singh and Pereira, feminine cultures value a focus on aesthetics and using more of a soft -sell
approach, where the focus on more aff ective, intangible aspects of the fi rm, its members, and
its products is more appropriate.
Obviously, operationalizing Hall’s and Hofstede’s dimensions for eff ective user interface
design is not easy. Furthermore, given all of the diff erent platforms on which a user interface
can be deployed, the level of complexity and diffi culty in designing eff ective and effi cient user
interfaces that take into consideration the global and multicultural world in which we live
is increasing. However, in a global market, ignoring cultural issues in user interface design,
44 See www.webtrust.org
410 Chapter 10 Human–Computer Interaction Layer Design
whether it is for an internal system used only by employees of the fi rm or an external system
that is used by customers, will most certainly cause a system to fail. Th is is especially true
when you consider mobile and social media sites.
NONFUNCTIONAL REQUIREMENTS AND HUMAN–COMPUTER
INTERACTION LAYER DESIGN
Th e human–computer interaction layer is heavily infl uenced by nonfunctional require-
ments. In this chapter, we dealt with issues such as layout of the user interface, awareness
of content, aesthetics, user experience, and consistency. We also have provided informa-
tion on how to design the navigation, inputs, and outputs of the user interface. Finally,
we have considered mobile computing, social media, immersive and multidimensional
environments, and international and cultural issues in user-interface design. None of
these have anything to do with the functional requirements of the system. However, if
they are ignored, the system can be unusable. As with the data management layer, there
are four primary types of nonfunctional requirements that can be important in designing
the human–computer interaction layer: operational, performance, security, and cultural
and political requirements.
Operational requirements, such as choice of hardware and soft ware platforms, infl uence
the design of the human–computer interaction layer. For example, something as simple as the
number of buttons on a mouse (one, two, three, or more) changes the interaction that the user
will experience. Other operational nonfunctional requirements that can infl uence the design
of the human–computer interaction layer include system integration and portability. In these
cases, a Web-based solution may be required, which can eff ect the design; not all features of
a user interface can be implemented effi ciently and eff ectively on the Web. Th is can require
additional user interface design. Obviously, the entire area of mobile computing can eff ect the
success or failure of the system.
Performance requirements, over time, have become less of an issue for this layer.
However, speed requirements are still paramount, especially with mobile computing. Most
users do not care for hitting return or clicking the mouse and having to take a coff ee break
while they are waiting for the system to respond, so effi ciency issues must be still addressed.
Depending on the user interface toolkit used, diff erent user interface components may be
required. Furthermore, the interaction of the human–computer interaction layer with the
other layers must be considered. For example, if the system response is slow, incorporating
more-effi cient data structures with the problem domain layer, including indexes in the tables
with the data management layer, and/or replicating objects across the physical architecture
layer could be required.
Security requirements aff ecting the human–computer interaction layer deal primarily
with the access controls implemented to protect the objects from unauthorized access.
Most of these controls are enforced through the DBMS on the data management layer and
the operating system on the physical architecture layer. However, the human–computer
interaction layer design must include appropriate log-on controls and the possibility of
encryption.
In addition to the international and cultural issues described previously, unstated
norms eff ect the cultural and political requirements that can eff ect the design of the human–
computer interaction layer. Unstated norm requirements include having the date displayed
in the appropriate format (MM/DD/YYYY versus DD/MM/YYYY). For a system to be truly
useful in a global environment, the user interface must be customizable to address local
cultural requirements.
Key Terms 411
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the six basic principles of user interface design.
Apply the use-case driven process described to design a user interface.
Describe the purpose of use scenarios in user interface design.
Describe how to use windows navigation diagrams, windows layout diagrams, storyboards, and user interface
prototypes during the design of a user interface.
Describe the diff erence between essential and real use cases.
Describe the importance and use of interface standards in user interface design.
Describe the four common approaches used to evaluate user interfaces.
Discuss the relationship between user interface design and requirements determination.
Design effi cient and eff ective navigation controls that are easy to use, prevent users from making mistakes, support
obvious approaches for users to recover from mistakes, and use a consistent grammar order.
Design effi cient and eff ective input mechanisms that capture the necessary information for the system.
Design effi cient and eff ective output that supports the users in their tasks.
Describe the unique issues related to designing user interfaces for mobile computing platforms.
Describe the unique navigation controls, input mechanisms, and outputs that mobile computing platforms possess.
Describe the unique issues related to designing user interfaces for social applications.
Describe the unique issues related to designing user interfaces for immersive and multidimensional applications.
Discuss the international and cultural issues that can aff ect the design of the human–computer interaction layer.
Describe how nonfunctional requirements may influence the actual design of the human–computer
interaction layer.
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Th e design of the human–computer interaction layer is particularly important because to the
user, the interface is the system. Th e challenge for the team was ensuring that the interface
will display accurately on devices of various sizes. Next, they developed the WND for the
Mobile Scheduling phase of the system and developed the design prototypes for the interface.
Aft er completing these tasks, the team developed the navigation documentation artifacts.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
KEY TERMS
Acknowledgment
message
Action-object order
Aesthetics
Augmented reality (AR)
Bar-code reader
Batch processing
Batch report
Bias
Button
Check box
Check digit check
Cognitive map
Collaboration
Collectivism
Color
Combo box
Command language
Completeness check
Confi rmation message
Consistency
Consistency check
Content awareness
Context
Cultural diff erences
2D space
Database check
Data-entry operator
Default value
Delay message
Density
Detail report
412 Chapter 10 Human–Computer Interaction Layer Design
Direct manipulation
Drill-down capability
Drop-down list box
Drop-down menu
Ease of learning
Ease of use
Edit check
Error message
Essential use case
Exception report
Femininity
Field
Field label
Form
Format check
Gamifi cation
GPS
Grammar order
Graph
Graphical user interface
(GUI)
Haptic feedback
Help message
Heuristic evaluation
High context
Hot key
Image map
Immersion
Individualism
Information load
Input mechanism
Interactive evaluation
Interface action
Interface design
prototype
Interface evaluation
Interface icon
Interface metaphor
Interface object
Interface standards
Interface template
Layout
Low context
Magnetic stripe readers
Masculinity
Menu
Menu bar
Mobile device
Monochronic time
Multilingual requirements
Multidimensional
information visualization
Natural language
Navigation controls
Navigation mechanism
Nonimmersive 3D
Number box
Object-action order
Object recognition
Occlusion
Online processing
On-screen list box
Optical character
recognition
Output mechanism
Polychronic time
Pop-up menu
Power distance
Pull
Push
Radio button
Range check
Real-time information
Real-time report
Real use case
Report
Screen
Selection box
Sequence diagrams
Slider
Smart card
Smartphone
Social media
Source data automation
Speed of messages
State
Stereotype
Storyboard
Summary report
System interface
Tab menu
Tablet
Text box
Th ree-clicks rule
Time
Toolbar
Touch screens
Transaction processing
Transition
Turnaround document
Uncertainty avoidance
Usability testing
Use case
Use scenario
User experience
User interface
User interface prototype
Validation
Virtual reality (VR)
Wayfi nding
Walkthrough evaluation
White space
Window
Windows layout diagram
Window navigation
diagram (WND)
QUESTIONS
1. Explain three important user interface design
principles.
2. What are three fundamental parts of most user inter-
faces?
3. Why is content awareness important?
4. What is white space, and why is it important?
5. Under what circumstances should densities be low?
High?
6. How can a system be designed to be used by both
experienced and fi rst-time users?
7. Why is consistency in design important? Why can too
much consistency cause problems?
8. How can diff erent parts of the interface be consistent?
9. Describe the basic process of user interface design.
10. What are use cases, and why are they important?
11. What is a WND, and why is it used?
12. Why are interface standards important?
13. Explain the purpose and contents of interface meta-
phors, interface objects, interface actions, interface
icons, and interface templates.
14. Why do we prototype the user interface design?
15. Why is it important to perform an interface evaluation
before the system is built?
16. Compare and contrast the four types of interface
evaluation.
17. Under what conditions is heuristic evaluation justi-
fi ed?
18. What are Krug’s three design principles?
19. Describe three basic principles of navigation design.
20. How can you prevent mistakes?
21. Explain the diff erences between object-action order
and action-object order.
Exercises 413
22. Describe four types of navigation controls
23. Why are menus the most commonly used navigation
control?
24. Compare and contrast four types of menus.
25. Under what circumstances would you use a drop-
down menu versus a tab menu?
26. Under what circumstances would you use an image
map versus a simple list menu?
27. Describe fi ve types of messages.
28. What are the key factors in designing an error mes-
sage?
29. What is context-sensitive help? Does your word pro-
cessor have context-sensitive help?
30. How do an essential use case and a real use case diff er?
31. What is the relationship between essential use cases
and use scenarios?
32. What is the relationship between real use cases and
use scenarios?
33. Explain three principles in the design of inputs.
34. Compare and contrast batch processing and online
processing. Describe one application that would
use batch processing and one that would use online
processing.
35. Why is capturing data at the source important?
36. Describe four devices that can be used for source data
automation.
37. Describe fi ve types of inputs.
38. Why is input validation important?
39. Describe fi ve types of input validation methods.
40. Explain three principles in the design of outputs.
41. Describe fi ve types of outputs.
42. What do you think are three common mistakes that
novice analysts make in navigation design?
43. What do you think are three common mistakes that
novice analysts make in input design?
44. What do you think are three common mistakes that
novice analysts make in output design?
45. What are the six challenges you face when developing
mobile applications?
46. What are the six suggestions to address the mobile
computing challenges?
47. What are the unique navigation controls, input mech-
anisms, and outputs that mobile computing supports?
48. With regard to social media, what is the diff erence
between “push” and “pull” approaches to interacting
with customers?
49. Why is it important to keep your social media sites
synced?
50. How can you keep your customers engaged with your
social media sites?
51. Why do people play games?
52. What is gamifi cation?
53. What is occlusion? Why is it an issue when developing
multidimensional information visualizations? Aug-
mented reality systems? Virtual reality systems?
54. What is augmented reality?
55. Name some of potential business applications of aug-
mented reality.
56. What is virtual reality?
57. Name some of potential business applications of vir-
tual reality.
58. When developing a virtual reality system, what are
some of the issues that need to be addresses?
59. What is a cognitive map?
60. What are some of the multilingual issues that you may
face when developing for a global audience?
61. How important is the proper use of color when
developing websites for a global audience? Give some
examples of potential pitfalls that you could run into.
62. Name the three cultural dimensions that are relevant
to user interface design identifi ed by Hall. Why are
they relevant?
63. Name the four cultural dimensions that are relevant to
user interface design identifi ed by Hofstede. Why are
they relevant?
64. What are some of the nonfunctional requirements
that can infl uence the design of the human–computer
interaction layer?
EXERCISES
A. Develop two use scenarios for a website that sells some
retail products (e.g., books, music, clothes).
B. Create a storyboard for a website that sells some retail
products (e.g., books, music, clothes).
C. Draw a WND for a website that sells some retail prod-
ucts (e.g., books, music, clothes).
D. Create a windows layout diagram for the home page
of a website that sells some retail products (e.g., books,
music, clothes).
E. Describe the primary components of the interface
standards for a website that sells some retail products
(metaphors, objects, actions, icons, and template).
414 Chapter 10 Human–Computer Interaction Layer Design
F. Using the Web, identify a set of games that are useful
in some aspect of business, e.g., advertising or training.
G. Using the Web, identify a set of multidimensional
information visualizations that are used to support
business decision-making.
H. Using the Web, fi nd businesses that are currently
using augmented and virtual reality systems.
I. For the A Real Estate Inc. problem in Chapter 4 (exer-
cises I, J, and K), Chapter 5 (exercises P and Q), Chap-
ter 6 (exercise D), Chapter 7 (exercise A), Chapter 8
(exercise A), and Chapter 9 (exercise L):
1. Develop two use scenarios.
2. Draw a WND.
3. Design a storyboard.
J. Based on your solution to exercise I:
1. Create windows layout diagrams for the interface
design.
2. Develop a real use case.
K. For the A Video Store problem in Chapter 4 (exercises
L, M, and N), Chapter 5 (exercises R and S), Chapter 6
(exercise E), Chapter 7 (exercise B), Chapter 8 (exer-
cise B), and Chapter 9 (exercise M):
1. Develop two use scenarios.
2. Draw a WND.
3. Design a storyboard.
L. Based on your solution to exercise K:
1. Create windows layout diagrams for the interface
design.
2. Develop a real use case.
M. For the gym membership problem in Chapter 4
(exercises O, P, and Q), Chapter 5 (exercises T and U),
Chapter 6 (exercise F), Chapter 7 (exercise C), Chapter 8
(exercise C), and Chapter 9 (exercise N):
1. Develop two use scenarios.
2. Draw a WND.
3. Design a storyboard.
N. Based on your solution to exercise M:
1. Create windows layout diagrams for the interface
design.
2. Develop a real use case.
O. For the Picnics R Us problem in Chapter 4 (exercises
R, S, and T), Chapter 5 (exercises V and W), Chapter 6
(exercise G), Chapter 7 (exercise D), Chapter 8 (exer-
cise D), and Chapter 9 (exercise O):
1. Develop two use scenarios.
2. Draw a WND.
3. Design a storyboard.
P. Based on your solution to exercise O:
1. Create windows layout diagrams for the interface
design.
2. Develop a real use case.
Q. For the Of-the-Month-Club problem in Chapter 4
(exercises U, V, and W), Chapter 5 (exercises X and
Y), Chapter 6 (exercise H), Chapter 7 (exercise E),
Chapter 8 (exercise E), and Chapter 9 (exercise N):
1. Develop two use scenarios.
2. Draw a WND.
3. Design a storyboard.
R. Based on your solution to exercise Q:
1. Create windows layout diagrams for the interface
design.
2. Develop a real use case.
S. Create a user interface design for a mobile solution
for the:
1. A Real Estate Inc. problem.
2. A Video Store problem.
3. Gym membership problem.
4. Picnics R Us problem.
5. Of-the-Month-Club problem.
T. How would your answers change to exercises I through
S if you were developing for a global marketplace?
MINICASES
1. Tots to Teens is a catalog retailer specializing in
children’s clothing. A project has been under way
to develop a new order entry system for the com-
pany’s catalog clerks. Th e old system had a charac-
ter-based user interface that corresponded to the
system’s COBOL underpinnings. Th e new system will
feature a graphical user interface more in keeping with
up-to-date PC products in use today. Th e company
hopes that this new user interface will help reduce the
turnover it has experienced with its order entry clerks.
Many newly hired order entry staff found the old sys-
tem very diffi cult to learn and were overwhelmed by
the numerous mysterious codes that had to be used to
communicate with the system.
A user interface walkthrough evaluation was
scheduled for today to give the user a fi rst look at
the new system’s interface. Th e project team was
careful to invite several key users from the order
Minicases 415
entry department. In particular, Norma was included
because of her years of experience with the order
entry system. Norma was known to be an informal
leader in the department; her opinion infl uenced
many of her associates. Norma had let it be known
that she was less than thrilled with the ideas she had
heard for the new system. Owing to her experience
and good memory, Norma worked very eff ectively
with the character-based system and was able to
breeze through even the most convoluted transac-
tions with ease. Norma had trouble suppressing a
sneer when she heard talk of such things as “icons”
and “buttons” in the new user interface.
Cindy was also invited to the walkthrough because
of her infl uence in the order entry department. Cindy
has been with the department for just one year, but
she quickly became known because of her successful
organization of a sick child daycare service for the
children of the department workers. Sick children are
the number-one cause of absenteeism in the depart-
ment, and many of the workers could not aff ord to
miss workdays. Never one to keep quiet when a sit-
uation needed improvement, Cindy has been a vocal
supporter of the new system.
a. Drawing upon the design principles presented in
the text, describe the features of the user interface
that will be most important to experienced users
like Norma.
b. Drawing upon the design principles presented in
the text, describe the features of the user interface
that will be most important to novice users like
Cindy.
2. Th e members of a systems development project team
have gone out for lunch together, and as oft en hap-
pens, the conversation turns to work. Th e team has
been working on the development of the user inter-
face design, and so far, work has been progressing
smoothly. Th e team should be completing work on
the interface prototypes early next week. A combi-
nation of storyboards and language prototypes has
been used in this project. Th e storyboards depict the
overall structure and fl ow of the system, but the team
developed language prototypes of the actual screens
because they felt that seeing the actual screens would
be valuable for the users.
Chris (the youngest member of the project team):
I read an article last night about a really cool way to
evaluate a user interface design. It’s called usabil-
ity testing, and it’s done by all the major software
vendors. I think we should use it to evaluate our
interface design.
Heather (systems analyst): I’ve heard of that, too,
but isn’t it really expensive?
Mark (project manager): I’m afraid it is expensive
and I’m not sure we can justify the expense for this
project.
Chris: But we really need to know that the interface
works. I thought this usability testing technique would
help us prove we have a good design.
Amy (systems analyst): It would, Chris, but there
are other ways too. I assumed we’d do a thorough
walkthrough with our users and present the interface
to them at a meeting. We can project each interface
screen so that the users can see it and give us their
reaction. Th is is probably the most effi cient way to get
the users’ response to our work.
Heather: Th at’s true, but I’d sure like to see the
users sit down and work with the system. I’ve always
learned a lot by watching what they do, seeing where
they get confused, and hearing their comments and
feedback.
Ryan (systems analyst): It seems to me that we’ve
put so much work into this interface design that all we
really need to do is review it ourselves. Let’s just make
a list of the design principles we’re most concerned
about and check it ourselves to make sure we’ve fol-
lowed them consistently. If we have, we should be fi ne.
We want to get moving on the implementation, you
know.
Mark: Th ese are all good ideas. It seems like we’ve
all got a diff erent view of how to evaluate the interface
design. Let’s try to sort out the technique that’s best for
our project.
Develop a set of guidelines that can help a project
team like this one select the most appropriate interface
evaluation technique for their project.
3. Th e menu structure for Holiday Travel Vehicle’s
existing character-based system is shown here. Develop
and prototype a new interface design for the system’s
functions using a graphical user interface. Also, develop
a set of real use cases for your new interface. Assume
the new system will need to include the same functions
as those shown in the menus provided. Include any
messages that will be produced as a user interacts with
your interface (error, confi rmation, status, etc.). Also,
prepare a written summary that describes how your
interface implements the principles of good interface
design as presented in the textbook.
416 Chapter 10 Human–Computer Interaction Layer Design
Holiday Travel Vehicles
Main Menu
1 Sales Invoice
2 Vehicle Inventory
3 Reports
4 Sales Staff
Type number of menu selection here:____
Holiday Travel Vehicles
Sales Invoice Menu
1 Create Sales Invoice
2 Change Sales Invoice
3 Cancel Sales Invoice
Type number of menu selection here:____
Holiday Travel Vehicles
Vehicle Inventory Menu
1 Create Vehicle Inventory Record
2 Change Vehicle Inventory Record
3 Delete Vehicle Inventory Record
Type number of menu selection here:____
Holiday Travel Vehicles
Reports Menu
1 Commission Report
2 RV Sales by Make Report
3 Trailer Sales by Make Report
4 Dealer Options Report
Type number of menu selection here:____
Holiday Travel Vehicles
Sales Staff Maintenance Menu
1 Add Salesperson Record
2 Change Salesperson Record
3 Delete Salesperson Record
Type number of menu selection here:____
4. One aspect of the new system under development at
Holiday Travel Vehicles will be the direct entry of the
sales invoice into the computer system by the sales-
person as the purchase transaction is being completed.
In the current system, the salesperson fi lls out a paper
form (shown on the next page).
Design and prototype an input screen that will per-
mit the salesperson to enter all the necessary informa-
tion for the sales invoice. Th e following information
may be helpful in your design process. Assume that
Holiday Travel Vehicles sells recreational vehicles and
trailers from four diff erent manufacturers. Each man-
ufacturer has a fi xed number of names and models of
RVs and trailers.
For the purposes of your prototype, use this format:
Mfg-A Name-1 Model-X
Mfg-A Name-1 Model-Y
Mfg-A Name-1 Model-Z
Mfg-B Name-1 Model-X
Mfg-B Name-1 Model-Y
Mfg-B Name-2 Model-X
Mfg-B Name-2 Model-Y
Mfg-B Name-2 Model-Z
Mfg-C Name-1 Model-X
Mfg-C Name-1 Model-Y
Mfg-C Name-1 Model-Z
Mfg-C Name-2 Model-X
Mfg-C Name-3 Model-X
Mfg-D Name-1 Model-X
Mfg-D Name-2 Model-X
Mfg-D Name-2 Model-Y
Also, assume there are ten diff erent dealer options that
could be installed on a vehicle at the customer’s request.
Th e company currently has ten salespeople on staff .
Minicases 417
5. Refer to the Professional and Scientifi c Staff Manage-
ment (PSSM) Minicase in Chapters 4, 6, 7, 8, and 9.
a. Develop two use scenarios, draw a WND, and
design a storyboard.
b. Based on your answers to part a, create windows
layout diagrams for the user interface and develop
a set of real use cases for the user interface.
c. How would your user interface design have to be
modifi ed if you were to deploy it on a tablet? What
about a smartphone?
d. What, if any, social media sites should PSSM con-
sider?
e. How would your answers change if you were devel-
oping the system for a global audience?
Holiday Travel Vehicles
Sales Invoice Invoice #: ____________
Invoice Date: _________
Customer Name: _____________________________________
Address: _____________________________________
City: _____________________________________
State: _____________________________________
Zip: _____________________________________
Phone: _____________________________________
New RV/TRAILER
(circle one) Name: _____________________________________
Model: _____________________________________
Serial #: ______________________ Year: _________
Manufacturer: _____________________________________
Trade-in RV/TRAILER
(circle one) Name: _____________________________________
Model: _____________________________________
Year: _____________________________________
Manufacturer: _____________________________________
Options: Code Description Price
Vehicle Base Cost: ________________
Trade-in Allowance: ________________ (Salesperson Name)
Total Options: ________________
Tax: ________________
License Fee: ________________
Final Cost: ________________ (Customer Signature)
418
C H A P T E R 1 1
Physical Architecture
Layer Design
An important component of the design of an information system is the design of the
physical architecture layer, which describes the system’s hardware, soft ware, and network
environment. Th e physical architecture layer design fl ows primarily from the nonfunctional
requirements, such as operational, performance, security, cultural, and political require-
ments. Th e deliverable from the physical architecture layer design includes the architecture
and the hardware and soft ware specifi cation.
OBJECTIVES
■ Understand the diff erent physical architecture components.
■ Understand server-based, client-based, and client–server physical architectures.
■ Be familiar with cloud computing, ubiquitous computing and the Internet of things
(IoT), and Green IT.
■ Be able to create a network model using a deployment diagram.
■ Be familiar with how to create a hardware and soft ware specifi cation.
■ Understand how operational, performance, security, cultural, and political requirements
aff ect the design of the physical architecture layer.
INTRODUCTION
In today’s environment, most information systems are spread across multiple computers. A
Web-based system, for example, runs in the browser on a desktop computer but interacts
with the Web server (and possibly other computers) over the Internet. A system that operates
completely inside a company’s network may have a Visual Basic program installed on one
computer but interact with a database server elsewhere on the network. Th erefore, an impor-
tant step of design is the creation of the physical architecture layer design, the plan for how
the system will be distributed across the computers, and what hardware and soft ware will be
used for each computer.
In many cases, systems are built to use the existing hardware and soft ware in the organi-
zation. Th erefore, the current architecture restricts the choice. Other factors such as corporate
standards, existing site-licensing agreements, and product–vendor relationships also can
mandate what architecture, hardware, and soft ware the project team must use. However,
many organizations now have a variety of infrastructures available or are openly looking for
pilot projects to test new architectures that enable a project team to select one on the basis of
other important factors.
Elements of the Physical Architecture Layer 419
Designing the physical architecture layer can be quite diffi cult; therefore, many organiza-
tions hire expert consultants or assign very experienced analysts to the task.1 In this chapter,
we examine the key factors in physical architecture layer design, but it is important to remem-
ber that it takes lots of experience to do it well. Th e nonfunctional requirements developed
during analysis (see Chapter 3) play a key role in physical architecture layer design. Th ese
requirements are reexamined and refi ned into more-detailed requirements that infl uence the
system’s architecture.
ELEMENTS OF THE PHYSICAL ARCHITECTURE LAYER
Th e objective of designing the physical architecture layer is to determine what parts of the
application soft ware will be assigned to what hardware. Although there are numerous ways
the soft ware components can be placed on the hardware components, there are three prin-
cipal application architectures in use today: server-based architectures, client-based architec-
tures, and client–server architectures.
Architectural Components
Th e major architectural components of any system are the soft ware and the hardware. Th e
major soft ware components of the system being developed have to be identifi ed and then
allocated to the various hardware components on which the system will operate. Each of these
components can be combined in a variety of diff erent ways.
All soft ware systems can be divided into four basic functions. Th e fi rst is data storage (asso-
ciated with the object persistence located on the data management layer—see Chapter 9). Most
application programs require data to be stored and retrieved, whether the information is a small
fi le such as a memo produced by a word processor or a large database that stores an organiza-
tion’s accounting records. Th ese are the data documented in the structural model (CRC cards
and class diagrams). Th e second function is data access logic (associated with the data access and
manipulation classes located on the data management layer—see Chapter 9), the processing
required to access data, which oft en means database queries in SQL (structured query language).
Th e third function is the application logic (located on the problem domain layer—see Chapters 4
through 8), which can be simple or complex, depending on the application. Th is is the logic doc-
umented in the functional (activity diagrams and use cases) and behavioral models (sequence,
communication, and behavioral state machines). Th e fourth function is the presentation logic
(located on the human–computer interaction layer—see Chapter 10), the presentation of infor-
mation to the user, and the acceptance of the user’s commands (the user interface). Th ese four
functions (data storage, data access logic, application logic, and presentation logic) are the basic
building blocks of any application.
Th e three primary hardware components of a system are client computers, servers, and the
network that connects them. Client computers are the input/output devices employed by the user
and are usually desktop or laptop computers, but they can also be handheld devices, cell phones,
special-purpose terminals, and so on. Servers are typically larger computers that are used to store
soft ware and hardware that can be accessed by anyone who has permission. Th e network that
connects the computers can vary in speed from a slow cell phone, to medium-speed always-on
frame relay networks, to fast always-on broadband connections such as cable modem, DSL, or
T1 circuits, to high-speed always-on ethernet, T3, or ATM circuits.2
1 For more information on the physical architecture layer, see Irv Englander, Th e Architecture of Computer Hardware
and Systems Soft ware: An Information Technology Approach, 5th Ed. (Hoboken, NJ: Wiley, 2014); Kalani Kirk Hausman
and Susan L. Cook, IT Architecture for DummiesTM (Hoboken, NJ: Wiley, 2011).
2 For more information on networks, see Alan Dennis, Networking in the Internet Age (New York: Wiley, 2002).
420 Chapter 11 Physical Architecture Layer Design
Server-Based Architectures
Th e very fi rst computing architectures were server-based architectures, with the server per-
forming all four functions. Th e clients enabled users to send and receive messages to and from
the server. Th e clients merely captured keystrokes and sent them to the server for processing
and accepted instructions from the server on what to display (see Figure 11-1).
Th is very simple architecture oft en works very well. Application soft ware is developed
and stored on one computer, and all data are on the same computer. Th ere is one point of
control, because all messages fl ow through the one central server. Th e fundamental problem
with server-based networks is that the server must process all messages. As the demands for
more and more applications grow, many server computers become overloaded and unable
to quickly process all the users’ demands. Response time becomes slower, and network
managers are required to spend increasingly more money to upgrade the server computer.
Unfortunately, upgrades come in large increments and are expensive; it is diffi cult to upgrade
“a little.”
Client-Based Architectures
With client-based architectures, the clients are personal computers on a local area network
(LAN), and the server computer is a server on the same network. Th e application soft ware on
the client computers is responsible for the presentation logic, the application logic, and the
data access logic; the server simply stores the data (see Figure 11-2).
FIGURE 11-1
Server-Based
Architecture
Server Host
(mainframe computer)
Client /(terminal)
Presentation logic
Application logic
Data access logic
Data storage
FIGURE 11-2
Client-Based
Architectures
Client
(microcomputer)
Presentation logic
Application logic
Data access logic
Server
(microcomputer)
Data storage
Elements of the Physical Architecture Layer 421
Th is simple architecture also oft en works well. However, as the demands for more and
more network applications grow, the network circuits can become overloaded. Th e funda-
mental problem in client-based networks is that all data on the server must travel to the
client for processing. For example, suppose the user wishes to display a list of all employees
with company life insurance. All the data in the database must travel from the server where
the database is stored over the network to the client, which then examines each record to see
whether it matches the data requested by the user. Th is can overload both the network and
the power of the client computers.
Client–Server Architectures
Most organizations today use client–server architectures, which attempt to balance the
processing between the client and the server by having both do some of the application
functions. In these architectures, the client is responsible for the presentation logic, whereas
the server is responsible for the data access logic and data storage. Th e application logic
may reside on either the client or the server or be split between both (see Figure 11-3). Th e
client shown in Figure 11-3 can be referred to as a thick, or fat, client if it contains the bulk
of application logic. A current practice is to create client–server architectures using thin cli-
ents because there is less overhead and maintenance in supporting thin-client applications.
For example, many Web-based systems are designed with the Web browser performing
presentation, with only minimal application logic using programming languages like Java
and the Web server having the application logic, data access logic, and data storage.
Client–server architectures have four important benefi ts. First, they are scalable. Th at
means it is easy to increase or decrease the storage and processing capabilities of the servers.
If one server becomes overloaded, you simply add another server so that many servers are
used to perform the application logic, data access logic, or data storage. Th e cost to upgrade
is much more gradual, and you can upgrade in smaller steps rather than spending hundreds
of thousands to upgrade a mainframe server.
Client–server architectures can support many diff erent types of clients and servers. It is
possible to connect computers that use diff erent operating systems so that users can choose
which type of computer they prefer (e.g., combining both Windows computers and Apple
Macintoshes on the same network). We are not locked into one vendor, as is oft en the case with
server-based networks. Middleware is a type of system soft ware designed to translate between
diff erent vendors’ soft ware. Middleware is installed on both the client computer and the server
computer. Th e client soft ware communicates with the middleware, which can reformat the
message into a standard language that can be understood by the middleware assisting the
server soft ware.
For thin-client server architectures that use Internet standards, it is simple to clearly
separate the presentation logic, the application logic, and the data access logic and design so
FIGURE 11-3
Client–Server
Architecture
Client
(microcomputer)
Server
(micro, mini, or mainframe)
Presentation logic
Application logic
Data access logic
Data storage
422 Chapter 11 Physical Architecture Layer Design
that each is somewhat independent. For example, the presentation logic can be designed in
HTML or XML to specify how the page will appear on the screen (see Chapter 10). Simple
program statements are used to link parts of the interface to specifi c application logic mod-
ules that perform various functions. Th ese HTML or XML fi les defi ning the interface can be
changed without aff ecting the application logic. Likewise, it is possible to change the applica-
tion logic without changing the presentation logic or the data, which are stored in databases
and accessed using SQL commands.
Finally, because no single server computer supports all the applications, the network is
generally more reliable. Th ere is no central point of failure that will halt the entire network if
it fails, as there is in server-based computing. If any one server fails in a client–server envi-
ronment, the network can continue to function using all the other servers (but, of course, any
applications that require the failed server will not work).
Client–server architectures also have some critical limitations, the most important of
which is its complexity. All applications in client–server computing have two parts, the soft –
ware on the client and the soft ware on the server. Writing this soft ware is more complicated
than writing the traditional all-in-one soft ware used in server-based architectures. Updating
the network with a new version of the soft ware is more complicated, too. In server-based
architectures, there is one place where application soft ware is stored; to update the soft ware,
we simply replace it there. With client–server architectures, we must update all clients and
all servers.
Much of the debate about server-based versus client–server architectures has centered on
cost. One of the great claims of server-based networks in the 1980s was that they provided econ-
omies of scale. Manufacturers of big mainframes claimed it was cheaper to provide computer
services on one big mainframe than on a set of smaller computers. Th e personal computer revolu-
tion changed this. Since the 1980s, the cost of personal computers has continued to drop, whereas
their performance has increased signifi cantly. Today, personal computer hardware is more than
1,000 times cheaper than mainframe hardware for the same amount of computing power.
With cost differences like these, it is easy to see why there has been a sudden rush
to microcomputer-based client–server computing. The problem with these cost compar-
isons is that they ignore the total cost of ownership, which includes factors other than
obvious hardware and software costs. For example, many cost comparisons overlook the
increased complexity associated with developing application software for client–server
networks. Most experts believe that it costs four to five times more to develop and
maintain application software for client–server computing than it does for server-based
computing.
Client–Server Tiers
Th ere are many ways the application logic can be partitioned between the client and the
server. Th e example in Figure 11-3 is one of the most common. In this case, the server is
responsible for the data, and the client is responsible for the application and presentation.
Th is is called a two-tiered architecture because it uses only two sets of computers, clients, and
servers.
A three-tiered architecture uses three sets of computers (see Figure 11-4). In this case, the
soft ware on the client computer is responsible for presentation logic, an application server (or
servers) is responsible for the application logic, and a separate database server (or servers) is
responsible for the data access logic and data storage.
An n-tiered architecture uses more than three sets of computers. In this case, the client
is responsible for presentation, database servers are responsible for the data access logic and
data storage, and the application logic is spread across two or more diff erent sets of servers.
Elements of the Physical Architecture Layer 423
Th is type of architecture is common in today’s e-commerce systems (see Figure 11-5). Th e
fi rst component is the Web browser on the client computer employed by a user to access the
system and enter commands (presentation logic). Th e second is a Web server that responds
to the user’s requests, either by providing (HTML) pages and graphics (application logic) or
by sending the request to the third component on another application server that performs
various functions (application logic). Th e fourth component is a database server that stores
all the data (data access logic and data storage). Each of these four components is separate,
making it easy to spread the diff erent components on diff erent servers and to partition the
application logic on two diff erent servers.
Th e primary advantage of an n-tiered client–server architecture compared with a two-
tiered architecture (or a three-tiered architecture with a two-tiered architecture) is that it sep-
arates the processing that occurs to better balance the load on the diff erent servers; it is more
scalable. In Figure 11-5, we have three separate servers, a confi guration that provides more
power than if we had used a two-tiered architecture with only one server. If we discover that
FIGURE 11-4
Three-Tiered
Client–Server
Architecture
Client
(microcomputer)
Presentation logic
Application server
(microcomputer)
Database server
(micro, mini, or mainframe)
Application logic Data access logic
Data storage
Client
(microcomputer)
Presentation logic
Web server
(micro, mini, or mainframe)
Web-related
Application logic
Data access logic
Data storage
Application server
(micro, mini, or mainframe)
Database server
(micro, mini, or mainframe)
Non–Web-related
Application logic
FIGURE 11-5
Four-Tiered
Client–Server
Architecture
424 Chapter 11 Physical Architecture Layer Design
the application server is too heavily loaded, we can simply replace it with a more powerful
server or just put in several more application servers. Conversely, if we discover the database
server is underused, we could store data from another application on it.
Th ere are two primary disadvantages to an n-tiered architecture compared with a two-
tiered architecture (or a three-tiered architecture with a two-tiered architecture). First, the
confi guration puts a greater load on the network. If you compare Figures 11-3, 11-4, and
11-5, you will see that the n-tiered model requires more communication among the servers;
it generates more network traffi c, so you need a higher-capacity network. It is also much
more diffi cult to program and test soft ware in n-tiered architectures than in two-tiered
architectures because more devices have to communicate to complete a user’s transaction.
Selecting a Physical Architecture
Most systems are built to use the existing infrastructure in the organization, so oft en the
current infrastructure restricts the choice of architecture. For example, if the new system
will be built for a mainframe-centric organization, a server-based architecture may be the
best option. Other factors such as corporate standards, existing licensing agreements, and
product/vendor relationships can also mandate what architecture the project team needs to
design. However, many organizations now have a variety of infrastructures available or are
openly looking for pilot projects to test new architectures and infrastructures, enabling a
project team to select an architecture based on other important factors.
Each of the computing architectures just discussed has its strengths and weaknesses, and
no architecture is inherently better than the others. Th us, it is important to understand the
strengths and weaknesses of each computing architecture and when to use each. Figure 11-6
presents a summary of the important characteristics of each.
Cost of Infrastructure One of the strongest driving forces to client–server architectures is
cost of infrastructure (the hardware, soft ware, and networks that will support the application
system). Simply put, personal computers are more than 1,000 times cheaper than mainframes
for the same amount of computing power. Th e personal computers on our desks today have
more processing power, memory, and hard disk space than the typical mainframe of the past,
and the cost of the personal computers is a fraction of the cost of the mainframe.
Th erefore, the cost of client–server architectures is low compared to server-based archi-
tectures that rely on mainframes. Client–server architectures also tend to be cheaper than
client-based architectures because they place less of a load on networks and thus require less
network capacity.
Cost of Development Th e cost of developing systems is an important factor when consid-
ering the fi nancial benefi ts of client–server architectures. Developing application soft ware for
client–server computing is extremely complex, and most experts believe that it costs four to
Characteristic Server-based Client-based Client–Server
Cost of infrastructure Very high Medium Low
Cost of development Medium Low High
Ease of development Low High Low to medium
Interface capabilities Low High High
Control and security High Low Medium
Scalability Low Medium High
FIGURE 11-6
Characteristics
of Computing
Architectures
Elements of the Physical Architecture Layer 425
fi ve times more to develop and maintain application soft ware for client–server computing
than it does for server-based computing. Developing application soft ware for client-based
architectures is usually cheaper still, because there are many GUI development tools for sim-
ple stand-alone computers that communicate with database servers.
Th e cost diff erential might change as more companies gain experience with client–server
applications, new client–server products are developed and refi ned, and client–server stand-
ards mature. However, given the inherent complexity of client–server soft ware and the need
to coordinate the interactions of soft ware on diff erent computers, there is likely to remain a
cost diff erence.
Ease of Development In most organizations today, there is a huge backlog of main-
frame applications, systems that have been approved but that lack the staff to implement
them. Th is backlog signals the diffi culty in developing server-based systems. Th e tools
for mainframe-based systems oft en are not user friendly and require highly specialized
skills—skills that new graduates oft en don’t have and aren’t interested in acquiring. In
contrast, client-based and client–server architectures can rely on graphical user interface
(GUI) development tools that can be intuitive and easy to use. Th e development of appli-
cations for these architectures can be fast and painless. Unfortunately, the applications for
client–server systems can be very complex because they must be built for several layers of
hardware (e.g., database servers, Web servers, client workstations) that need to communi-
cate eff ectively with one another. Project teams oft en underestimate the eff ort involved in
creating secure, effi cient client–server applications.
Interface Capabilities Typically, server-based applications contain plain, character-based
interfaces. For example, think about airline reservation systems such as SABRE, which can
be quite diffi cult to use unless the operator is well trained on the commands and hundreds
of codes that are used to navigate through the system. Today, most users of systems expect
a GUI or a Web-based interface that they can operate using a mouse and graphical objects.
GUI and Web development tools typically are created to support client-based or client–server
applications; rarely can server-based environments support these types of applications.
Control and Security Th e server-based architecture was originally developed to control
and secure data, and it is much easier to administer because all the data are stored in a single
location. In contrast, client–server computing requires a high degree of coordination among
many components, and the chance for security holes or control problems is much more likely.
Also, the hardware and soft ware used in client–server architecture are still maturing in terms
of security. When an organization has a system that absolutely must be secure, then the pro-
ject team may be more comfortable with the server-based alternative on highly secure and
control-oriented mainframe computers.
Scalability Scalability refers to the ability to increase or decrease the capacity of the
computing infrastructure in response to changing capacity needs. Th e most scalable archi-
tecture is client–server computing because servers can be added to (or removed from) the
architecture when processing needs change. Also, the types of hardware that are used in
client–server situations typically can be upgraded at a pace that most closely matches the
growth of the application. In contrast, server-based architectures rely primarily on main-
frame hardware that needs to be scaled up in large, expensive increments, and client-based
architectures have ceilings above which the application cannot grow because increases
in use and data can result in increased network traffi c to the extent that performance is
unacceptable.
426 Chapter 11 Physical Architecture Layer Design
CLOUD COMPUTING3
Cloud computing is the idea of treating IT as a utility or commodity. Essentially, cloud com-
puting is the latest approach to support distributed computing in a client–server type of
architecture (see previous section) where the server is “in the cloud” and the client is on the
desktop. Th e cloud can be the fi rm’s corporate data center, an external data center, or some
combination of the two; however, more and more it generally is seen as an external, rather
than an internal, service. Consequently, the idea of multitenancy, where the cloud vendor has
multiple customers using the same resource at the same time, becomes a real issue for both
the cloud vendor and the cloud customer. Cloud computing may become the greatest enabler
for IT outsourcing (see Chapter 7).
Th ere are three diff erent classifi cations of clouds: private, public, and hybrid. Private
clouds are available only to employees of the fi rm, public clouds are available to the general
public, and hybrid clouds combine the private and public cloud ideas to form a single cloud.
In some senses, all e-commerce sites could run in a hybrid cloud environment where the cus-
tomer sales transaction portion of the system would need to be public while all other aspects
would be private.
Fundamentally, cloud computing is an umbrella technology that encompasses the ideas
of virtualization, service-oriented architectures, and grid computing. Th e idea of virtualiza-
tion is not new. Virtualization is the idea of treating any computing resource, regardless of
where it is located, as if it is “in” the client machine. Th is idea evolved from virtual memory.
Virtual memory was developed originally in the 1960s. Virtual memory allowed the user/
programmer to act as if the amount of main memory in the computer was unlimited. Th is
was done by swapping “pages” of main memory out to disk when the content of the pages
was not being used and by swapping a page from disk back to main memory when it was
needed. Before virtual memory was created, the programmer had to write code to perform
the paging function for each application. Virtualization is simply the scaling up of this idea
to all computing resources, not simply main memory. Th is includes treating a mainframe
computer as if it is a set of virtual servers, each of which can be running diff erent operating
and/or application systems.
Web services basically support connections between diff erent services to form service-oriented
architectures.4 Basically, a service is a piece of soft ware that supports some aspect of a business
process. A service can be an implementation of part of a business process, it can be an implemen-
tation of an entire business process, or it can be object persistence support for the data manage-
ment layer (see Chapter 9). Th ese services can be either internal or external to the fi rm. Services
can be combined to support business processes. A service-oriented architecture allows business
processes to be supported by “plugging and playing” services together in a static and/or dynamic
manner.5 Some of the pluggable and playable services can be purchased outright, or they can be
billed to the fi rm based on their use, a sort of pay-as-you-go model.
Grid computing6 tends to be the underlying hardware technology that supports the cloud.
A grid is a very large set of networked computers that tend to be geographically dispersed.
For example, the grid that supports SalesForce.com’s CRM application contains about 1,000
3 Judith Hurwitz, Marcia Kaufman, Fern Halper, and Robin Bloor, Cloud Computing for DummiesTM (Hoboken, NJ:
Wiley 2010).
4 Douglas K. Barry, Web Services and Service-Oriented Architectures (San Francisco: Morgan Kaufman, 2003).
5 P. Ghandforoush, T.K. Sen, and D. Tegarden, R. Ramaswamy, “Designing Systems Using Business Components:
A Case Study in Call Center Automation.” International Journal of Electronic Customer Relationship Management
4, no. 2 (2010): 161–179.
6 Pawel Plaszczak and Richard Welner, Jr., Grid Computing (San Francisco: Morgan Kaufman, 2006).
Cloud Computing 427
computers. Th e computers do not have to be of the same type. For example, they can be a
mixture of Linux servers and mainframes. With grid computing, fi rms have the ability to add
and remove computers to support a business process based on the current level of activity
taking place in that particular business process. Th is provides an enormous amount of fl ex-
ibility in confi guring the underlying physical architecture that supports business processes.
Combining virtualization, service-oriented architectures, and grid computing is what
all the hoopla is about with regard to cloud computing. Cloud computing is highly elastic
and scalable, it supports a demand-driven approach to provisioning and deprovisioning of
resources, and it supports a billing model that only charges for the resources being used.
From a business perspective, cloud computing supports the idea of IT being a commodity.
Th e cloud can contain the fi rm’s IT infrastructure, IT platform, and soft ware. Infrastructure
as a Service (IaaS) refers to the cloud providing the computing hardware to the fi rm as a remote
service. Th e hardware typically includes the computing hardware that supports application
servers, networking, and data storage. Amazon’s EC2 (aws.amazon.com/ec2/) service is a good
example of this. With Platform as a Service (PaaS), the cloud vendor not only provides
hardware support to a customer but also provides the customer with either package-based
solutions, diff erent services that can be combined to create a solution, or the development
tools necessary to create custom solutions in the PaaS vendor’s cloud. SalesForce.com is a
good example of the vendor providing a package-based solution, Amazon’s SimpleDB and
Simple Query Service are examples of diff erent services being supported, and Google’s App
Engine is an example of a cloud vendor providing good development tools. Like most things
in IT, Soft ware as a Service (SaaS) is not a new idea. SaaS has been around for more than
thirty years. In the 1970s, there were many “service bureaus” that supported timesharing of
hardware and soft ware to many diff erent customers; that is, they supported multitenancy.
For example, ADP has supported payroll functions for many fi rms for a very long time.
Today, SalesForce.com’s CRM system is a good example of a SaaS cloud-based solution.
However, cloud computing must overcome certain obstacles before it becomes the pri-
mary approach to provision the physical architecture layer.7 Th e fi rst obstacle is the mixed
level of cloud performance. One issue is whether the vendor has the resources to provide the
fi rm with enough “power” during a peak load. Th e issue here is that a typical cloud vendor
is supporting many diff erent fi rms. If the vendor does not have enough computing resources
to handle all of the fi rms’ peak loads at the same time, then there will have to be some deg-
radation of some or all of the fi rms’ support. Th is is primarily a result of the unpredictability
of the overall performance requirements with disk I/O and network traffi c. Given the mul-
titenancy typical of a cloud vendor’s hardware, bottlenecks with disks will occur. However,
given the dependency on networks, data transfer rates are critical. In an enlightening exam-
ple, Armbrust and colleagues show that when dealing with large volumes of data, it is faster
to transfer data using overnight shipping. In their example, they showed that if you were to
transfer 10 terabytes of data with an average transfer rate of 20 Mbits/sec, then it would take
more than 45 days to complete the transfer. If you shipped the data overnight instead, you
would eff ectively be using a transfer rate of 1500 Mbits/sec.
Th e second obstacle deals with the level of dependency that a customer’s fi rm has on a
cloud vendor. Firms are dependent on cloud vendors based on the type of service that they
are using (IaaS, PaaS, and SaaS), the actual level of service availability, and the potential of
data lock-in. Currently, most cloud vendor’s API to storage is proprietary. Consequently, the
customer’s data become “locked in” to the specifi c cloud vendors storage. Th is is also true for
7 Michael Armbrust, Armando Fox, Rean Griffi th, Anthony D. Joseph, Randy Katz, Andy Konwinski, Gunho Lee,
David Patterson, Areil Rabkin, Ion Stoica, and Matei Zahara, “A View of Cloud Computing,” Communications of
the ACM 53, no. 4 (2010): 50–58.
428 Chapter 11 Physical Architecture Layer Design
much of the actual service APIs. Consequently, customers fi nd themselves hoping that the
cloud vendor will be the equivalent of a benevolent dictator that will act in the interest of the
customer; otherwise, actual level of service being provided could suff er. Given the potential
for data and/or service lock-in, a customer must pay close attention to the viability of the
cloud vendor. If the vendor goes out of business, the customer could be following suit very
quickly. If the cloud vendor also has outsourced to other cloud vendors, such as to a disk
farm company, then they could fi nd themselves in the same situation. Th is could lead to a
cascading eff ect of business failures. Consequently, when a fi rm is considering outsourcing its
IT area into the cloud, the fi rm had better understand the total risk involved.
Th e third major obstacle to cloud adoption is the perceived level of security available in
the cloud. Not only does a fi rm have to worry about security from the outside, but due to
multitenancy, the fi rm must seriously consider potential attacks from within its cloud from
other cloud users. From a service availability perspective, a denial-of-service attack against
another tenant within the cloud can cause performance degradation of the fi rm’s systems.
Finally, a fi rm must consider protecting itself from the cloud vendor. Th e cloud vendor is
responsible only for physical security and fi rewalls. All application-level security tends to be
the responsibility of the cloud customer. Obviously, security in the cloud is a very complex
endeavor. Given the confi dentiality and auditability requirements of Sarbanes-Oxley (SOX)
and the Health and Human Services Health Insurance Portability and Accountability Act
(HIPAA), security in the cloud becomes a major concern when a fi rm considers moving
any of its confi dential data, including e-mail, to the cloud. In many ways, when using a
cloud a fi rm is simply taking a leap of faith that the cloud is secure.
UBIQUITOUS COMPUTING AND THE INTERNET OF THINGS 8
Oft en, ubiquitous computing and the Internet of Th ings (IoT) are the beginning of the real-
ization of all of the dreams (or nightmares) of science fi ction writers. Th is ranges from the
dystopian views portrayed in Blade Runner and the Terminator movies to the Precrime unit
of Minority Report and fi nally to the extremely optimistic future portrayed in Th e Jetsons car-
toon of the 1960s. Essentially, ubiquitous computing is the idea that computing takes place
everywhere and in everything. With ubiquitous computing, computing becomes so engrained
into everyday things that computing eff ectively disappears into the background. In other
words, computing becomes so deeply rooted into everyday things that the things themselves
seem to become magical. Th e IoT is the idea that, in addition to things having some form
of computing capacity built into them, everyday things become connected via the Internet.
So, in addition to having some form of computing capacity, everyday things can communi-
cate with each other. Th is raises the importance of understanding mobile computing, social
media, and cloud computing even further. Obviously, the opportunities (or pitfalls) that this
provides may be endless.
Currently, there are two major approaches to support ubiquitous computing: general
computing devices and specialized computing devices. General computing devices include
devices such as smartphones and tablets. Th ese devices can be loaded with many diff er-
ent apps that provide all types of computing and communication support. For example,
your smartphone can be used as a GPS, an e-book reader, a music or video player, a game
8 Th is section is based on material contained in Adam Greenfi eld, Everywhere: Th e Dawning If the Age of Ubiqui-
tous Computing (Berkeley, CA: New Riders, 2006); Bo Begole, Ubiquitous Computing for Business (Upper Saddle
River, NJ: FT Press, 2011); Adrian McEwen and Hakim Cassimally, Designing the Internet of Th ings (Chichester,
West Sussex, UK: Wiley, 2014); David Rose, Enchanted Objects: Design, Human Desire, and the Internet of Th ings
(New York, NY: Scribner, 2014); Gershon Dublon and Joseph A. Paradiso, “Extra Sensory Perception,” Scientifi c
American 311, no. 1 (July 2014): 36–41.
Ubiquitous Computing and the Internet of Things 429
console, a WWW interface, a camera, a “tape” recorder, a restaurant advisor, etc. In other
words, if there is an app for it, you can have it loaded on your smartphone to give you that
capability. Today’s smartphones are essentially general computers that happen to support
voice communications, i.e., it also is a phone. And, like general-purpose computers, the
smartphone typically requires you to activate the app before it can do anything for you.
Even though it is very impressive to have that amount of computing capability at your
fi ngertips, it only supports the dream of ubiquitous computing in a very limited manner.
Essentially, from an information systems development perspective, this is not new; it is no
diff erent than having a computer connected to the Internet. Th us, developing apps for these
devices should follow the same basic development approach used throughout this book.
Th e second approach, having specialized computing devices, goes a long way toward
realizing the dream of ubiquitous computing. With this approach, we have so-called
enchanted objects that can interact with each other. An enchanted object is an everyday
object that has a very specialized processor embedded in it that augments the object such
that the object seems to be magical. For example, an umbrella that, since there is a good
chance of rain, lets you know that you should take it with you today, or a wallet that lets you
know that you are reaching your monthly budget limits or that your account just received
a deposit. In the case of the umbrella, the umbrella is connected to AccuWeather. If the
forecast is for rain, the umbrella activates a set of LEDs in the handle that informs you that
you should take it with you when you leave. In the case of the wallet, as you deplete your
monthly budget, the wallet becomes more diffi cult to open, or if you receive a deposit to
your account, the wallet “puff s” up to let you know that your wallet is fatter, i.e., you have
more cash available.
Th e general information systems development approach used in this book is applica-
ble to the development of enchanted objects. However, given that enchanted objects, by
defi nition, are enhanced everyday things, additional issues must also be addressed. Th ese
issues include a set of unique design principles, a set of characteristics, and a set of levels
of enchantment.
McEwen and Cassimally identify a set of unique design principles that need to be con-
sidered when developing enchanted objects. First, enchanted objects should be in the back-
ground simply providing its message for you to receive at your leisure, not “in your face.”
Th is is in contrast with most apps today. Typically, apps will notify you about some topic at
their leisure by interrupting you. Second, magic is a useful metaphor for people to adopt an
enchanted object. Th e umbrella mentioned earlier is a good example of this principle. Th e
umbrella simply sits by the door letting you know whether it wants to be taken with you or
not. Th ird is the whole issue of privacy. With all of these enchanted objects “sharing” data
about you, all of the issue related to Orwell’s “Big Brother” creeps into focus. How will you
keep anything secret and, possibly even more important, who actually owns the data being
collected? However, this issue is not unique to enchanted objects. It is equally applicable
to smartphones and their apps. Fourth, we need to consider how to “mash-up” a set of
enchanted objects that are loosely connected to support a larger purpose. In fact, Brynjolfsson
and McAfee suggest that this type of recombinant innovation may provide the basis for
a new type of economy that will increase both progress and prosperity.9 Fift h, the idea of
aff ordances becomes increasingly important. For an enchanted object to be adopted, it must
be very simple to use. Th e object itself must imply how to use it. Th e umbrella, for example,
simply lets you know that you should take it with you by drawing your attention to it.
9 Erik Brynjolfsson and Andrew McAfee, Th e Second Machine Age: Work, Progress, and Prosperity in a Time of
Brilliant Technologies (New York, NY: Norton, 2014).
430 Chapter 11 Physical Architecture Layer Design
Rose provides a set of characteristics that enchanted objects should possess if we are
to adopt them. First, they should be glanceable. Th e umbrella, again, is a great example.
You don’t have to do anything but glance at the umbrella to know whether you should
take it with you or not. Second, enchanted objects should be gestureable. Th is is related to
the idea aff ordances. It must be intuitively obvious as to how to use an enchanted object.
For example, years ago Th e A.T. Cross Company sold a notebook and pen combination
(CrossPadTM) that you could use to take notes. Th e aff ordance of this product was the fact
that you simply used a special pen to write your notes on the paper contained in the note-
book. Th e enchanted part was the fact that the product also had a radio transmitter built
in to the pen that enabled it to store your notes in electronic form that could be uploaded
to you computer later. Th ird, the enchanted object must be aff ordable. In this case, given
the falling cost of computing hardware, if the object isn’t that aff ordable at fi rst, it should
be fairly quickly. Fourth, the objects should be wearable. Th e Nike FuelBandTM is a perfect
example. You simply wear it like a watch. Fift h, an enchanted object should be indestructi-
ble. Obviously, this one would only be true as it is related to the underlying object. For
example, the enchanted umbrella is as indestructible as any normal umbrella, but it is not as
indestructible as other things in the real world. Sixth, an enchanted object must be capable
of doing its thing with minimal interaction with the user. For example, you simply wear
the Nike FuelBandTM and plug it up at night to your computer, and it will update itself,
recharge itself, update your profi le, and be waiting for you to put it on in the morning.
Seventh, an enchanted object should be loveable. By that we mean that they should be easy
to anthropomorphize. We should enjoy using them, and we should miss them when we
don’t. Obviously, you should recognize that there are trade-off s among some of the charac-
teristics and, as such, not all objects will possess all of them. However, as a designer, your
enchanted objects should have as many as possible.
Rose also suggests a set of levels (or steps) of enchantment of which enchanted objects
designers should be aware. For the fi rst level, he suggests that enchanted objects should be
augmented everyday objects that are connected to the network. Th is allows them to send
and receive data that can be used by other enchanted objects or other systems. Given the
amount of data to be collected about ourselves and everyone else for both current time and
in the future, the second level for enchanted objects is to be able to be personalized such
that they can interact with us in a customized manner. Currently, to a small degree Amazon
and Netfl ix are already doing this, e.g., with their list of recommendations that they make
to you. Th eir recommendations are based on your past interactions with them and match-
ing those interactions to the interactions of others. Th e potential for this type of activity in
health systems is enormous. Th e third level is where our enchanted objects interact with our
social networks to automatically inform our colleagues, or a special subset of them, of our
activities with the enchanted object. Th is again could be very useful in health systems where
the object informs our physician’s system or our health support group of certain types of
positive (or negative) activities. Th e fourth level adapts gaming ideas to our enchanted
objects, i.e., gamifi cation. Nike’s FuelBandTM is a perfect example of using gamifi cation to
keep a user intrinsically motivated to reach his or her individual goals. Th e last level is that
designers of enchanted objects will improve their adoption if the objects can be part of a
story; Rose refers to this as story-ifi cation. Th rough the use of stories, users can more easily
understand the purpose of and the utility provided by the enchanted objects, thus increas-
ing the likelihood of a user “bonding” with the enchanted object.
Given the potential of ubiquitous computing and the IoT, you should begin consid-
ering possible applications that may benefi t from them. For example, today, through the
use of RFID and GPS, it is possible to know the location of each and every one of a fi rm’s
inventory items. Even though it can be argued that an inventory item with an RFID tag
Green IT 431
that has GPS ability isn’t an enchanted object, it is useful one. And, even though the cost of
these types of augmentations is dropping, it can be further argued that you may not want to
tag each and every item. However, before the enchanted object vision can become a reality,
two possible technical problems will need to be addressed.10 First, given the current set of
communication networks, can the current networks handle the additional communication
volume required? Do the Internet, cell phone, and WiFi networks have suffi cient band-
width to support all of these additional things? When you start to hook up everything to
the Net, it is doubtful that the capacity is there. Second, is it reasonable to expect the simple
special-purpose devices that are embedded in enchanted objects to handle the complexity
of the required communication protocols of the existing networks? To address these prob-
lems could mean that a new physical architecture could be necessary.
GREEN IT11
Given all of the computing power being deployed to solve today’s business problems, Green
IT has become important. Green IT is a broad term that encompasses virtually anything that
helps reduce the environmental impact of IT. Some of the topics included are e-waste, green-
ing data centers, and the dream of the paperless offi ce.
First, when it comes to disposing old electronic devices, care must be taken. Old com-
puters contain very toxic material, including lead, PCBs, mercury, and cadmium. One of
the major Green IT issues is how to dispose of this e-waste. One of the most disturbing
trends in dealing with e-waste is the shipping of the e-waste from the developed world to
the developing world where environmental standards are virtually nonexistent. Owing to
“backyard recycling” techniques used in these locations, the toxic material contained in the
e-waste shows up in the soil, water, and air. Alternatives to simply dumping old comput-
ers into the trash include extending the replacement cycles of the machines by converting
the machines from Windows-based machines to Linux-based machines. Linux takes less
“horsepower” to run than Windows. Th erefore, for certain applications, a Linux-based
desktop is more than suffi cient to implement parts of the physical architecture layer.
Second, large data centers use as much electricity in a day as a small city. Consequently,
given this level of power consumption, creating green data centers in the future will be cru-
cial. Th ere are a whole set of ways to create a green data center. One way is to pay very close
attention to where the data center is to be located. Placing the data center in the shade of a
mountain or tall building will reduce the cost of energy required. For example, HP placed
one of its new data centers in northeast England so that it could be cooled by the cold winds
that blow onto shore from the North Sea.12 Looking into alternative energy possibilities is
another way to deal with energy consumption. For example, Google has been in the business
of buying wind farms to generate the power for its data centers, and HP has shown how a
cow manure-based methane power plant could be created to generate the power to run a data
center in dairy country.13
10 Francis da Costa, Rethinking the Internet of Th ings: A Scalable Approach to Connecting Everything (New York,
NY: Apress Media, 2013).
11 Caril Baroudi, Jeff rey Hill, Arnold Reinhold, and Jhana Senxian, Green IT for DummiesTM (Hoboken, NJ: Wiley,
2009).
12 Andrew Nusca, “Smart Takes: HP Opens First Wind-Cooled Green Data Center; Most Effi cient to Date,” SMART-
PLANET (February 11, 2010). Retrieved August 2014 from www.smartplanet.com/blog/smart-takes/hp-opens-fi rst-
wind-cooled-green-data-center-most-effi cient-to-date.
13 Google Data Centers, Renewable Energy. Retrieved August 2014 from www.google.com/about/datacenters/renewable.
432 Chapter 11 Physical Architecture Layer Design
Th e third way to consider making your IT infrastructure greener is to consider the cloud
(see earlier section). With the cloud’s virtualization capabilities, the number of high-powered
servers and desktops can be reduced. However, you will need to perform some trade-off s
between the obstacles of moving to the cloud and the move toward a greener IT. Th e fourth
way to address the power demands for a modern IT infrastructure is by only purchasing Energy
Star compliant electronics. Th e fi fth way is to encourage employees to have their machines go
to “sleep” to save energy when the machines have been idle for some period of time.
Th e paperless offi ce idea has been around for a very long time. However, up until now,
the idea has been more fantasy than reality. Today, with the advent of multiuse tablets, such
as Apple’s iPadTM, the paperless offi ce is becoming a reality. When considering the cloud and
the apps available on the iPadTM, it is possible not only to create a paperless offi ce but also to
have the paperless offi ce eff ectively be a portable offi ce.
INFRASTRUCTURE DESIGN
In most cases, a system is built for an organization that has a hardware, soft ware, and commu-
nications infrastructure already in place. Th us, project teams are usually more concerned with
how an existing infrastructure needs to be changed or improved to support the requirements
that were identifi ed during analysis, as opposed to how to design and build an infrastructure
from scratch. Coordination of infrastructure components is very complex, and it requires
highly skilled technical professionals. As a project team, it is best to allow the infrastructure
analysts to make changes to the computing infrastructure.
Deployment Diagram
Deployment diagrams are used to represent the relationships between the hardware components
used in the physical infrastructure of an information system. For example, when designing a
distributed information system that will use a wide area network, a deployment diagram can
be used to show the communication relationships among the diff erent nodes in the network.
Th ey also can be used to represent the soft ware components and how they are deployed over
the physical architecture or infrastructure of an information system. In this case, a deployment
diagram represents the environment for the execution of the soft ware.
Th e elements of a deployment diagram include nodes, artifacts, and communication
paths (see Figure 11-7). Other elements can also be included in this diagram. In our case,
we include only the three primary elements and the element that portrays an artifact being
deployed onto a node.
A node represents any piece of hardware that needs to be included in the model of the
physical architecture layer design. For example, nodes typically include client computers,
servers, separate networks, or individual network devices. Typically, a node is labeled with its
name and, possibly, with a stereotype. Th e stereotype is modeled as a text item surrounded
by “<>” symbols. Th e stereotype represents the type of node being represented on the
diagram. For example, typical stereotypes include device, mobile device, database server, Web
server, and application server. Th ere are times that the notation of a node should be extended
to better communicate the design of the physical architecture layer. Figure 11-8 includes a set
of typical network node symbols that can be used instead of the standard notation.
An artifact represents a piece of the information system that is to be deployed onto the
physical architecture (see Figure 11-7). Typically, an artifact represents a soft ware component, a
subsystem, a database table, an entire database, or a layer (data management, human–computer
interaction, or problem domain). Artifacts, like nodes, can be labeled with both a name and a
stereotype. Stereotypes for artifacts include source fi le, database table, and executable fi le.
Infrastructure Design 433
FIGURE 11-7 Development Diagram Syntax
A node:
■ Is a computational resource, e.g., a client computer, server, separate network, or
individual network device.
■ Is labeled by its name.
■ May contain a stereotype to specifically label the type of node being represented,
e.g., device, client workstation, application server, mobile device, etc.
An artifact:
■ Is a specification of a piece of software or database, e.g., a database or a table or
view of a database, a software component or layer.
■ Is labeled by its name.
■ May contain a stereotype to specifically label the type of artifact, e.g., source file,
database table, executable file, etc.
A communication path:
■ Represents an association between two nodes.
■ Allows nodes to exchange messages.
■ May contain a stereotype to specifically label the type of communication path
being represented, (e.g., LAN, Internet, serial, parallel).
A node with a deployed artifact:
■ Portrays an artifact being placed on a physical node.
<>
Node Name
<>
<>
Artifact Name
<>
Node Name
<>
Artifact Name
A communication path represents a communication link between the nodes of the phys-
ical architecture (see Figure 11-7). Communication paths are stereotyped based on the type
of communication link they represent (e.g., LAN, Internet, serial, parallel, or USB) or the
protocol that is being supported by the link (e.g., TCP/IP).
FIGURE 11-8 Extended Node Syntax for Development Diagram
Workstation Server Mainframe
Subnetwork Data-
base
Firewall
434 Chapter 11 Physical Architecture Layer Design
Figure 11-9 portrays three diff erent versions of a deployment diagram. Version a uses
only the basic standard notation. Version b introduces the idea of deploying an artifact
onto a node (see Figure 11-7). In this case, the artifacts represent the diff erent layers of the
appointment system described in earlier chapters. Version c uses the extended notation to
represent the same architecture. As you can see, all three versions have their strengths and
weaknesses. When comparing version a and version b, the user can glean more information
from version b with little additional eff ort. However, when comparing version a to version
c, the extended node notation enables the user to quickly understand the hardware require-
ments of the architecture. When comparing version b to version c, version b supports the
soft ware distribution explicitly but forces the user to rely on the stereotypes to understand the
required hardware, whereas version c omits the soft ware distribution information entirely.
We recommend that you use the combination of symbols to best portray the physical archi-
tecture to the user community.
Network Model
Th e network model is a diagram that shows the major components of the information system
(e.g., servers, communication lines, networks) and their geographic locations throughout the
organization. Th ere is no one way to depict a network model, and in our experience analysts
FIGURE 11-9 Three Versions of Appointment System Deployment Diagram
<>
<>
Receptionist PC
<>
Office Server
<>
Receptionist PC
<>
Appt System
<>
Office Server
<>
Appt System
<>
Appt System
Receptionist PC Office Server
Appt
Data
Base
<>
<>
(a)
(b)
(c)
Infrastructure Design 435
create their own standards and symbols, using presentation applications (e.g., PowerPoint) or
diagramming tools (e.g., Visio). In this text, we use UML’s deployment diagram.
Th e purpose of the network model is twofold: to convey the complexity of the system
and to show how the system’s soft ware components will fi t together. Th e diagram also helps
the project team develop the hardware and soft ware specifi cation that is described later in
this chapter.
Th e components of the network model are the various clients (e.g., personal computers,
kiosks), servers (e.g., database, network, communications, printer), network equipment (e.g.,
WiFi connections, ethernet, cell phone network, satellite links), and external systems or
networks (e.g., Internet service providers) that support the application. Locations are the geo-
graphic sites related to these components. For example, if a company created an application
for users at four of its plants in Canada and eight plants in the United States and it used one
external system to provide Internet service, the network model to depict this would contain
twelve locations (4 1 8 5 12).
Creating the network model is a top-down exercise whereby we fi rst graphically depict
all the locations where the application will reside. Placing symbols that represent the locations
for the components on a diagram and then connecting them with lines that are labeled with
the approximate amount of data or types of network circuits between the separated compo-
nents accomplish this.
Companies seldom build networks to connect distant locations by buying land and lay-
ing cable (or sending up their own satellites). Instead, they usually lease services provided by
large telecommunications fi rms such as AT&T, Sprint, and Verizon. Figure 11-10 shows a
typical network. Th e clouds in the diagram represent the networks at diff erent locations (e.g.,
Toronto, Atlanta). Th e lines represent network connections between specifi c points (e.g.,
Toronto to Brampton). In other cases, a company might lease connections from many points
to many others, and rather than trying to show all the connections, a separate cloud may be
drawn to represent this many-to-many type of connection (e.g., the cloud in the center of
Figure 11-10 represents a network of many-to-many connections provided by a telecom fi rm
like Verizon).
Th is high-level diagram has several purposes. First, it shows the locations of the com-
ponents needed to support the application; therefore, the project team can get a good
understanding of the geographic scope of the new system and how complex and costly the
communications infrastructure will be to support. (For example, an application that supports
one site will probably have less communications costs as compared to a more- complex appli-
cation that will be shared all over the world.) Th e diagram also indicates the external compo-
nents of the system (e.g., customer systems, supplier systems), which may impact security or
global needs (discussed later in this chapter).
Th e second step of the network model is to create low-level network diagrams for each
of the locations shown on the top-level diagram. First, hardware is drawn on the model
in a way that depicts how the hardware for the new system will be placed throughout the
location. It usually helps to use symbols that resemble the hardware that will be used. Th e
amount of detail to include on the network model depends on the needs of the project.
Some low-level network models contain text descriptions below each of the hardware
components that describe in detail the proposed hardware confi gurations and processing
needs; others include only the number of users that are associated with the diff erent parts
of the diagram.
Next, lines are drawn connecting the components that will be physically attached to
each other. In terms of soft ware, some network models list the required soft ware for each
network model component right on the diagram, whereas other times, the soft ware is
436 Chapter 11 Physical Architecture Layer Design
FIGURE 11-10 Deployment Diagram Representation of a Top-Level Network Model
<>
<>
<
>
<>
<>
<
>
<><>
<
>
<
><>
<>
<>
Verizon Network
Ottawa,
ONT
Tyson’s Corner,
VA
Research
Triangle Park,
NC
Ossining,
NY
Sunrise,
FL
Atlanta,
GA
Nashville,
TN
Richardson,
TX
Santa Clara,
CA
Toronto,
ONT
Brampton,
ONT
Mariline,
ONT
described in a memo attached to the network model. Figure 11-11 shows a deployment
diagram that portrays two levels of detail of a low-level network model. Notice, we use both
the standard and extended node notation in this fi gure. In this case, we have included a
package (see Chapter 7) to represent a set of connections to the router in the MFA building.
By including a package, we show only the detail necessary. Th e extended notation in many
cases aids the user in understanding the topology of the physical architecture layer much
better than the standard notation. We recommend using the symbols that get the message
across best.
Infrastructure Design 437
FIGURE 11-11 Deployment Diagram Representation of a Low-Level Network Model
<>
Main Financial Aid Building
<>
<>
DOE
Washington, DC
Offices
DOE Regional Offices
(a)
(b)
Main Financial Aid Building
<>
<>
ServerServer
Ethernet
LAN
Ethernet
LAN
<><>
<>
<>
ServerServer
Ethernet
LAN
Ethernet
LAN
<>
Ethernet
LAN
<>
<>
<>
<>
Ethernet
LAN
<>
<>
<>
MFA Building
<>
<>
<>
DOE
Washington, DC
Offices
DOE Regional Offices
438 Chapter 11 Physical Architecture Layer Design
Our experiences have shown that most project teams create a memo for the project fi les
that provides additional detail about the network model. Th is information is helpful to the
people who are responsible for creating the hardware and soft ware specifi cations (described
later in this chapter) and who will work more extensively with the infrastructure develop-
ment. Th is memo can include special issues that aff ect communications, requirements for
hardware and soft ware that might not be obvious from the network model, or specifi c hard-
ware or soft ware vendors or products that should be acquired.
Th e primary purpose of the network model diagram is to present the proposed infra-
structure for the new system. Th e project team can use the diagrams to understand the scope
of the system, the complexity of its structure, any important communication issues that might
aff ect development and implementation, and the actual components that need to be acquired
or integrated into the environment.
HARDWARE AND SYSTEM SOFTWARE SPECIFICATIONS
Th e time to begin acquiring the hardware and soft ware that will be needed for a future system
is during the design of the system. In many cases, the new system will simply run on the exist-
ing equipment in the organization. Other times, however, new hardware and soft ware must be
purchased. Th e hardware and soft ware specifi cation is a document that describes what hardware
and soft ware are needed to support an application. Th e actual acquisition of hardware and soft –
ware should be left to the purchasing department or the area in the organization that handles
capital procurement. However, the project team writes the hardware and soft ware specifi cation
to communicate the project needs to the appropriate people. Th ere are several steps involved
in creating the document. Figure 11-12 shows a sample hardware and soft ware specifi cation.
First, we need to defi ne the soft ware that will run on each component. Th is usually
starts with the operating system (e.g., Windows, Linux) and includes any special-purpose
soft ware on the client and servers (e.g., Oracle database). Th is document should consider any
additional costs, such as technical training, maintenance, extended warranties, and licensing
agreements (e.g., a site license for a soft ware package). Th e listed needs are infl uenced by
decisions that are made in the other design activities.
Second, we must create a list of the hardware that is needed to support the future system.
With the advent of mobile computing (see Chapter 10), cloud computing (see earlier in this
chapter), the IoT (see earlier in this chapter), and Green IT (see earlier in this chapter), this
Specifi cation
Standard
Client
Standard
Web Server
Standard
Application Server
Standard
Database Server
Operating System • Windows
• Internet Explorer
• Linux • Linux • Linux
Special Software • Acrobat Reader
• Adobe Flash
• QuickTime
• Apache • Java • Oracle
Hardware • 8 GB Memory
• 500 GB disk drive
• Intel Core i5
• 2–22” monitors
• 16 GB Memory
• 1TB disk drive
• Intel Xenon
E%-2400
• 1–22” monitor
• 32 GB Memory
• 2–1 TB disk drives
• Intel Xenon E5-2600
• 1–22” monitor
• 32 GB Memory
• 4–1 TB Hotplug
disk drives
• Intel Xenon E5-2600
• 1–22” monitor
Network • 100 Mbps Ethernet
High-speed Wireless
• 100 Mbps Ethernet • 100 Mbps Ethernet • 100 Mbps Ethernet
FIGURE 11-12 Sample Hardware and Software Specifi cation
Hardware and System Software Specifi cations 439
step is much more involved than it used to be. However, the low-level network model pro-
vides a good starting point for recording the project’s hardware needs because each compo-
nent on the diagram corresponds to an item on this list. In general, the list can include things
like database servers, network servers, peripheral devices (e.g., printers, scanners), backup
devices, storage components, and any other hardware component that is needed to support an
application. At this time, you also should note the quantity of each item that will be needed.
Th ird, we must describe, in as much detail as possible, the minimum requirements for
each piece of hardware. Typically, the project team must convey requirements like the amount
of processing capacity, the amount of storage space, and any special features that should be
included. Many organizations have standard lists of approved hardware and soft ware that must
be used; so in many cases, this step simply involves selecting items from the lists. Other times,
however, the team is operating in new territory and is not constrained by the need to select from
an approved list. Th is step becomes easier with experience; however, there are some hints that
can help you describe hardware needs (see Figure 11-13). For example, consider the hardware
standards within the organization or those recommended by vendors. Talk with experienced
system developers or other companies with similar systems. Finally, think about the factors that
aff ect hardware performance, such as the response-time expectations of the users, data volumes,
soft ware memory requirements, the number of users accessing the system, the number of exter-
nal connections, and growth projections.
Th e last step to consider is to evaluate vendor proposals (see Chapter 7). Th e easiest way
to do this is to create an alternative matrix (see Chapters 2 and 7). In this case, the evaluation
criteria in the alternative matrix should include all architectural requirements, both optional
and mandatory, and each criterion should be weighted. Some general criteria include CPU
speed, bus speed, disk size, disk access time, cache size, cache speed, RAM size, RAM speed,
data transfer rate, video RAM size and speed, monitor size, and printer resolution. Of course,
in today’s connected world, the networking hardware and soft ware would also need to be
specifi ed, including routers, print servers, hubs, and switches. Mobile devices such as smart-
phones and tablets may be part of the physical architecture solution. Depending on the prob-
lem domain requirements, additional hardware and system soft ware could be required, such
as speech recognition and generation soft ware and hardware, digitizing tablets, and possibly
head-mounted displays, shutter glasses, force feedback pointing devices, and 3D printers.
Each of these types of specialized devices has its own specialized evaluation criteria. In a nut-
shell, when creating a hardware and system soft ware specifi cation, most systems analysts fi nd
that they need help from IT and CS personnel.
FIGURE 11-13
Factors in Hardware
and Software
Selection
Functions and Features What specifi c functions and features are needed (e.g., size of monitor,
software features)?
Performance How fast does the hardware and software operate (e.g., processor, number of
database writes per second)?
Legacy Databases and Systems How well does the hardware and software interact with legacy
systems (e.g., can it write to this database)?
Hardware and OS Strategy What are the future migration plans (e.g., the goal is to have all of one
vendor’s equipment)?
Cost of Ownership What are the costs beyond purchase (e.g., incremental license costs, annual
maintenance, training costs, salary costs)?
Political Preferences People are creatures of habit and are resistant to change, so changes should
be minimized.
Vendor Performance Some vendors have reputations or future prospects that are different from
those of a specifi c hardware or software system they currently sell.
440 Chapter 11 Physical Architecture Layer Design
Depending on the overall cost and size of the project, one thing that should be seriously
considered is the use of a benchmark. A benchmark is essentially a sample of programs that
would be expected to run on the new physical architecture. Even though benchmarks can be
expensive to create, they tend to provide a more realistic picture of how the proposed physical
architecture layer will perform.
When evaluating hardware, there is a set of issues that you should recognize.14
■ Not only should you provide sample programs for the benchmarks, but you also
need to provide actual data. Otherwise, the benchmark results could be misleading.
■ You need to carefully review the mix of system soft ware and hardware. For
example, in many cases, Linux performs better on the same hardware when
compared against Windows, but some applications might not be available under
Linux. Consequently, there may be some trade-off s that should be considered.
■ When considering adding additional hardware, be sure to evaluate the additional
hardware based on marginal utility, not actual utility.
■ Do not specify the physical architecture before you understand the problem
domain requirements. Th is might seem obvious, but when you consider the time
it takes for a mainframe computer, a large number of servers, or a large number
of client machines to be specifi ed, ordered, and delivered, it can be tempting to
specify the hardware and system soft ware prematurely. Th is could lead to either
under- or over-specifi cation.
■ Recognize the reality of Parkinson’s Law. From an IT perspective, Parkinson’s Law
implies that regardless of the users’ real needs, their imagined needs will always
fi ll up whatever capacity the system has. Consequently, it is imperative that the
physical architecture layer design be based on the current and expected future
architecture of the problem domain layer.
■ Do not limit choices to a single vendor. Th is is especially true when you consider
commodity hardware, such as displays, desktops, and department-size servers.
■ Given the rate of technological change that is taking place in IT, consider lead-
ing-edge ideas. For example, even though tablet computers have been around for
a while, the iPadTM was not on most people’s radar. Today, it is considered to be a
game changer when considering client-based hardware. Consequently, you really
must stay up to date when it comes to the design of the physical architecture layer.
NONFUNCTIONAL REQUIREMENTS AND PHYSICAL
ARCHITECTURE LAYER DESIGN
Th e design of the physical architecture layer specifi es the overall architecture and the place-
ment of soft ware and hardware that will be used. Each of the architectures discussed before
has its strengths and weaknesses. Most organizations use client–server architectures for cost
reasons, so in the event that there is no compelling reason to choose one architecture over
another, cost usually suggests client–server.
Creating a physical architecture layer design begins with the nonfunctional requirements.
Th e fi rst step is to refi ne the nonfunctional requirements into more-detailed requirements
that are then used to help select the architecture to be used (server-based, client-based, or
client–server) and what soft ware components will be placed on each device. In a client–server
14 Alton R. Kindred, Data Systems and Management: An Introduction to Systems Analysis and Design, 2nd Ed.
(Englewood Cliff s, NJ: Prentice-Hall, 1980).
Nonfunctional Requirements and Physical Architecture Layer Design 441
architecture, one also has to decide whether to use a two-tier, three-tier, or n-tier architecture.
Th en the nonfunctional requirements and the architecture design are used to develop the
hardware and soft ware specifi cation.
Four primary types of nonfunctional requirements can be important in designing the
architecture: operational requirements, performance requirements, security requirements,
and cultural/political requirements. Furthermore, each of these requirements must be fully
verifi ed and validated.
Operational Requirements
Operational requirements specify the operating environment(s) in which the system must
perform and how those might change over time. Th is usually refers to operating systems,
system soft ware, and information systems with which the system must interact, but on occa-
sion it also includes the physical environment if the environment is important to the applica-
tion (e.g., it’s located on a noisy factory fl oor, so no audible alerts can be heard). Figure 11-14
summarizes four key operational requirement areas and provides some examples of each.
Technical Environment Requirements Technical environment requirements specify the type
of hardware and soft ware system on which the system will work. Th ese requirements usually
focus on the operating system soft ware (e.g., Windows, Linux, Mac OS), database system
soft ware (e.g., Oracle), and other system soft ware (e.g., Firefox). In today’s distributed world,
issues related to mobile computing (see Chapter 10), cloud computing (see the earlier section
in this chapter), the IoT (see earlier section in this chapter), and Green IT (see the earlier sec-
tion in this chapter) are very relevant. Consequently, it also includes all of the diff erent types of
Type of Requirement Defi nition Examples
Technical Environment
Requirements
Special hardware, software, and network
requirements imposed by business
requirements
• The system will work over the Web
environment with Internet Explorer.
• All offi ce locations will have an always-on
network connection to enable real-time data-
base updates.
• A version of the system will be provided for
customers connecting over the Internet via a
tablet or smartphone.
System Integration
Requirements
The extent to which the system will
operate with other systems
• The system must be able to import and
exportExcel spreadsheets.
• The system will read and write to the main
inventory database in the inventory system.
Portability Requirements The extent to which the system will need
to operate in other environments
• The system must be able to work with differ-
entoperating systems (e.g., Linux, Mac OS,
and Windows).
• The system might need to operate with
handheld devices, such as Android and
Apple iOS devices.
Maintainability
Requirements
Expected business changes to which the
system should be able to adapt
• The system will be able to support more than
one manufacturing plant with six months’
advance notice.
• New versions of the system will be released
every six months.
FIGURE 11-14 Operational Requirements
442 Chapter 11 Physical Architecture Layer Design
hardware from mainframe computers to smartphones. Depending on the applications being
deployed over the physical architecture, specialized hardware could be required, such as 3D
displays, 3D printing, 3D sound systems, and tablets with accelerometers. With today’s tech-
nology, the possible combinations of hardware that can be used to solve a problem are nearly
endless. Consequently, this is one area where additional expertise might be required.
System Integration Requirements System integration requirements are those that require
the system to operate with other information systems, either inside or outside the company.
Th ese typically specify interfaces through which data will be exchanged with other systems.
Portability Requirements Information systems never remain constant. Business needs
change and operating technologies change, so the information systems that support them
and run on them must change, too. Portability requirements defi ne how the technical oper-
ating environments might change over time and how the system must respond (e.g., the sys-
tem currently runs on Windows, whereas in the future the system might have to be deployed
on Linux). Portability requirements also refer to potential changes in business requirements
that drive technical environment changes. For example, in the future users might want to
access a website from their cell phones.
Maintainability Requirements Maintainability requirements specify the business require-
ment changes that can be anticipated. Not all changes are predictable, but some are. For
example, suppose a small company has only one manufacturing plant but is anticipating the
construction of a second plant in the next fi ve years. All information systems must be written
to make it easy to track each plant separately, whether for personnel, budgeting, or inventory
systems. Th e maintainability requirements attempt to anticipate future requirements so that
the systems designed today will be easy to maintain if and when those future requirements
appear. Maintainability requirements can also defi ne the update cycle for the system, such as
the frequency with which new versions will be released.
Performance Requirements
Performance requirements focus on performance issues, such as response time, capacity, and
reliability. Figure 11-15 summarizes three key performance requirement areas and provides
some examples.
Speed Requirements Speed requirements are exactly what they say: How fast should the sys-
tem operate? First is the response time of the system: How long it takes the system to respond to
a user request. Although everyone would prefer low response times, with the system respond-
ing immediately to each user request, this is not practical. We could design such a system, but it
would be expensive. Most users understand that certain parts of a system will respond quickly,
whereas others are slower. Actions that are performed locally on the user’s computer must be
almost immediate (e.g., typing, dragging, and dropping), whereas others that require commu-
nicating across a network can have longer response times (e.g., a Web request).
Th e second aspect of speed requirements is how long it takes transactions in one part
of the system to be refl ected in other parts. For example, how soon aft er an order is placed
will the items it contained be shown as no longer available for sale to someone else? If the
inventory is not updated immediately, then someone else could place an order for the same
item, only to fi nd out later it is out of stock. Th is is especially true when one considers NoSQL
database that does not update all copies of the data immediately (see Chapter 9). Or how soon
aft er an order is placed is it sent to the warehouse to be picked from inventory and shipped?
Nonfunctional Requirements and Physical Architecture Layer Design 443
Type of Requirement Defi nition Examples
Speed Requirements The time within which the system must
perform its functions
• Response time must be less than 7 seconds
for any transaction over the network.
• The inventory database must be updated in
real time.
• Orders will be transmitted to the factory
fl oor every 30 minutes.
Capacity Requirements The total and peak number of users
and the volume of data expected
• There will be a maximum of 100–200
simultaneous users at peak use times.
• A typical transaction will require the trans-
mission of 10K of data.
Availability and Reliability
Requirements
The extent to which the system will
be available to the users and the
permissible failure rate due to errors
• The system will store data on approximately
5,000 customers for a total of about 2 MB
of data.
• Scheduled maintenance shall not exceed
one 6-hour period each month.
• The system shall have 99% uptime perfor-
mance.
FIGURE 11-15 Performance Requirements
Capacity Requirements Capacity requirements attempt to predict how many users the sys-
tem will have to support, both in total and simultaneously. Capacity requirements are impor-
tant in understanding the size of the databases, the processing power needed, and so on. Th e
most important requirement is usually the peak number of simultaneous users because this
has a direct impact on the processing power of the computer(s) needed to support the system.
It is oft en easier to predict the number of users for internal systems designed to support
an organization’s own employees than it is to predict the number of users for customer-facing
systems, especially those on the Web. How does Weather.com estimate the peak number of
users who will simultaneously seek weather information? Th is is as much an art as a science,
so oft en the team provides a range of estimates, with wider ranges used to signal a less-accu-
rate estimate.
Availability and Reliability Requirements Availability and reliability requirements focus
on the extent to which users can assume that the system will be available for them to use.
Although some systems are intended to be used only during the forty-hour workweek, some
systems are designed to be used by people around the world. For such systems, project team
members need to consider how the application can be operated, supported, and maintained
24/7 (i.e., 24 hours a day, 7 days a week). Th is 24/7 requirement means that users might need
help or have questions at any time, and a support desk that is available eight hours a day will
not be suffi cient support. It is also important to consider what reliability is needed in the sys-
tem. A system that requires high reliability (e.g., a medical device or telephone switch) needs
far greater planning and testing than one that does not have such high-reliability needs (e.g.,
personnel system, Web catalog).
It is more diffi cult to predict the peaks and valleys in use of the system when the system
has a global audience. Typically, applications are backed up on weekends or late evenings
when users are no longer accessing the system. Such maintenance activities need to be
rethought with global initiatives. For example, what day(s) of the week is considered a “down”
day. In diff erent parts of the world, business does not take place every day. In some parts,
444 Chapter 11 Physical Architecture Layer Design
Friday is sacred; in other parts, it’s Saturday or Sunday. Consequently, political and cultural
issues (described below and in Chapter 10) can impact the performance requirements. Th e
development of Web interfaces, in particular, has escalated the need for 24/7 support; by
default, the Web can be accessed by anyone at any time. For example, the developers of a
Web application for U.S. outdoor gear and clothing retailer Orvis were surprised when the
fi rst order aft er going live came from Japan.
Security Requirements15
Security is the ability to protect the information system from disruption and data loss, whether
caused by an intentional act (e.g., a hacker, a terrorist attack) or a random event (e.g., disk
failure, tornado). Security is primarily the responsibility of the operations group—the staff
responsible for installing and operating security controls, such as fi rewalls, intrusion-detection
systems, and routine backup and recovery operations. Nonetheless, developers of new systems
must ensure that the system’s security requirements produce reasonable precautions to prevent
problems; system developers are responsible for ensuring security within the information
systems themselves.
Security is an ever-increasing problem in today’s Internet-enabled world. Historically,
the greatest security threat has come from inside the organization itself. Ever since the early
1980s when the FBI fi rst began keeping computer crime statistics and security fi rms began
conducting surveys of computer crime, organizational employees have perpetrated the vast
majority of computer crimes. For years, 80 percent of unauthorized break-ins, theft s, and
sabotage have been committed by insiders, leaving only 20 percent to hackers external to the
organizations.
In 2001, that changed. Depending on what survey you read, the percentage of incidents
attributed to external hackers in 2001 increased to 50 to 70 percent of all incidents, meaning
that the greatest risk facing organizations is now from the outside. Although some of this shift
may be due to better internal security and better communications with employees to prevent
security problems, much of it is simply due to an increase in activity by external hackers. With
cloud computing and the IoT, security has become even more important.
Developing security requirements usually starts with some assessment of the value of the
system and its data. Th is helps pinpoint extremely important systems so that the operations
staff is aware of the risks. Security within systems usually focuses on specifying who can access
what data, identifying the need for encryption and authentication, and ensuring the applica-
tion prevents the spread of viruses (see Figure 11-16).
System Value Th e most important computer asset in any organization is not the equipment;
it is the organization’s data. For example, suppose someone destroyed a mainframe com-
puter worth $10 million. Th e mainframe could be replaced, simply by buying a new one. It
would be expensive, but the problem would be solved in a few weeks. Now suppose someone
destroyed all the student records at your university so that no one knew what courses anyone
had taken or their grades. Th e cost would far exceed the cost of replacing a $10 million com-
puter. Th e lawsuits alone would easily exceed $10 million, and the cost of staff to fi nd paper
records and reenter the data from them would be enormous and certainly would take more
than a few weeks.
15 For more information, see Brett C. Tjaden, Fundamentals of Secure Computer Systems (Wilsonville, OR: Franklin,
Beedle, and Associates, 2004); for security controls associated with the Sarbanes–Oxley act, see Dennis C. Brewer,
Security Controls for Sarbanes–Oxley Section 404 IT Compliance: Authorization, Authentication, and Access
(Indianapolis: Wiley, 2006).
Nonfunctional Requirements and Physical Architecture Layer Design 445
FIGURE 11-16 Security Requirements
Type of Requirement Defi nition Examples
System Value Estimates Estimated business value of the system and
its data
• The system is not mission critical, but a sys-
tem outage is estimated to cost $50,000 per
hour in lost revenue.
• A complete loss of all system data is esti-
mated to cost $20 million.
Access Control
Requirements
Limitations on who can access what data • Only department managers will be able to
change inventory items within their own
department.
• Telephone operators will be able to read
and create items in the customer fi le but
cannot change or delete items.
Encryption and the
Authentication
Requirements
Defi nes what data will be encrypted Where
and whether authentication will be
needed for user access
• Data will be encrypted from the user’s com-
puter to website to provide secure ordering.
• Users logging in from outside the offi ce will
be required to authenticate.
Virus Control
Requirements
Requirements to control the spread
of viruses
• All uploaded fi les will be checked for
viruses before being saved in the system.
In some cases, the information system itself has value that far exceeds the cost of the
equipment as well. For example, for an Internet bank that has no brick and mortar branches,
the website is a mission-critical system. If the website crashes, the bank cannot conduct busi-
ness with its customers. A mission-critical application is an information system that is literally
critical to the survival of the organization. It is an application that cannot be permitted to fail,
and if it does fail, the network staff drops everything else to fi x it. Mission-critical applications
are usually clearly identifi ed so that their importance is not overlooked.
Even temporary disruptions in service can have signifi cant costs. Th e costs of disruptions
to a company’s primary website or the LANs and backbones that support telephone sales oper-
ations are oft en measured in the millions of dollars. Amazon.com, for example, has revenues
of more than $10 million per hour, so if its website were unavailable for an hour or even part
of an hour, it would lose millions of dollars in revenue. Companies that do less e-business or do
telephone sales have lower costs, but recent surveys suggest losses of $100,000 to $200,000 per
hour are not uncommon for major customer-facing information systems.
Access Control Requirements Some of the data stored in the system need to be kept confi den-
tial; some data need special controls on who is allowed to change or delete it. Personnel records,
for example, should be able to be read only by the personnel department and the employee’s
supervisor; changes should be permitted to be made only by the personnel department. Access
control requirements state who can access what data and what type of access is permitted: whether
the individual can create, read, update and/or delete the data. Th e requirements reduce the
chance that an authorized user of the system can perform unauthorized actions. One approach
to address these requirements is through the use of access control lists, which can be implemented
via the operating system or database management system.
Encryption and Authentication Requirements One of the best ways to prevent unauthorized
access to data is encryption, which is a means of disguising information by the use of mathemat-
ical algorithms (or formulas). Encryption can be used to protect data stored in databases or data
446 Chapter 11 Physical Architecture Layer Design
that are in transit over a network from a database to a computer. Th ere are two fundamentally
diff erent types of encryption: symmetric and asymmetric. A symmetric encryption algorithm
[such as Data Encryption Standard (DES) or Advanced Encryption Standard (AES)] is one in
which the key used to encrypt a message is the same as the one used to decrypt it, which means
that it is essential to protect the key and that a separate key must be used for each person or
organization with whom the system shares information (or else everyone can read all the data).
An asymmetric encryption algorithm (such as public key encryption) is one in which the key
used to encrypt data (called the public key) is diff erent from the one used to decrypt it (called the
private key). Even if everyone knows the public key, once the data are encrypted, they cannot be
decrypted without the private key. Public key encryption greatly reduces the key-management
problem. Each user has its public key that is used to encrypt messages sent to it. Th ese public
keys are widely publicized (e.g., listed in a telephone book style directory)—that’s why they’re
called public keys. Th e private key, in contrast, is kept secret (which is why it’s called private).
Public key encryption also permits authentication (or digital signatures). When one user
sends a message to another, it is diffi cult to legally prove who actually sent the message. Legal proof
is important in many communications, such as bank transfers and buy/sell orders in currency
and stock trading, which normally require legal signatures. Public key encryption algorithms are
invertible, meaning that text encrypted with either key can be decrypted by the other. Normally,
we encrypt with the public key and decrypt with the private key. However, it is possible to do the
reverse: encrypt with the private key and decrypt with the public key. Because the private key is
secret, only the real user can use it to encrypt a message. Th us, a digital signature or authentication
sequence is used as a legal signature on many fi nancial transactions. Th is signature is usually the
name of the signing party plus other unique information from the message (e.g., date, time, or
dollar amount). Th is signature and the other information are encrypted by the sender using the
private key. Th e receiver uses the sender’s public key to decrypt the signature block and compares
the result to the name and other key contents in the rest of the message to ensure a match.
Th e only problem with this approach lies in ensuring that the person or organization
that sent the document with the correct private key is the actual person or organization.
Anyone can post a public key on the Internet, so there is no way of knowing for sure who
actually used it. For example, it would be possible for someone other than Organization A in
this example to claim to be Organization A when, in fact, he or she is an imposter.
Th is is where the Internet’s public key infrastructure (PKI) becomes important.16 Th e
PKI is a set of hardware, soft ware, organizations, and polices designed to make public key
encryption work on the Internet. PKI begins with a certifi cate authority (CA), which is a
trusted organization that can vouch for the authenticity of the person or organization using
authentication (e.g., VeriSign). A person wanting to use a CA registers with the CA and must
provide some proof of identify. Th ere are several levels of certifi cation, ranging from a simple
confi rmation from a valid e-mail address to a complete police-style background check with
an in-person interview. Th e CA issues a digital certifi cate that is the requestor’s public key,
encrypted using the CA’s private key as proof of identify. Th is certifi cate is then attached
to the user’s e-mail or Web transactions in addition to the authentication information. Th e
receiver then verifi es the certifi cate by decrypting it with the CA’s public key and must also
contact the CA to ensure that the user’s certifi cate has not been revoked by the CA.
Th e encryption and authentication requirements state what encryption and authentica-
tion requirements are needed for what data. For example, will sensitive data such as customer
credit-card numbers be stored in the database in encrypted form, or will encryption be used
to take orders over the Internet from the company’s website? Will users be required to use a
digital certifi cate in addition to a standard password?
16 For more on the PKI, see http://datatracker.ietf.org/wg/pkix/charter/.
Nonfunctional Requirements and Physical Architecture Layer Design 447
Virus Control Requirements Virus control requirements address the single most common
security problem: viruses. Studies have shown that almost 90 percent of organizations suff er a
virus infection each year. Viruses cause unwanted events—some harmless (such as nuisance
messages), some serious (such as the destruction of data). Any time a system permits data to
be imported or uploaded from a user’s computer, there is the potential for a virus infection.
Many systems require that all information systems that permit the import or upload of user
fi les to check those fi les for viruses before they are stored in the system.
Cultural and Political Requirements
Cultural and political requirements are those specifi c to the countries in which the system will
be used. In today’s global business environment, organizations are expanding their systems
to reach users around the world. Although this can make great business sense, its impact on
application development should not be underestimated. Yet another important part of the
design of the system’s physical architecture is understanding the global cultural and political
requirements for the system (see Chapter 10 and Figure 11-17).
Customization Requirements For global applications, the project team needs to give some
thought to customization requirements: How much of the application will be controlled by a
central group, and how much of the application will be managed locally? For example, some
companies allow subsidiaries in some countries to customize the application by omitting or
adding certain features. Th is decision has trade-off s between fl exibility and control because
customization oft en makes it more diffi cult for the project team to create and maintain the
application. It also means that training can diff er among diff erent parts of the organization,
and customization can create problems when staff moves from one location to another.
Owing to the use of diff erent languages, in some cases, specialized hardware that has been
customized to the local culture is required. For example, having specialized keyboards makes
sense for any language that does not use the typical Roman alphabet, e.g., Arabic, Hebrew,
Greek, Japanese, Korean, Mandarin, or Russian. Th ere are also emulators available for many
diff erent languages. Depending on the users being served, assistive devices could be required,
such as Braille devices, eye-tracking devices, head pointers, head/mouth stick keyboards, or
adaptive ability switches. Depending on the cultural and political requirements, many diff er-
ent hardware platforms might need to be considered.
FIGURE 11-17 Cultural and Political Requirements
Type of Requirement Defi nition Examples
Customization
Requirements
Specifi cation of what aspects of the system
can be changed by local users
• Country managers will be able to defi ne new
fi elds in the product database to capture
country-specifi c information.
• Country managers will be able to change the
format of the telephone number fi eld in the
customer database.
Legal Requirements The laws and regulations that impose
requirements on the system
• Personal information about customers can-
not be transferred out of European Union
countries into the United States.
• It is against U.S. federal law to divulge infor-
mation on who rented what videotape, so
access to a customer’s rental history is per-
mitted only to regional managers.
448 Chapter 11 Physical Architecture Layer Design
Legal Requirements Legal requirements are requirements imposed by laws and government
regulations. System developers sometimes forget to think about legal regulations; unfortu-
nately, forgetting comes at some risk because ignorance of the law is no defense. For example,
in 1997 a French court convicted the Georgia Institute of Technology of violating French lan-
guage law. Georgia Tech operated a small campus in France that off ered summer programs
for American students. Th e information on the campus Web server was primarily in English
because classes are conducted in English, which violated the law requiring French to be the
predominant language on all Internet servers in France. By formally considering legal regu-
lations, you are less likely to overlook them. Another major example is the recent European
court ruling regarding the user’s right to be forgotten.
Synopsis
In many cases, the technical environment requirements as driven by the business require-
ments can simply define the physical architecture layer. In this case, the choice is simple:
Business requirements dominate other considerations. For example, the business require-
ments might specify that the system needs to work over the Web using the customer’s
Web browser. In this case, the architecture probably should be a thin client–server. Such
business requirements are most likely in systems designed to support external custom-
ers. Internal systems can also impose business requirements, but usually they are not as
restrictive.
In the event that the technical environment requirements do not stipulate a specifi c
architecture, then the other nonfunctional requirements become important. Even in cases
when the business requirements drive the architecture, it is still important to work through
and refi ne the remaining nonfunctional requirements because they are important in later
stages of design and implementation.
Operational Requirements System integration requirements can lead to one architecture
being chosen over another, depending on the architecture and design of the system(s) with
which the system needs to integrate. For example, if the system must integrate with a desktop
system (e.g., Excel), this might suggest a thin or thick client–server architecture, whereas if
it must integrate with a server-based system, a server-based architecture may be indicated.
Systems that have extensive portability requirements tend to be best suited for a thin client–
server architecture because it is simpler to write for Web-based standards (e.g., HTML, XML)
that extend the reach of the system to other platforms, rather than trying to write and rewrite
extensive presentation logic for diff erent platforms in the server-based, client-based, or thick
client–server architectures. Systems with extensive maintainability requirements might not
be well suited to client-based or thick client–server architectures because of the need to rein-
stall soft ware on the desktops.
Performance Requirements Generally speaking, information systems that have high perfor-
mance requirements are best suited to client–server architectures. Client–server architectures
are more scalable, which mean they respond better to changing capacity needs and thus
enable the organization to better tune the hardware to the speed requirements of the system.
Client–server architectures that have multiple servers in each tier should be more reliable
and have greater availability, because if any one server crashes, requests are simply passed to
other servers, and users might not even notice (although response time could be worse). In
practice, however, reliability and availability depend greatly on the hardware and operating
system, and Windows-based computers tend to have lower reliability and availability than
Linux or mainframe computers.
Verifying and Validating the Physical Architecture Layer 449
Security Requirements Generally speaking, because all soft ware is in one location and
because mainframe operating systems are more secure than microcomputer operating systems,
server-based architectures tend to be more secure. For this reason, high-value systems are more
likely to be found on mainframe computers, even if the mainframe is used as a server in client–
server architectures. In today’s Internet-dominated world, authentication and encryption tools
for Internet-based client–server architectures are more advanced than those for mainframe
server-based architectures. Viruses are potential problems in all architectures because they
easily spread on desktop computers. If a server-based system can reduce the functions needed
on desktop systems, then they may be more secure.
Cultural and Political Requirements As cultural and political requirements become
more important, the ability to separate the presentation logic from the application logic
and the data becomes important. Such separation makes it easier to develop the presenta-
tion logic in different languages while keeping the application logic and data the same. It
also makes it easier to customize the presentation logic for different users and to change
it to better meet cultural norms. To the extent that the presentation logic provides access
to the application and data, it also makes it easier to implement different versions that
enable or disable different features required by laws and regulations in different coun-
tries. This separation is the easiest in thin client–server architectures, so systems with
many cultural and political requirements often use thin client–server architectures. As
with system integration requirements, the impact of legal requirements depends on the
specific nature of the requirements, but in general, client-based systems tend to be less
flexible.
VERIFYING AND VALIDATING THE PHYSICAL
ARCHITECTURE LAYER
Like the models on the other layers, the infrastructure design and the hardware and soft-
ware specifications of the physical architecture layer need to be verified and validated.
Verifying and validating the design of the data management layer fall into three basic
groups.
First, we recommend verifying and validating deployment diagrams by ensuring that all
of them are in fact consistent and balanced. For example, each of the nodes in a top-level net-
work model deployment diagram should be associated with a separate deployment diagram
that represents the low-level network model for the node.
Second, the hardware and soft ware specifi cations should be consistent with the
“lowest-level” network models. For example, if a low-level network model for an offi ce
describes a set of workstations, servers, printers, switches, routers, etc., then the hardware
and soft ware specifi cation for that location should be the details for each of the IT artifacts
for that location.
Th ird, once the system has been implemented, testing of the nonfunctional requirements
becomes crucial. In this case, tests must be designed and performed for each of the nonfunc-
tional requirements. For example, for the performance requirements, load testing must be
performed to identify possible performance bottlenecks in the network. We will return to
this topic in Chapter 12.
450 Chapter 1 Physical Architecture Layer Design
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
Sam Wilson, the infrastructure analyst, took the lead in designing this layer. Sam was
responsible for ensuring that the new system adheres to the infrastructure standards
in place at Patterson while making sure that the infrastructure can support the new
system. Th e team had to decide which architecture would best meet the needs of the
Integrated Health Clinic Delivery system and to determine what soft ware would be
placed on which hardware. Next, the team had to map the network model on a deploy-
ment diagram. Accomplishing this task required close communication with the other
layers, including the problem domain, data management, and human–computer inter-
action layers.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the four major architectural components.
Describe the three major physical architectures.
Discuss the trade-off s in selecting a physical architecture.
Discuss cloud computing.
Describe Infrastructure as a Service, Platform as a Service, and Soft ware as a Service.
Describe the three primary obstacles to cloud computing.
Discuss the potential impact of ubiquitous computing and the Internet of Th ings.
Describe enchanted objects.
Discuss the three general questions that Green IT tries to address.
Create an infrastructure design using deployment diagrams.
Create a high-level hardware and soft ware specifi cation.
Describe how nonfunctional requirements may infl uence the actual design of the physical
architecture layer.
KEY TERMS
24/7
Access control list
Access control requirements
Alternative matrix
Application logic
Architectural component
Artifact
Asymmetric encryption
algorithm
Authentication
Availability and reliability
requirements
Benchmark
Business process
Capacity requirements
Certifi cate authority (CA)
Client-based architecture
Client computer
Client–server architecture
Cloud computing
Communication path
Cultural and political
requirements
Customization requirements
Data access logic
Data storage
Deployment diagrams
E-waste
Enchanted objects
Encryption
Fat client
Graphical user interface
(GUI)
Green data centers
Green IT
Grid computing
Questions 451
Hardware and soft ware
specifi cation
Hybrid cloud
Infrastructure as a service
(IaaS)
Internet of Th ings (IoT)
Invertible
Legal requirements
Locations
Maintainability
requirements
Middleware
Mission critical system
Multitenancy
Network
Network model
Node
N-tiered architecture
Operational requirements
Outsourcing
Parkinson’s law
Performance requirements
Paperless offi ce
Platform as a service (PaaS)
Portability requirements
Presentation logic
Private cloud
Private key
Public cloud
Public key
Public key encryption
Response time
Scalable
Security requirements
Server
Server-based architecture
Service
Service-oriented
architecture
Soft ware as a service
(SaaS)
Speed requirements
SQL (structured query
language)
Symmetric encryption
algorithm
System integration
requirements
Technical environment
requirements
Th ick client
Th in client
Th ree-tiered architecture
Timesharing
Total cost of ownership
Two-tiered architecture
Ubiquitous computing
Virtual memory
Virtualization
Virus
Virus control requirements
Web services
QUESTIONS
1. What are the four basic functions of any information
system?
2. What are the three primary hardware components of
any physical architecture?
3. Name two examples of a server.
4. Compare and contrast server-based architectures,
client-based architectures, and client–server-based
architectures.
5. What is the biggest problem with server-based com-
puting?
6. What is the biggest problem with client-based computing?
7. Describe the major benefi ts and limitations of thin
client–server architectures.
8. Describe the major benefi ts and limitations of thick
client–server architectures.
9. Describe the diff erences among two-tiered, three-
tiered, and n-tiered architectures.
10. Defi ne scalable. Why is this term important to system
developers?
11. What six criteria are helpful to use when comparing
the appropriateness of computing alternatives?
12. Why should the project team consider the exist-
ing physical architecture in the organization when
designing the physical architecture layer of the new
system?
13. Name the three diff erent types of clouds. How do they
diff er from one another?
14. What is meant by a service-oriented architecture?
15. Defi ne virtualization. How does it relate to the cloud?
16. What are the diff erences among IaaS, PaaS, and SaaS?
17. What are the obstacles for provisioning the physical
architecture layer with cloud technologies?
18. What, if any, are the issues related to security in the
cloud?
19. What are SOX and HIPAA, and how could they aff ect
a fi rm’s decision to adopt cloud technology?
20. What is meant by ubiquitous computing? How about
the Internet of Th ings?
21. What is an enchanted object? Give a set of examples
of them.
22. What is e-waste?
23. What is the problem with backyard recycling of e-waste?
24. What is meant by a green data center?
25. How do tablets, such as the iPadTM, enable the paper-
less offi ce?
26. What additional hardware- and soft ware-associated
costs might need to be included on the hardware and
soft ware specifi cation?
27. Who is ultimately in charge of acquiring hardware
and soft ware for a project?
28. What is a benchmark, and why is it important?
29. Why is Parkinson’s Law relevant to the design of the
physical architecture layer?
30. What do you think are three common mistakes that
novice analysts make in architecture design and hard-
ware and soft ware specifi cation?
31. Describe the major nonfunctional requirements and
how they infl uence physical architecture layer design.
452 Chapter 1 Physical Architecture Layer Design
32. Why is it useful to defi ne the nonfunctional require-
ments in more detail even if the technical environ-
ment requirements dictate a specifi c architecture?
33. What does the network model communicate to the
project team?
34. What are the diff erences between the top-level net-
work model and the low-level network model?
35. Are some nonfunctional requirements more impor-
tant than others in infl uencing the architecture design
and hardware and soft ware specifi cation?
36. What do you think are the most important security
issues for a system?
EXERCISES
A. Using the Web (or past issues of computer industry
magazines such as Computerworld), locate a system
that runs in a server-based environment. Based on
your reading, why do you think the company chose
that computing environment?
B. Using the Web (or past issues of computer industry
magazines such as Computerworld), locate a system
that runs in a client–server environment. Based on
your reading, why do you think the company chose
that computing environment?
C. Using the Web, locate examples of a mainframe com-
ponent, a minicomputer component, and a micro-
computer component. Compare the components in
terms of price, speed, available memory, and disk
storage. Did you fi nd large diff erences in prices when
the performances of the components are considered?
D. You have been selected to fi nd the best client–server
architecture for a Web-based order entry system that
is being developed for L.L. Bean. Write a short memo
that describes to the project manager your reason for
selecting an n-tiered architecture over a two-tiered
architecture. In the memo, give some idea as to what
diff erent components of the architecture you would
include.
E. Th ink about the system that your university currently
uses for career services, and suppose that you are
in charge of replacing the system with a new one.
Describe how you would decide on the computing
architecture for the new system using the criteria
presented in this chapter. What information will you
need to fi nd out before you can make an educated
comparison of the alternatives?
F. Using the Web, fi nd information on the eff ects that
e-waste and backyard recycling has on developing
countries. Based on you fi nd, what Green IT policies
would you suggest a fi rm put in place to minimize the
negative eff ects of e-waste.
G. Using the Web, fi nd examples of company’s pursuing
a Green IT strategy. Describe what they are doing.
H. Energy Star is a joint program between the US Depart-
ment of Energy and the Environmental Protection
Agency. What are the requirements for various IT
devices to be certifi ed as being Energy Star compliant?
I. Using the Web, fi nd examples of fi rms using the cloud
as a basis for the physical architecture layer. Describe
exactly what they are doing.
J. Locate a consumer products company on the Web
and read its company description (so that you get a
good understanding of the geographic locations of the
company). Pretend that the company is about to cre-
ate a new application to support retail sales over the
Web. Create a high-level network model that depicts
the locations that would include components that
support this application.
K. Create a low-level network diagram for the building
that houses the computer labs at your university.
Choose an application (e.g., course registration, stu-
dent admissions) and include only the components
that are relevant to that application.
L. An energy company with headquarters in Dallas,
Texas, is thinking about developing a system to track
the effi ciency of its oil refi neries in North America.
Each week, the ten refi neries—as far as Valdez, Alaska,
and as close as San Antonio, Texas—will upload
performance data via satellite to the corporate main-
frame in Dallas. Production managers at each site will
use a personal computer to connect to an Internet ser-
vice provider and access reports via the Web. Create
a high-level network model that depicts the locations
that have components supporting this system.
M. Suppose that your mother is a real estate agent, and
she has decided to automate her daily tasks using a
laptop computer. Consider her potential hardware and
soft ware needs, and create a hardware and soft ware
Minicases 453
specifi cation that describes them. Th e specifi cation
should be developed to help your mother buy her
hardware and soft ware on her own.
N. Suppose that the admissions offi ce in your university
has a Web-based application so that students can apply
for admission online. Recently, there has been a push to
admit more international students into the university.
What do you recommend that the application include
to ensure that it supports this global requirement?
O. Based on the A Real Estate Inc. problem in Chapter 4
(exercises I, J, and K), Chapter 5 (exercises P and Q),
Chapter 6 (exercise D), Chapter 7 (exercise A), Chapter
8 (exercise A), Chapter 9 (exercise L), and Chapter 10
(exercises I and J), suggest a physical architecture design
and portray it with a deployment diagram.
P. Based on the A Video Store problem in Chapter 4
(exercises L, M, and N), Chapter 5 (exercises R and
S), Chapter 6 (exercise E), Chapter 7 (exercise B),
Chapter 8 (exercise B), Chapter 9 (exercise M), and
Chapter 10 (exercises K and L), suggest a physical
architecture design and portray it with a deployment
diagram.
Q. Based on the gym membership problem in Chapter
4 (exercises O, P, and Q), Chapter 5 (exercises T and
U), Chapter 6 (exercise F), Chapter 7 (exercise C),
Chapter 8 (exercise C), Chapter 9 (exercise N), and
Chapter 10 (exercises M and N), suggest a physical
architecture design and portray it with a deployment
diagram.
R. Based on the Picnics R Us exercises in Chapter 4
(exercises R, S, and T), Chapter 5 (exercises V and
W), Chapter 6 (exercise G), Chapter 7 (exercise D),
Chapter 8 (exercise D), and Chapter 9 (exercise O),
and Chapter 10 (exercises O and P), suggest a physical
architecture design and portray it with a deployment
diagram.
S. Based on the Of-the-Month-Club problem in Chapter
4 (exercises U, V, and W), Chapter 5 (exercises X and
Y), Chapter 6 (exercise H), Chapter 7 (exercise E),
Chapter 8 (exercise E), Chapter 9 (exercise N), and
Chapter 10 (exercises Q and R), suggest a physical
architecture design and portray it with a deployment
diagram.
MINICASES
1. Th e system development project team at Birdie
Masters golf schools has been working on defi ning
the physical architecture design for the system. Th e
major focus of the project is a networked school loca-
tion operations system, allowing each school location
to easily record and retrieve all school location trans-
action data. Another system element is the use of the
Internet to enable current and prospective students
to view class off erings at any of the Birdie Masters’
locations, schedule lessons and enroll in classes at any
Birdie Masters location, and maintain a student pro-
gress profi le—a confi dential analysis of the student’s
golf skill development.
Th e project team has been considering the globali-
zation issues that should be factored into the archi-
tecture design. Th e school’s plan for expansion into
the golf-crazed Japanese market is moving ahead. Th e
fi rst Japanese school location is tentatively planning
to open about six months aft er the target completion
data for the system project. Th erefore, it is important
that issues related to the international location be
addressed now during design.
Assume that you have been given the responsibility
of preparing a summary memo on the globalization
issues that should be factored into the design. Prepare
this memo discussing the globalization issues that are
relevant to Birdie Masters’ new system.
2. Jerry is a relatively new member of a project team that
is developing a retail store management system for a
chain of sporting goods stores. Company headquarters
is in Las Vegas, and the chain has twenty-seven loca-
tions throughout Nevada, Utah, and Arizona. Several
cities have multiple stores.
Th e new system will be a networked client–server
architecture. Stores will be linked to one of three
regional servers, and the regional servers will be linked
to corporate headquarters in Las Vegas. Th e regional
servers also link to one another. Each retail store
will be outfi tted with similar confi gurations of two
PC-based point-of-sale terminals networked to a local
fi le server. Jerry has been given the task of developing
a network model that will document the geographic
structure of this system. He has not faced a system of
this scope before and is a little unsure how to begin.
454 Chapter 11 Physical Architecture Layer Design
a. Prepare a set of instructions for Jerry to follow in
developing this network model.
b. Using a deployment diagram, draw a network
model for this organization.
c. Prepare a set of instructions for Jerry to follow in
developing a hardware and soft ware specifi cation.
3. Refer to the Professional and Scientifi c Staff Manage-
ment (PSSM) minicase in Chapters 4, 6, 7, 8, 9, and 10.
Based on the solutions developed for those problems,
suggest a physical architecture design and portray it
with a deployment diagram.
4. Refer to the Holiday Travel Vehicles minicase in Chap-
ters 5, 6, 7, 8, 9, and 10. Based on the solutions devel-
oped for those problems, suggest a physical architecture
design and portray it with a deployment diagram.
During construction, the actual system is built. Building a successful
information system requires a set of activities: programming,
testing, and documenting the system. In today’s global economy,
cultural issues also play an important role in managing these
activities. Installing an information system requires switching from
the current system to the new system. Th is conversion process can
be quite involved; for example, cultural diff erences among the users,
the development team, and the two groups can be quite challenging.
Furthermore, not only does conversion involve shutting the old
system down and turning the new one on, it also can involve
a signifi cant training eff ort. Finally, operating the system may
uncover additional requirements that may have to be addressed by
the development team.
Program
s
Test Plan
D
ocum
entation
C
onversion
Plan
C
hange
M
anagem
ent
Plan
Support
Plan
Project
A
ssessm
ent
CHAPTER 12
Construction
P A R T T H R E E
Construction,
Installation, and
Operations
CHAPTER 13
Installation and
Operations
456
This chapter discusses the activities needed to successfully build an information system:
programming, testing, and documenting the system. Programming is time-consuming and
costly, but except in unusual circumstances, it is the simplest for the systems analyst because
it is well understood. For this reason, the systems analyst focuses on testing (proving that the
system works as designed) and developing documentation.
OBJECTIVES
■ Understand the basic issues related to managing programmers.
■ Understand how cultural issues can impact the effi ciency, eff ectiveness, and focus of
soft ware development teams.
■ Be familiar with the diff erent types of documentation.
■ Understand how to develop documentation.
■ Understand how object-orientation eff ects soft ware testing.
■ Understand the diff erent types of and purpose of unit tests.
■ Understand the diff erent types of and purpose of integration tests.
■ Understand the diff erent types of and purpose of system tests.
■ Understand the diff erent types of and purpose of acceptance tests.
INTRODUCTION
When people fi rst learn about developing information systems, they usually think imme-
diately about writing programs. Programming can be the largest single component of any
systems development project in terms of time and cost. However, it also can be the best
understood component and therefore—except in rare circumstances—off ers the fewest prob-
lems of all aspects of system development. When projects fail, it is usually not because the
programmers were unable to write the programs, but because the analysis, design, installa-
tion, and/or project management were done poorly.
Construction is the development of all parts of the system, including the soft ware
itself, documentation, and new operating procedures. Looking back at Figure 1-18, we see
that the Construction phase of the Enhanced Unifi ed Process deals predominantly with
the Implementation, Testing, and Confi guration and Change Management workfl ows.
Implementation obviously deals with programming. Programming is oft en seen as the focal
point of systems development. Aft er all, systems development is writing programs. It is the
reason we do all the analysis and design. And it’s fun. Many beginning programmers see
testing and documentation as bothersome aft erthoughts. Testing and documentation aren’t
fun, so they oft en receive less attention than the creative activity of writing programs.
However, programming and testing are very similar to writing and editing. No profes-
sional writer (or good student writing an important term paper) would stop aft er writing
the fi rst draft . Rereading, editing, and revising the initial draft into a good paper are the
C H A P T E R 1 2
Construction
Managing Programming 457
hallmarks of good writing. Likewise, thorough testing is the hallmark of professional soft –
ware developers. Most professional organizations devote more time and money to testing
(and the subsequent revision and retesting) than to writing the programs in the fi rst place.
Th e reasons are simple economics: Downtime and failures caused by soft ware bugs1 are
extremely expensive. Many large organizations estimate the costs of downtime of critical appli-
cations at $50,000 to $200,000 per hour.2 One serious bug that causes an hour of downtime can
cost more than one year’s salary of a programmer—and how oft en are bugs found and fi xed
in one hour? Testing is, therefore, a form of insurance. Organizations are willing to spend a
lot of time and money to prevent the possibility of major failures aft er the system is installed.
Th erefore, a program is usually not considered fi nished until the test for that program
is passed. For this reason, programming and testing are tightly coupled, and because pro-
gramming is the primary job of the programmer (not the analyst), testing (not program-
ming) oft en becomes the focus of the construction stage for the systems analysis team.
Th e Confi guration and Change Management workfl ow keeps track of the state of the
evolving system. Th e evolving information system comprises a set of artifacts that include,
for example, diagrams, source code, and executables. During the development process,
these artifacts are modifi ed. Th e amount of work, and hence dollars, that goes into the
development of the artifacts is substantial. Th erefore, the artifacts themselves should be
handled as any expensive asset would be handled: Access controls must be put into place
to safeguard the artifacts from being stolen or destroyed. Because the artifacts are modifi ed
on a regular, if not on a continuous basis, good version control mechanisms should be
established. Th e traceability of the artifacts back through the various artifacts developed,
such as data management layer designs, class diagrams, package diagrams, and use-case
diagrams, to the specifi c requirements is also very important. Without this traceability, we
will not know which aspects of a system to modify when—not if—the requirements change.
MANAGING PROGRAMMING
In general, systems analysts do not write programs; programmers write programs. Th erefore,
the primary task of the systems analysts during programming is . . . waiting. However, the
project manager is usually very busy managing the programming eff ort by assigning the pro-
grammers, coordinating the activities, and managing the programming schedule.3
Assigning Programmers
Th e fi rst step in programming is assigning modules to the programmers. As discussed in
Chapter 8, each module (class, object, or method) should be as separate and distinct as
possible from the other modules (i.e., cohesion should be maximized and coupling should
be minimized). Th e project manager fi rst groups together classes that are related so that
each programmer is working on related classes. Th ese groups of classes are then assigned to
programmers. A good place to start is to look at the package diagrams.
1 When I (Alan Dennis) was an undergraduate, I had the opportunity to hear Admiral Grace Hopper tell how the
term bug was introduced. She was working on one of the early Navy computers when suddenly it failed. Th e computer
would not restart properly, so she began to search for failed vacuum tubes. She found a moth inside one tube and
recorded in the log book that a bug had caused the computer to crash. From then on, every computer crash was jok-
ingly blamed on a bug (as opposed to programmer error), and eventually the term bug entered the general language
of computing.
2 See Billie Shea, “Quality Patrol: Eye on the Enterprise,” Application Development Trends (November 5, 1998): 31–38.
3 One of the best books on managing programming (even though it was fi rst written more than 30 years ago) is that by
Frederick P. Brooks, Jr. Th e Mythical Man-Month, 20th Anniversary Edition (Reading, MA: Addison-Wesley, 1995).
458 Chapter 12 Construction
One of the rules of systems development is that the more programmers who are
involved in a project, the longer the system will take to build. Th is is because as the size of
the programming team increases, the need for coordination increases exponentially, and
the more coordination required, the less time programmers can spend actually writing
systems (see Chapter 2). Th e best size is the smallest possible programming team. When
projects are so complex that they require a large team, the best strategy is to try to break the
project into a series of smaller parts that can function as independently as possible.
Coordinating Activities
Coordination can be done through both high-tech and low-tech means. Th e simplest
approach is to have a weekly project meeting to discuss any changes to the system that have
arisen during the past week—or any issues that have come up. Remember, agile development
approaches such as Scrum encourage daily meetings (see Chapter 2). Regular meetings, even
if they are brief, encourage the widespread communication and discussion of issues before
they become problems.
Another important way to improve coordination is to create and follow standards that
can range from formal rules for naming fi les, to forms that must be completed when goals
are reached, to programming guidelines (see Chapter 2). When a team forms standards and
then follows them, the project can be completed faster because task coordination is less
complex.
Th e analysts also must put mechanisms in place to keep the programming eff ort well
organized. Many project teams set up three areas in which programmers can work: a devel-
opment area, a testing area, and a production area. Th ese areas can be diff erent directories
on a server hard disk, diff erent servers, or diff erent physical locations, but the point is that
fi les, data, and programs are separated based on their status of completion. At fi rst, pro-
grammers access and build fi les within the development area and then copy them to the
testing area when the programmers are fi nished. If a program does not pass a test, it is sent
back to development. Once all programs are tested and ready to support the new system,
they are copied into the production area—the location where the fi nal system will reside.
Keeping fi les and programs in diff erent places based on completion status helps man-
age change control, the action of coordinating a system as it changes through construction.
Another change control technique is keeping track of which programmer changes which
classes and packages by using a program log. Th e log is merely a form on which programmers
sign out classes and packages to write and sign in when they are completed. Both the pro-
gramming areas and program log help the analysts understand exactly who has worked on
what and the system’s current status. Without these techniques, fi les can be put into produc-
tion without the proper testing (e.g., two programmers can start working on the same class
or package at the same time).
If a CASE tool is used during the construction step, it can be very helpful for change
control because many CASE tools are set up to track the status of programs and help manage
programmers as they work. In most cases, maintaining coordination is not conceptually com-
plex. It just requires a lot of discipline and attention to tracking small details.
Managing the Schedule
Th e time estimates that were produced during project identifi cation and refi ned during analy-
sis and design almost always need to be refi ned as the project progresses during construction
because it is virtually impossible to develop an exact assessment of the project’s schedule. As
we discussed in Chapter 2, a well-done set of time estimates usually has a 10 percent margin
of error by the time we reach the construction step. It is critical that the time estimates be
Managing Programming 459
4 Based upon material from Steve McConnell, Rapid Development (Redmond, WA: Microsoft Press, 1996).
revised as construction proceeds. If a program module takes longer to develop than expected,
then the prudent response is to move the expected completion date later by the same amount.
One of the most common causes for schedule problems is scope creep. Scope creep
occurs when new requirements are added to the project aft er the system design was fi nalized
(see Chapter 2). If you recall, Scrum encourages adding new requirements to the product
backlog instead of letting the scope of the project change between sprints. Scope creep can
be very expensive because changes made late in system development can require much of the
completed system design (and even programs already written) to be redone. Any proposed
change during construction must require the approval of the project manager and should
only be done aft er a quick cost–benefi t analysis has been done.
Another common cause is the unnoticed day-by-day slippages in the schedule. One package
is a day late here; another one is a day late there. Pretty soon these minor delays add up and the
project is noticeably behind schedule. Once again, the key to managing the programming eff ort
is to watch these minor slippages carefully and update the schedule accordingly.
Typically, a project manager creates a risk assessment that tracks potential risks along
with an evaluation of their likelihood and potential impact. As the construction step moves
to a close, the list of risks changes as some items are removed and others surface. Th e best
project managers, however, work hard to keep risks from having an impact on the schedule
and costs associated with the project.
In previous chapters, we discussed classic mistakes and
how to avoid them. Here, we summarize four classic
mistakes in implementation:4
■ Research-oriented development: Using state-of-the-art
technology requires research-oriented development
that explores the new technology because “bleeding
edge” tools and techniques are not well understood,
are not well documented, and do not function exactly
as promised.
Solution: If you use state-of-the-art technology, you
need to signifi cantly increase the project’s time and
cost estimates even if (some experts would say espe-
cially if) such technologies claim to reduce time and
effort.
■ Using low-cost personnel: You get what you pay for.
The lowest-cost consultant or staff member is signifi –
cantly less productive than the best staff. Several stud-
ies have shown that the best programmers produce
software six to eight times faster than the least produc-
tive (yet cost only 50 to 100 percent more).
Solution: If cost is a critical issue, assign the best, most
expensive personnel; never assign entry-level person-
nel in an attempt to save costs.
■ Lack of code control: On large projects, programmers
need to coordinate changes to the program source
code (so that two programmers don’t try to change the
same program at the same time and overwrite each
other’s changes). Although manual procedures appear
to work (e.g., sending e-mail notes to others when you
work on a program to tell them not to), mistakes are
inevitable.
Solution: Use a source code library that requires
programmers to “check out” programs and prohibits
others from working on them at the same time.
■ Inadequate testing: The number one reason for project
failure during implementation is ad hoc testing—where
programmers and analysts test the system without formal
test plans.
Solution: Always allocate suffi cient time in the project
plan for formal testing.
12-1 Avoiding Classic Implementation MistakesPRACTICAL
TIP
460 Chapter 12 Construction
Cultural Issues
One of the major issues facing information systems development organizations is the off shor-
ing of the implementation aspects of information systems development. Confl icts caused by
diff erent national and organizational cultures are now becoming a real area of concern. With
the potential of cloud computing (see Chapter 11) potentially enabling even more outsourc-
ing, the potential of cultural confl ict is even greater.
A simple example that can demonstrate cultural diff erences with regard to student
learning is the idea of plagiarism. What exactly does plagiarism really imply? Diff erent
cultures have very diff erent views. In some cultures, one of the highest forms of respect is
simply to quote an expert. However, in these same cultures, there is no need to reference
the expert. Th e act of quoting the expert itself is the act of respect. In some cases, actually
referencing the expert through the use of quotation marks and a footnote may be viewed as
an insult to the expert and the reader because it is obvious to the reader that the writer did
not expect the reader to recognize the expert’s quote. Th is expectation was caused by either
the reader’s own ignorance or the expert’s lack of reputation. Either way, the writer would
be insulting someone through the use of quotation marks and footnotes. Th ese cultures tend
to be collectivist in nature (see Chapters 10 and 13). Consequently, since the collective owns
all ideas, there is no concept of theft of ideas. However, in the United States, the opposite is
true. If a writer does not use quotation marks and footnotes to appropriately give credit to the
source of the quote (or paraphrase), then the writer is guilty of theft .5 Obviously, in today’s
global world, plagiarism is not a simple issue.
Another simple example of cultural diff erences, with regard to student learning, is the
idea of students working together to complete homework assignments. Even though we all
know that research has shown that students learn better in groups, in the United States, we
view students who turn in the same assignment as cheaters.6 In other cultures, individual per-
formance is not as important as the performance of the group. Again, these cultures are col-
lectivist in nature. Consequently, helping a fellow student to understand the assignment and
to perform better in the class would be the expectation. Furthermore, this attitude extends
to test taking. If a fellow student is struggling on a test and if you were from a collectivist
culture, it would be your duty to allow your fellow student to copy your answer. Obviously,
this is another example of a substantive cultural diff erence. From a business perspective, these
diff erent views of plagiarism and cheating could have serious implications for the protection
of intellectual property.
As we stated previously, with off shore outsourcing, information systems development
teams can be geographically dispersed and multicultural in their membership. Given the
above issues and when we consider the cultural diff erences Hall and Hofstede identifi ed
(see Chapter 10), cultural issues add a new wrinkle in the management of developing a suc-
cessful information system.7 From an information systems development perspective, context
could infl uence the ability of a team member to see (or not see) potential creative solutions
that are out of the box or aff ect a team member’s ability (or inability) to understand the
entire problem under consideration. Furthermore, given this dimension, the level of detail in
5 A wonderful little book on plagiarism is Richard A. Posner, Th e Little Book of Plagiarism (New York: Pantheon
Books, 2007).
6 In this case, the recent work of Roger Schank is very enlightening. For example, see Roger C. Schank, Making Minds
Less Well Educated than Our Own (Mahwah, NJ: Lawrence Erlbaum Associates, 2004).
7 See Geert Hofstede, Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations across
Nations, 2nd Ed. (Th ousand Oaks, CA: Sage, 2001); Geert Hofstede, Gert Jan Hofstede, and Michael Minkov, Cultures
and Organizations: Soft ware of the Mind, 3rd Ed. (New York, NY: McGraw-Hill, 2010); Edward T. Hall, Beyond Culture
(New York: Anchor Books, 1981).
Managing Programming 461
direction could be varied between cultures. Hofstede’s individualism and collectivism dimen-
sion partially explains the results regarding plagiarism and cheating described above. Given
the importance that intellectual property plays in IT, this potentially could be a real problem
when off shoring development to a collectivist culture. Furthermore, Hall’s speed of messages
and context dimensions could also aff ect the way this could be addressed. Depending on the
culture, too much detail could be insulting, but attempting to put this issue in to a contextual
frame that is culturally sensitive is diffi cult.
When managing programmers in a multicultural setting, Hall’s time dimension must also
be considered. In monochronic time cultures, deadlines are critical. Th is is probably why time-
boxing has been relatively successful as a method to control projects (see Chapter 2). However,
in a polychronic time culture, a deadline is nothing more than a suggestion. Obviously, when
managing programmers, understanding how the culture considers time is very important to
have both a successful product delivery and a successful development process.
Hofstede’s other previously mentioned dimensions are power distance, uncertainty
avoidance, and masculinity versus femininity. Managing programmers in a culture with a
high power distance value is diff erent than with a culture with a low power distance. For
example, in the United States, programmers see themselves as equals to their managers.
In fact, in some fi rms, the president of the fi rm can be found “coding” solutions along
side of a brand new hire. Th is somewhat explains the growing popularity of agile methods
(see Chapter 1). In comparison, in a high power distance culture, the president of the fi rm
would never stoop to performing the same tasks as a new hire. It would be insulting to the
president and embarrassing to the new hire.
With regard to uncertainty avoidance, the choice of systems development approach
could be aff ected. In a culture that prefers everything to be neat and ordered, a systems devel-
opment methodology that is very rule-driven would be benefi cial. Also, development team
member professional certifi cation and team and fi rm ISO or CMMI certifi cations would lend
credibility to the team, whereas in a culture that willingly takes on risk, certifi cations might
not increase the perceived standing of the development team.
When managing programmers in a masculine culture, it is critical to provide recognition to
the top-performing members of the development team and also to recognize the top-performing
teams. On the other hand, when considering a feminine culture, it is more important to ensure
that the workplace is a supportive, noncompetitive, and nurturing environment.
Hofstede has identifi ed a fi ft h dimension, long-versus short-term orientation, which deals
with how the culture views the past and the future. In a long-term focused culture, team
development and a deep relationship with a client is very important, while in a culture that
emphasizes the short term, delivering a high-quality product on time is all that really matters.
For years, project managers in the United States have had to bring together individuals
from very diff erent backgrounds. Moreover, there was always a common spoken and writ-
ten language, English, and the melting pot idea that guaranteed some level of commonality
among the team members.8 However, in today’s “fl at world,” there is no longer any common
culture or common spoken and written language. From an information systems develop-
ment perspective, the common languages tend to be UML, Java, SQL, C11, Objective-C, and
Visual Basic, not English. However, at this time, there is no common culture. Consequently,
understanding cultural issues will be extremely important for the near future to successfully
manage international and multicultural development teams.
8 People who grew up in diff erent areas of the United States (e.g., New York City, Nashville, Minneapolis, Denver, and
Los Angeles) are, in a very real sense, culturally diff erent. For an interesting take on this, see Joel Garreau, Th e Nine
Nations of North America (New York NY: Avon Books, 1981). However, the prevalence of the Internet and cable TV
has created much more of a shared culture in the United States than in many other parts of the world. Obviously, the
Internet and cable TV also could aff ect the world in the long run.
462 Chapter 12 Construction
DEVELOPING DOCUMENTATION
Developing documentation of the system must be done throughout system development.
In many ways, the documentation of a system is a system. So, developing documenta-
tion can follow a similar, but simpler, approach as soft ware development. In this case,
creating use cases and developing the user interfaces to the documentation make sense.
Th ere are two fundamentally diff erent types of documentation: system documentation
and user documentation. System documentation is intended to help programmers and
systems analysts understand the application soft ware and enable them to build it or
maintain it aft er the system is installed. System documentation is largely a by-product of
the systems analysis and design process and is created as the project unfolds. Each step
and phase produces documents that are essential in understanding how the system is or
is to be built, and these documents are stored in the project binder(s). In many object-
oriented development environments, it is possible to somewhat automate the creation of
detailed documentation for classes and methods. For example, in Java, if the programmers
use javadoc-style comments, it is possible to create HTML pages that document a class and
its methods automatically by using the javadoc utility.9 Because most programmers look
on documentation with much distaste, anything that can make documentation easier to
create is useful.
User documentation (such as user’s manuals, training manuals, and online help sys-
tems) is designed to help the user operate the system. Although most project teams expect
users to have received training and to have read the user’s manuals before operating the
system, unfortunately, this is not always the case. It is more common today—especially in
the case of commercial soft ware packages for microcomputers—for users to begin using the
soft ware without training or reading the user’s manuals. In this section, we focus on user
documentation.10
User documentation is oft en left until the end of the project, which is a dangerous
strategy. Developing good documentation takes longer than many people expect because it
requires much more than simply writing a few pages. Producing documentation requires
designing the documents (whether on paper or online), writing the text, editing the docu-
ments, and testing them. For good-quality documentation, this process usually takes about
three hours per page (single-spaced) for paper-based documentation or two hours per screen
for online documentation. Th us “simple” documentation, such as a ten-page user’s manual
and a set of twenty help screens, takes seventy hours. Of course, lower-quality documentation
can be produced faster.
Th e time required to develop and test user documentation should be built into the pro-
ject plan. Most organizations plan for documentation development to start once the interface
design and program specifi cations are complete. Th e initial draft of documentation is usually
scheduled for completion immediately aft er the unit tests are complete. Th is reduces (but
doesn’t eliminate) the chance that the documentation will need to be changed due to soft ware
changes and still leaves enough time for the documentation to be tested and revised before
the acceptance tests are started.
Although paper-based manuals are still important, online documentation is becoming
more pervasive. Paper-based documentation is simpler to use because it is more familiar to
users, especially novices who have less computer experience; online documentation requires
the users to learn one more set of commands. Paper-based documentation is also easier to
9 For those who have used Java, javadoc is how the JDK documentation from Sun is created.
10 For more information on developing documentation, see Th omas T. Barker, Writing Soft ware Documentation
(Boston: Allyn & Bacon, 1998).
Developing Documentation 463
fl ip through and gain a general understanding of its organization and topics and can be used
far away from the computer itself.
Th ere are four key strengths of online documentation that all but guarantee it will
become the dominant form for the 21st century. Searching for information is oft en simpler
(provided the help search index is well designed) because the user can type in a variety of
keywords to view information almost instantaneously, rather than having to search through
the index or table of contents in a paper document. Th e same information can be presented
several times in many diff erent formats, so that the user can fi nd and read the information
in the most informative way (such redundancy is possible in paper documentation, but
the cost and intimidating size of the resulting manual make it impractical). Online docu-
mentation provides many new ways for the user to interact with the documentation that
is not possible in static paper documentation. For example, it is possible to use links or
“tool tips” (i.e., pop-up text; see Chapter 10) to explain unfamiliar terms, and one can write
“show-me” routines that demonstrate on the screen exactly what buttons to click and text to
type. Finally, online documentation is signifi cantly less expensive to distribute than paper
documentation.
Types of Documentation
Th ere are three fundamentally diff erent types of user documentation: reference documents,
procedures manuals, and tutorials. Reference documents (also called the help system) are
designed to be used when the user needs to learn how to perform a specifi c function (e.g.,
updating a fi eld, adding a new record). Oft en people read reference information when they
have tried and failed to perform the function; writing reference documents requires special
care because the user is oft en impatient or frustrated when he or she begins to read them.
Procedures manuals describe how to perform business tasks (e.g., printing a monthly
report, taking a customer order). Each item in the procedures manual typically guides the
user through a task that requires several functions or steps in the system. Th erefore, each
entry is typically much longer than an entry in a reference document.
Tutorials—obviously—teach people how to use major components of a system
(e.g., an introduction to the basic operations of the system). Each entry in the tutorial is
typically longer still than the entries in procedures manuals, and the entries are usually
designed to be read in sequence (whereas entries in reference documents and procedures
manuals are designed to be read individually).
Regardless of the type of user documentation, the overall process for developing it is
similar to the process of developing interfaces (see Chapter 10). Th e developer fi rst designs
the general structure for the documentation and then develops the individual components
within it.
Designing Documentation Structure
Th e general structure used in most online documentation, whether reference documents,
procedures manuals, or tutorials, is to develop a set of documentation navigation controls that
lead the user to documentation topics. Th e documentation topics are the material that user
wants to read, whereas the navigation controls are the way the user locates and accesses a
specifi c topic. As such, using storyboards, windows navigation diagrams, and windows layout
diagrams is useful (see Chapter 10).
Designing the structure of the documentation begins by identifying the diff erent types of
topics and navigation controls that need to be included. Figure 12-1 shows a commonly used
structure for online reference documents (i.e., the help system). Th e documentation topics
generally come from three sources. Th e fi rst and most obvious source of topics is the set of
464 Chapter 12 Construction
FIGURE 12-1
Organizing Online
Reference
Documents
Topics
Tasks
Commands
Definitions
Links
Navigation Controls
Contents
Introduction
Basic Features
Finding Albums
Index
Finding
Finding Albums
Finding Artists
Finding Songs
Text Search
albums
announcements
articles
artists
Agent Search
Enter a question
Full
Search
———
Music
Category
———
Credit
Card
———
Move…
———
Copy…
———
Delete…
———
How to…
———
How to…
———
How to…
———
commands and menus in the user interface. Th is set of topics is very useful if the user wants
to understand how a particular command or menu is used.
However, the users oft en don’t know what commands to look for or where they are in
the system’s menu structure. Instead, users have tasks they want to perform, and rather than
thinking in terms of commands, they think in terms of their tasks. Th erefore, the second
Developing Documentation 465
and oft en more useful set of topics focuses on how to perform certain tasks, usually those in
the use scenarios, WND, and the real use cases from the user interface design (see Chapter
10). Th ese topics walk the user through the set of steps (oft en involving several keystrokes or
mouse clicks) needed to perform some task.
Th e third topic is defi nitions of important terms. Th ese terms are usually the use cases
and classes in the system, but sometimes they also include commands.
Th ere are fi ve general types of navigation controls for topics, but not all systems use all
fi ve types (see Figure 12-1). Th e fi rst is the table of contents that organizes the information
in a logical form, as though the users were to read the reference documentation from start
to fi nish. Th e index provides access into the topics based on important keywords, in the
same way that the index at the back of a book helps us fi nd topics. Text search provides
the ability to search through the topics either for any text the user types or for words that
match a developer-specifi ed set of words that is much larger than the set of words in the
index. Unlike the index, text search typically provides no organization to the words (other
than alphabetical). Some systems provide the ability to use an intelligent agent to help in the
search. Th e fi fth and fi nal navigation controls to topics are the hyperlinks between topics that
enable the user to click and move among topics.
Procedures manuals and tutorials are similar but oft en simpler in structure. Topics for
procedures manuals usually come from the use scenarios, WNDs, and the real use cases
developed during interface design and from other basic tasks the users must perform. Topics
for tutorials are usually organized around major sections of the system and the level of expe-
rience of the user. Most tutorials start with the basic, most commonly used commands and
then move into more complex and less commonly used commands.
Writing Documentation Topics
Th e general format for topics is fairly similar across application systems and operating
systems. Topics typically start with very clear titles, followed by some introductory text that
defi nes the topic and then by detailed, step-by-step instructions on how to perform what is
being described. Many topics include screen images to help the user fi nd items on the screen;
some also have tutorials and videos available online that demonstrate the functions of inter-
est to the user. Most also include navigation controls to enable the movement among topics,
usually at the top of the window, plus links to other topics. Some also include links to related
topics that include options or other commands and tasks the user might want to perform in
concert with the topic being read.
Writing the topic content can be challenging. It requires a good understanding of the
user (or more accurately the range of users) and a knowledge of what skills the users cur-
rently have and can be expected to import from other systems and tools they are using or
have used (including the system that the new system is replacing). Topics should always be
written from the viewpoint of the user and describe what the user wants to accomplish, not
what the system can do. Figure 12-2 provides some general guidelines to improve the quality
of documentation text.11
Identifying Navigation Terms
As we write the documentation topics, we also begin to identify the terms that will be
used to help users fi nd topics. Th e table of contents is usually the most straightforward,
because it is developed from the logical structure of the documentation topics, whether
11 One of the best books to explain the art of writing is William Strunk and E. B. White, Elements of Style, 4th Ed.
(Needham Heights, MA: Allyn & Bacon, 2000).
466 Chapter 12 Construction
Figure 12-2 Guidelines for Crafting Documentation Topics
Guideline Before the Guideline After the Guideline
Use the active voice: The active voice
creates more active and readable text by
putting the subject at the start of the
sentence, the verb in the middle, and the
object at the end.
Finding albums is done using the album
title, the artist’s name, or a song title.
You can fi nd an album by using the album
title, the artist’s name, or a song title.
Use e-prime style: E-prime style creates
more active writing by omitting all forms
of the verb to be.
The text you want to copy must be
selected before you click on the copy
button.
Select the text you want to copy before
you click on the copy button.
Use consistent terms: Always use the same
term to refer to the same items, rather
than switching among synonyms (e.g.,
change, modify, update).
Select the text you want to copy. Pressing
the copy button will copy the marked
text to the new location.
Select the text you want to copy. Pressing
the copy button will copy the selected
text to the new location.
Use simple language: Always use the
simplest language possible to accurately
convey the meaning. This does not mean
you should “dumb down” the text but
that you should avoid artifi cially infl ating
its complexity. Avoid separating subjects
and verbs and try to use the fewest words
possible. (When you encounter a com-
plex piece of text, try eliminating words;
you may be surprised at how few words
are really needed to convey meaning.)
The Georgia Statewide Academic and
Medical System (GSAMS) is a coopera-
tive and collaborative distance learning
network in the state of Georgia. The
organization in Atlanta that administers
and manages the technical and overall
operations of the currently more than
300 interactive audio and video telecon-
ferencing classrooms throughout Georgia
system is the Department of Administrative
Service (DOAS). (56 words)
The Department of Administrative Service
(DOAS) in Atlanta manages the Georgia
Statewide Academic and Medical System
(GSAMS), a distance learning network
with more than 300 teleconferencing
classrooms throughout Georgia. (29
words)
Use friendly language: Too often, docu-
mentation is cold and sterile because
it is written in a very formal manner.
Remember, you are writing for a person,
not a computer.
Blank disks have been provided to you
by Operations. It is suggested that you
ensure your data are not lost by making
backup copies of all essential data.
You should make a backup copy of all
data that are important to you. If you
need more diskettes, contact Operations.
Use parallel grammatical structures: Par-
allel grammatical structures indicate the
similarity among items in list and help
the reader understand content.
Opening fi les
Saving a document
How to delete fi les
Opening a fi le
Saving a fi le
Deleting a fi le
Use steps correctly: Novices often inter-
sperse action and the results of action
when describing a step-by-step process.
Steps are always actions.
1. Press the customer button.
2. The customer dialogue box will
appear.
3. Type the customer ID and press the
submit button and the customer record
will appear.
1. Press the customer button.
2. Type the customer ID in the customer
dialogue box when it appears.
3. Press the submit button to view the cus-
tomer record for this customer.
Use short paragraphs: Readers of documen-
tation usually quickly scan text to fi nd the
information they need, so the text in the
middle of long paragraphs is often over-
looked. Use separate paragraphs to help
readers fi nd information more quickly.
Source: Based upon material from T. T. Barker, Writing Software Documentation (Boston: Allyn & Bacon, 1998).
reference topics, procedure topics, or tutorial topics. Th e items for the index and search
engine require more care because they are developed from the major parts of the system
and the users’ business functions. Every time we write a topic, we must also list the terms
that will be used to fi nd the topic. Terms for the index and search engine can come from
four distinct sources.
Designing Tests 467
The first source for index terms is the set of the commands in the user interface,
such as open file, modify customer, and print open orders. All commands contain two
parts (action and object). It is important to develop the index for both parts because
users could search for information using either part. A user looking for more information
about saving files, for example, might search by using the term save or the term files.
Th e second source is the set of major concepts in the system, which are oft en use cases
and classes. In the case of the Appointment system, for example, this might include appoint-
ment, symptoms, or patient.
A third source is the set of business tasks the user performs, such as ordering replacement
units or making an appointment. Oft en these are contained in the command set, but some-
times they require several commands and use terms that do not always appear in the system.
Good sources for these terms are the use scenarios and real use cases developed during inter-
face design (see Chapter 10).
A fourth, oft en controversial, source is the set of synonyms for the three sets of
preceding items. Users sometimes don’t think in terms of the nicely defi ned terms used
by the system. Th ey might try to fi nd information on how to stop or quit rather than exit,
or erase rather than delete. Including synonyms in the index increases the complexity and
size of the documentation system but can greatly improve the usefulness of the system to
the users.
DESIGNING TESTS
In object-oriented systems, the temptation is to minimize testing. Aft er all, through the use
of patterns, frameworks, class libraries, and components, much of the system has been tested
previously. Th erefore, we should not have to test as much. Right? Wrong! Testing is more
critical to object-oriented systems than to systems developed in the past. Based on encapsu-
lation (and information hiding), polymorphism (and dynamic binding), inheritance, reuse,
and the actual object-oriented products, thorough testing is much more diffi cult and critical.
Given the complexity of the development processes used and the global nature of information
systems development, testing becomes even more crucial. Th us, object-oriented testing must
be done systematically, and the results must be documented so that the project team knows
what has and has not been tested. Testing object-oriented systems is therefore very complex.
Consequently, a complete coverage of the topic is beyond the scope of this book.12
Th e purpose of testing is not to demonstrate that the system is free of errors. It is not
possible to prove that a system is error free. Th e purpose of testing is to uncover diff erences
between what the system actually does and what the system should do. In other words, the
purpose of testing is to try and break the system. Th is is similar to theory testing. You cannot
prove a theory. If a test fails to fi nd problems with a theory, your confi dence in the theory is
increased. However, if a test succeeds in fi nding a problem, then the theory has been falsifi ed.
Soft ware testing is similar in that it can only show the existence of errors. So, the point of
testing is to uncover as many errors as feasible. It is simply not cost-eff ective to try to get every
error out of the soft ware. Except in simple examples, it is, in fact, impossible. Th ere are simply
too many combinations to check.
12 For a good introduction to testing object-oriented soft ware, see John D. McGregor and David A. Sykes, A Prac-
tical Guide to Testing Object-Oriented Soft ware (Boston: Addison-Wesley, 2001). For a thorough coverage of testing
object-oriented soft ware, see Robert V. Binder, Testing Object-Oriented Systems: Models, Patterns, and Tools (Reading,
MA: Addison-Wesley, 1999), this book provides more than 1,000 pages of information with regard to how to test the
diff erent artifacts and processes included in object-oriented systems development.
468 Chapter 12 Construction
Th ere are four general stages of tests: unit tests, integration tests, system tests, and accept-
ance tests. Although each application system is diff erent, most errors are found during integra-
tion and system testing. In addition to the diff erent stages of tests, the tests must address both
the functional and nonfunctional requirements. However, before going into the specifi c types
of tests, we describe the eff ect that the object-oriented characteristics have on testing and the
necessary planning and management activities that must take place to have a successful
testing program.
Testing and Object Orientation
Most testing techniques have been developed to support non–object-oriented development.
Th erefore, most of the testing approaches have had to be adapted to object-oriented systems.
Th e characteristics of object-oriented systems that aff ect testing the most are encapsulation
(and information hiding); polymorphism (and dynamic binding); inheritance; and the use
of patterns, class libraries, frameworks, and components. Also, the sheer volume of products
that come out of a typical object-oriented development process has increased the importance
of testing in object-oriented systems development.
Encapsulation and Information Hiding Encapsulation and information hiding allow pro-
cesses and data to be combined to create holistic entities (i.e., objects). Th ey support hiding
everything behind a visible interface. Although this allows the system to be modifi ed and
maintained in an eff ective and effi cient manner, it makes testing the system problematic.
What do you need to test to build confi dence in the system’s ability to meet the user’s need?
You need to test the business process that is represented in the use cases. However, the busi-
ness process is distributed over a set of collaborating classes and contained in the methods
of those classes. Th e only way to know the eff ect that a business process has on a system is
to look at the state changes that take place in the system. But in object-oriented systems, the
instances of the classes hide the data behind a class boundary. How is it possible then to see
the impact of a business process?
A second issue raised by encapsulation and information hiding is the defi nition of a “unit”
for unit testing. What is the unit to be tested? Is it the package, class, or method? In traditional
approaches, the answer would be the process that is contained in a function. However, the pro-
cess in object-oriented systems is distributed over a set of classes. Th erefore, testing individual
methods makes no sense. Th e answer is the class. Th is dramatically changes the way unit testing
is done.
A third issue raised is the impact on integration testing. In this case, objects can be aggre-
gated to form aggregate objects; for instance, a car has many parts, or they can be grouped
together to form collaborations. Furthermore, they can be used in class libraries, frameworks,
and components. Based on all of these diff erent ways classes can be grouped together, how
does one eff ectively do integration testing?
Polymorphism and Dynamic Binding Polymorphism and dynamic binding dramati-
cally aff ect both unit and integration testing. Because an individual business process is
implemented through a set of methods distributed over a set of objects, as shown before,
the unit test makes no sense at the method level. However, with polymorphism and
dynamic binding, the same method (a small part of the overall business process) can be
implemented in many diff erent objects. Th erefore, testing individual implementations
of methods makes no sense. Again, the unit that makes sense to test is the class. Except
for trivial cases, dynamic binding makes it impossible to know which implementation is
going to be executed until the system does it. Th erefore, integration testing becomes very
challenging.
Designing Tests 469
Inheritance When taking into consideration the issues raised about inheritance (see Chapter 8),
it should not be a surprise that inheritance aff ects the testing of object-oriented systems. Th rough
the use of inheritance, bugs can be propagated instantaneously from a superclass to all its direct
and indirect subclasses. However, the tests that are applicable to a superclass are also applicable
to all its subclasses. As usual, inheritance is a double-edged sword. Finally, even though we have
stated this many times before, inheritance should support only a generalization and specializa-
tion type of semantics. Remember, when using inheritance, the principle of substitutability is
critical (see Chapter 5). All these issues aff ect unit and integration testing.
Reuse On the surface, reuse should decrease the amount of testing required. However, each time
a class is used in a diff erent context, the class must be tested again. Th erefore, any time a class
library, framework, or component is used, unit testing and integration testing are important. In
the case of a component, the unit to test is the component itself. Remember that a component has
a well-defi ned API (application program interface) that hides the details of its implementation.
Object-Oriented Development Process and Products In virtually all textbooks, including
this one, testing is covered near the end of system development. Th is seems to imply that test-
ing is something that takes place only aft er the programming has ended. However, every prod-
uct13 that comes out of the object-oriented development process must be tested. For example,
it is a lot easier to ensure that the requirements are captured and modeled correctly through
testing the use cases, and it is a lot cheaper to catch this type of error back in analysis than it is
in implementation. Obviously, this is also true for testing collaborations. By the time we have
implemented a collaboration as a set of layers and partitions, we could have expended a great
deal of time—and time is money—on implementing the wrong thing. So testing collaborations
by role-playing the CRC cards in analysis actually saves the team lots of time and money.
Testing is something that must take place throughout system development, not simply
at the end. However, the type of testing that can take place on nonexecutable representations,
such as use cases and CRC cards, is diff erent from those on code written in an object-oriented
programming language. Th e primary approach to testing nonexecutable representations is some
form of an inspection or walkthrough of the representation.14 In the earlier chapters, we focused
on verifying and validating the diff erent analysis and design representations. We also made sure
that the diff erent representations were consistent and balanced. As such, we have dealt with test-
ing the nonexecutable representations throughout the development process (see Chapters 4–11).
Test Planning
Testing starts with the development of a test plan, which defi nes a series of tests that will be
conducted. Because testing takes place throughout the development of an object-oriented
system, a test plan should be developed at the very beginning of system development and
continuously updated as the system evolves. For example, the representation of a class
evolves from a simplistic CRC card to a set of classes that are implemented in a program-
ming language. In Figure 12-3 we see a CRC card representation of an Order class that
13 For example, activity diagrams, use-case descriptions, use-case diagrams, CRC cards, class diagrams, object
diagrams, sequence diagrams, communication diagrams, behavioral state machines, package diagrams, contracts,
method specifi cations, use scenarios, window navigation diagrams, storyboards, windows layout diagrams, real use
cases, and source code.
14 See Michael Fagan, “Design and Code Inspections to Reduce Errors in Program Development,” IBM Systems
Journal 15, no. 3 (1976); Daniel P. Freedman and Gerald M. Weinberg, Handbook of Walkthrough, Inspections, and
Technical Reviews: Evaluating Programs, Projects, and Products, 3rd Ed. (New York: Dorset House Publishing, 1990).
Also, Chapters 4, 5, 6, and 7 describe the walkthrough process in detail in relation to the verifi cation and validation
of the analysis models.
470 Chapter 12 Construction
Front:
Class Name: Order ID: 2
Calculate tax
Calculate subtotal
Calculate shipping
Calculate total
Responsibilities
Associated Use Cases: 3Description: An Individual who needs to receive or has received
medical attention
Type: Concrete, Domain
Collaborators
(a)
Back:
Attributes:
Relationships:
Generalization (a-kind-of):
Aggregation (has-parts):
Other Associations: Customer {1..1} State {1..1} Product {1..*}
(b)
Order Number (1..1) (unsigned long)
Date (1..1) (Date)
Sub Total (0..1) (double) {Sub Total = ProductOrder.sum(GetExtension())}
Tax (0..1) (double) (Tax = State.GetTaxRate() * Sub Total)
Shipping (0..1) (double)
Total (0..1) (double)
Customer (1..1) (Customer)
Cust ID (1..1) (unsigned long) {Cust ID = Customer. GetCustID()}
State (1..1) (State)
StateName (1..1) (String) {State Name = State. GetState()}
FIGURE 12-3
Order CRC Card
(see Figure 8-19)
Designing Tests 471
contains invariants. Each of these invariants must be tested and enforced for the Order class
to be considered to be of suffi cient quality. One simple invariant test would be to attempt to
assign a value to the Cust ID attribute that was not associated with the Customer object that
is contained in the Customer attribute. Another invariant test would be to try and assign
more than one date to the Date attribute. Finally, a trickier invariant test would be to try
to assign an integer value to the Shipping attribute. Th is one is more diffi cult because most
programming languages allow an integer to be “cast” to a double. If the value contained in
the Shipping attribute really is supposed to be a double, then casting the integer value to a
double would be an error. Th ese tests should be done using a walkthrough approach when
the class is specifi ed, as we did in Chapters 4, 5, 6, and 7, and a more rigorous approach
once the class has been fully implemented. Th is is an example of unit testing a class, which is
described later in this chapter. To ensure the quality of a class, it should be tested each time
its representation is changed.
Th e test plan should address all products that are created during the development of the
system. For example, tests should be created that can be used to test completeness of a CRC
card. Each individual test has a specifi c objective and describes a set of very specifi c test cases
to examine. In the case of invariant-based tests, a description of the invariant is given, and the
original values of the attribute, the event that will cause the attribute value to change, the actual
results observed, the expected results, and whether it passed or failed are shown. Test specifi ca-
tions are created for each type of constraint that must be met by the class. Also, similar types of
specifi cations are done for integration, system, and acceptance tests.
Not all classes are likely to be fi nished at the same time, so the programmer usually
writes stubs for the unfi nished classes to enable the classes around them to be tested. A stub
is a placeholder for a class that usually displays a simple test message on the screen or returns
some hardcoded value15 when it is selected. For example, consider an application system that
provides creating, changing, deleting, fi nding, and printing functions for some object such as
CDs, patients, or employees. Depending on the fi nal design, these diff erent functions could
end up in diff erent objects on diff erent layers. Th erefore, to test the functionality associated
with the classes on the problem-domain layer, a stub would be written for each of the classes
on the other layers that interact with the problem-domain classes. Th ese stubs would be the
minimal interface necessary to be able to test the problem-domain classes. For example, they
would have methods that could receive the messages being sent by the problem-domain
layer objects and methods that could send messages to the problem-domain layer objects.
Typically, the methods would display a message on the screen notifying the tester that the
method was successfully reached (e.g., Delete item from Database method reached). In this
way, the problem-domain classes could pass class testing before the classes on the other layers
were completed.
Finally, as you may suspect, test planning should be performed throughout the develop-
ment process. It is a lot easier to design tests when you are creating the diff erent analysis and
design representations than to wait and design them during the construction of the system.
Unit Tests
Unit tests focus on a single unit—the class. Th ere are two approaches to unit testing: black-
box testing and white-box testing (see Figure 12-4). Black-box testing is the most commonly
used because each class represents an encapsulated object. Black-box testing is driven by
15 Hardcoded means written into the program. For example, suppose you were writing a unit to calculate the net
present value of a loan. Th e stub might be written to always display (or return to the calling module) a value of 100
regardless of the input values.
472 Chapter 12 Construction
FIGURE 12-4 Types of Tests
Unit Testing Black-Box Testing CRC Cards For normal unit testing • Tester focuses on whether the class
Treats class as a Class Diagrams meets the requirements stated in the
black box Contracts specifi cations.
White-Box Testing Method Specifi cations When complexity is • By looking inside the class to review the
Looks inside the high code itself, the tester may discover errors
class to test its or assumptions not immediately obvious
major elements to someone treating the class as a black
box.
Integration User Interface Testing Interface Design For normal integration • Testing is done by moving through each
Testing The tester tests each testing and every menu item in the interface
interface function either in a top-down or bottom-up manner.
Use-Case Testing Use Cases When the user • Testing is done by moving through each
The tester tests each interface is important use case to ensure that they work correctly.
use case • Usually combined with user interface
testing because it does not test all
interfaces.
Interaction Testing Class Diagrams When the system • The entire system begins as a set of
Tests each process in Sequence Diagrams performs data stubs. Each class is added in turn and
step-by-step Communication processing the results of the class compared to the
fashion Diagrams correct result from the test data; when a
class passes, the next class is added
and the test rerun. This is done for each
package. Once each package has passed
all tests, then the process repeats
integrating the packages.
System Interface Testing Use-Case Diagram When the system • Because data transfers between systems
Tests the exchange of exchanges data are often automated and not monitored
data with other directly by the users, it is critical to
systems design tests to ensure that they are being
done correctly.
System Requirements Testing System Design, Unit For normal system • Ensures that changes made as a result
Testing Tests to whether Tests, and Integration testing of integration testing did not create new
original business Tests errors.
requirements • Testers often pretend to be uninformed
are met users and perform improper actions to
ensure that the system is immune to invalid
actions (e.g., adding blank records).
Usability Testing Interface Design and When user interface • Often done by analyst with experience
Tests how convenient Use Cases is important in how users think and in good interface
the system is to use design.
• Sometimes uses formal usability testing
procedures discussed in Chapter 10.
Documentation Testing Help System, For normal system • Analysts spot check or check every item
Tests the accuracy of Procedures, Tutorials testing on every page in all documentation to
the documentation ensure that the documentation items
and examples work properly.
Performance Testing System Proposal When the system is • High volumes of transactions are
Examines the ability to important generated and given to the system.
perform under high Infrastructure Design • Often done by using special-purpose
loads testing software.
Security Testing Infrastructure Design When the system is • Security testing is a complex task, usually
Tests disaster recovery important done by an infrastructure analyst assigned
and unauthorized to the project.
access • In extreme cases, a professional fi rm may
be hired.
Acceptance Alpha Testing System Tests For normal • Often repeats previous tests but are con-
Testing Conducted by users to acceptance testing ducted by users themselves to ensure that
ensure that they they accept the system.
accept the system
Beta Testing System Requirements When the system is • Users closely monitor system for errors or
Uses real data, not important useful improvements.
test data
Stage Types of Tests Test Plan Source When to Use Notes
Designing Tests 473
the CRC cards, behavior state machines, and contracts associated with a class, not by the
programmers’ interpretation. In this case, the test plan is developed directly from the spec-
ifi cation of the class: each item in the specifi cation becomes a test, and several test cases
are developed for it. White-box testing is based on the method specifi cations associated
with each class. However, white-box testing has had limited impact in object-oriented
development. Th is is due to the rather small size of the individual methods in a class. Most
approaches to testing classes use black-box testing to ensure their correctness.
Class tests should be based on the invariants on the CRC cards, the behavioral state
machines associated with each class, and the pre- and post-conditions contained on each
method’s contract. Assuming all the constraints have been captured on the CRC cards and
contracts, individual test cases can be developed fairly easily. For example, suppose the
CRC card for an order class gave an invariant that the order quantity must be between 10
and 100 cases. Th e tester would develop a series of test cases to ensure that the quantity is
validated before the system accepts it. It is impossible to test every possible combination of
input and situation; there are simply too many possible combinations. In this example,
the test requires a minimum of three test cases: one with a valid value (e.g., 15), one with
a low invalid value (e.g., 7), and one with a high invalid value (e.g., 110). Most tests would
also include a test case with a nonnumeric value to ensure the data types were checked (e.g.,
ABCD). A really good test would include a test case with nonsensical but potentially valid
data (e.g. 21.4).
Using a behavioral state machine is a useful way to identify tests for a class. Any class
that has a behavioral state machine associated with it has a potentially complex life cycle.
It is possible to create a series of tests to guarantee that each state can be reached. For
example, Figure 12-5 portrays the behavioral state machine for the Order class just dis-
cussed. In this case, there are many transitions between the diff erent states of an instance
of the Order class. Tests should be created to guarantee that the only transitions allowed
from an instance of the Order class are the ones specifi cally defi ned. In this case, it should
be impossible for an Order object to go from the In process state to the Placed state
without traversing the Ordered and Processing states via the Customer submits order, Order
sent for credit authorization, and Authorization = Approved transitions. Th is state-based test-
ing can be done throughout the development of the class via walkthroughs and role-playing
early in the evolution of the class and more rigorous testing once it has been implemented in
a programming language.
Tests also can be developed for each contract associated with the class. In the case
of a contract, a set of tests for each pre- and post-condition is required. For example, the
contract of the addOrder method of the Customer class shown in Figure 12-6 has both a
pre- and post-condition that essentially requires the new order to have not existed with
the instance of the Customer class before the method executes, and that the new order is
associated with the Customer object aft er the method executes. Tests must be created to
enforce these constraints. If the class is a subclass of another class, then all the tests asso-
ciated with the superclass must be executed again. Th e interactions among the constraints,
invariants, and the pre- and post-conditions in the subclass and the superclass(es) must be
also addressed.
Finally, owing to good object-oriented design, to fully test a class, special testing methods
might have to be added to the class being tested. For example, how can invariants be tested?
Th e only way to really test them is to have methods that are visible to the outside of the class
that can be used to manipulate the values of the class’s attributes. However, adding these types
of methods to a class does two things. First, they add to the testing requirements because they
themselves have to be tested. Second, if they are not removed from the deployed version of the
system, the system will be less effi cient, and the advantage of information hiding eff ectively
474 Chapter 12 Construction
O
rd
er
In
p
ro
ce
ss
O
rd
er
ed
D
en
ie
d
O
rd
er
is
cr
ea
te
d
C
us
to
m
er
su
bm
its
o
rd
er
C
us
to
m
er
ed
its
o
rd
er
in
fo
rm
at
io
n
A
ut
ho
ri
za
tio
n
=
D
en
ie
d
Pr
oc
es
si
ng
O
rd
er
s
en
t
fo
r
cr
ed
it
au
th
or
iz
at
io
n
C
us
to
m
er
w
ith
dr
aw
s
or
de
r
re
qu
es
t
Pl
ac
ed
O
rd
er
s
en
t t
o
cu
st
om
er
C
us
to
m
er
ac
ce
pt
s
sh
ip
m
en
t
Sh
ip
pe
d
R
ec
ei
ve
d
A
ut
ho
ri
za
tio
n
=
A
pp
ro
ve
d
O
rd
er
is
cl
os
ed
FI
G
U
R
E
1
2
-5
O
rd
er
B
eh
av
io
ra
l S
ta
te
M
ac
h
in
e
(s
ee
F
ig
u
re
6
-1
8)
Designing Tests 475
16 We describe some of the diff erent types of user interface testing in Chapter 10.
Method Name: Class Name: ID:
Associated Use Cases:
Clients (Consumers):
Type of Value Returned:
Description of Responsibilities:
Arguments Received:
Pre-Conditions:
Post-Conditions:
addOrder Customer 36
addCustomerOrder
anOrder:Order
void
not Orders.includes(anOrder)
Orders = Orders@pre.including(anOrder)
Implement the necessary behavior to add a new order to an existing customer keeping
the orders in sorted order by the order’s order number.
FIGURE 12-6
addOrder Contract
(see Figure 8-25)
is lost. As is readily apparent, testing classes is complex. Th erefore, great care must be taken
when designing tests for classes.
Integration Tests
Integration tests assess whether a set of classes that must work together do so without error. Th ey
ensure that the interfaces and linkages between diff erent parts of the system work properly. At this
point, the classes have passed their individual unit tests, so the focus now is on the fl ow of control
among the classes and on the data exchanged among them. Integration testing follows the same
general procedures as unit testing: Th e tester develops a test plan that has a series of tests, which, in
turn, have a test. Integration testing is oft en done by a set of programmers and/or systems analysts.
From an object-oriented systems perspective, integration testing can be diffi cult. A single class
can be in many diff erent aggregations, because of the way objects can be combined to form new
objects, class libraries, frameworks, components, and packages. Where is the best place to start the
integration? Typically, the answer is to begin with the set of classes, a collaboration, that are used
to support the highest-priority use case (see Chapter 4). Also, dynamic binding makes it crucial to
design the integration tests carefully to ensure that the combinations of methods are tested.
Th ere are four approaches to integration testing: user interface testing,16 use-case testing,
interaction testing, and system interface testing (see Figure 12-4). Most projects use all four
approaches. However, like unit testing, integration testing must be carefully planned. In the
case of use-case testing, only the aspects of the class and class invariants related to the spe-
cifi c use case are included in these use-case context-dependent class tests. In fact, typically
use-case testing is performed one scenario at a time. In many ways, use-case testing can be
viewed as a more rigorous role-playing exercise (see Chapter 5). Like unit testing, integration
476 Chapter 12 Construction
testing should be performed throughout the evolution of the system. In the early stages of the
system’s development, you should be working with the CRC cards and role-playing them.
Later on, you will have the contracts and method specifi cations completed. Gradually, you
will have implemented the problem domain classes, the user interface classes, and the data
management layer classes in a programming language. As in unit testing, each time a new
representation (diagram, text, program) is created, a new integration test needs to be per-
formed. Th erefore, as the system evolves to more completely support the use case, we can
more rigorously test whether the use case is fully supported or not.
One of the major problems with integration testing and object-oriented systems is the
diffi culty caused by the interaction of inheritance and dynamic binding. Th is specifi c prob-
lem has become known as the yo-yo problem. Th e yo-yo problem occurs when the analyst or
designer must bounce up and down through the inheritance graph to understand the control
fl ow through the methods being executed. In most cases, this is caused by a rather deep inher-
itance graph; that is, the subclass has many superclasses above it in the inheritance graph. Th e
yo-yo problem becomes even more of a nightmare in testing object-oriented systems when
inheritance confl icts exist and when multiple inheritance is used (see Chapter 8). About the
only realistic approach to testing through the yo-yo problem is through an interactive debug-
ger that is typically part of a systems development environment, such as Eclipse, Netbeans,
or Visual Studio.
System Tests
To ensure that all classes work together without error, systems analysts usually conduct the
system tests. System testing is similar to integration testing but is much broader in scope.
Whereas integration testing focuses on whether the classes work together without error,
system tests examine how well the system meets both the functional and nonfunctional
requirements, e.g., usability, documentation, performance, and security (see Figure 12-4).
Th e purpose of functional requirements testing is to ensure that the functional require-
ments uncovered are indeed met. Like integration testing, this is primarily driven by the
system’s use cases and their scenarios. However, in many cases, integration testing requires
modifi cations to the system. So, the focus of requirements testing is to ensure that the modi-
fi cations made did not cause additional errors.
Usability testing is essentially a combination of the user interface and use-case testing
that takes place during integration testing. Where user interface and use-case testing focused
on whether the user interface works and whether the use case was supported, respectively,
usability testing focuses on how well the user interface supports the use cases. Th at is, how
effi cient and eff ective the user interface is. In many cases, this could include formal usability
testing (see Chapter 10).
Given that documentation is basically a system in itself, documentation testing should
involve both unit and integration testing. In this case, the unit is a documentation entry, and
the user interface is either the paper or help screen. From an integration testing perspective,
the focus is on whether the documentation works or not. And, like system testing of the soft –
ware, the focus of system testing of the documentation is how well the documentation works.
Th e reason that documentation is not typically tested in parallel with the system, i.e., when the
classes, use cases, and user interface are tested, is to minimize the amount of documentation
testing required. Even though the documentation should be developed in parallel with the
soft ware, until the soft ware is tested, it is unclear exactly what to test in the documentation.
As the soft ware “passes” its tests, the documentation that goes along with the soft ware can
then be fi nalized and tested.
Designing Tests 477
Performance testing focuses on trying to break the system with regard to the amount of
work the system can handle. Th ese types of tests typically fall into two categories: stress tests
and volume tests. Th e purpose of stress tests, also known as load tests, is to ensure that the
system can handle a certain number of simultaneous requests. For example, if the system is
supposed to be able to handle 10,000 simultaneous requests, a stress test would attempt to
push the system into handling more than that. If the performance of the test is insuffi cient,
various soft ware and database optimizations (see Chapters 8 and 9) can be investigated. In
other cases, additional hardware could be required. Th e purpose of volume tests is to push
the implementation so that it may break when there is a large amount of data required to
answer a user request. Again, if it is discovered that the system fails this type of test, then
database and soft ware optimizations and additional hardware could be required. For exam-
ple, sometimes it is more effi cient to create a set of temporary tables by “selecting” the data
from the actual tables before “joining” the tables together. By performing the “selects” fi rst,
the “join” works on less data. In this case, it could both speed up and lessen the amount
of temporary storage required to handle the request. In other cases, denormalization of
the data and storing the data at multiple locations could be called for. You typically do
not want the user to make a request for a report and have to wait “too long” for the report
to be processed. In some cases, giving up some functionality to improve performance can
be crucial to the success of the system. So, the results of performance testing can make or
break a system.
Obviously, in today’s networked world, security testing is crucial. Security testing
involves three primary areas: authentication, authorization, and virus control. Authentication
testing deals with ensuring that the logged in user is who he or she claims to be. Typically,
this has been addressed with user IDs and passwords and through the use of encryption
techniques (see Chapter 11). Today, in addition to these approaches, various biometric
identifi ers have been used, e.g., retinal scans and fi ngerprints. Authorization testing deals
with ensuring that the logged in user actually has the authority to use the system(s) being
accessed. Authorization has been controlled through the use of roles, access control lists,
and capability lists. Security roles are the same as actor roles in a use-case model. Depending
on the role being played by a user, diff erent capabilities are made available to the user in the
form of a capability list. However, in this case, a role can be specifi ed down to the individual
user level and not be limited to a group of users. Also, an access control list can be associated
with each use case and with each class. In this case, an access control list specifi es which
roles have access to the resource (use case or class). Given that many system break-ins are a
function of viruses, virus controls also need to be enforced. Anytime a fi le is received or sent
by a user, the fi les should be scanned for potential viruses. Th is includes e-mail attachments,
Web downloads, and the insertion of fl ash drives on desktop computers as well on all forms
of “client” machines that can be attached to the system. Obviously, security requirements
will impact the performance of the system. Th erefore, trade-off s between these two sets of
requirements may be necessary.
Acceptance Tests
Acceptance testing is done primarily by the users with support from the project team. Th e
goal is to confi rm that the system is complete, meets the business needs that prompted the
system to be developed, and is acceptable to the users. Acceptance testing is done in two
stages: alpha testing, in which users test the system using made-up data, and beta testing,
in which users begin to use the system with real data but are carefully monitored for errors
(see Figure 12-4).
478 Chapter 12 Construction
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe the basic issues related to managing programmers.
Describe cultural issues as they are related to intellectual property.
Describe how Hall’s and Hofstede’s cultural dimensions can aff ect systems development.
Describe the diff erent types of documentation associated with an information system.
Describe how to develop the documentation of an information system.
Describe how object-orientation eff ects soft ware testing.
Describe and discuss unit testing.
Describe and discuss integration testing.
Describe and discuss system testing.
Describe and discuss acceptance testing.
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
While many of the activities, including testing and documentation, were done through-
out the development process, the construction phase for phase one of the Integrated
Health Clinic Delivery System was a busy time for the team who conducted integration,
system, and acceptance testing for the system. In addition, they fi nalized user and system
documentation of the system.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
KEY TERMS
Acceptance test
Access control list
Alpha test
Beta test
Black-box testing
Capability list
Change control
Collectivism
Construction
Context
Documentation navigation
control
Documentation topic
Femininity
Hardcoded
Individualism
Integration test
Interaction testing
Load test
Long-term orientation
Masculinity
Monochronic time
Polychronic time
Power distance
Procedures manual
Program log
Reference document
Requirements testing
Role
Security testing
Short-term orientation
Speed of messages
Stress test
Stub
System documentation
System interface testing
System test
Test case
Test plan
Test specifi cation
Time
Timeboxing
Traceability
Tutorial
Uncertainty avoidance
Unit test
Usability testing
Use-case testing
User documentation
User interface testing
Volume test
White-box testing
Yo-yo problem
Exercises 479
QUESTIONS
1. Why is testing important?
2. How can diff erent national or organizational cultures
aff ect the management of an information systems
development project?
3. What is the primary role of systems analysts during
the programming stage?
4. In Th e Mythical Man-Month, Frederick Brooks argues
that adding more programmers to a late project makes
it later. Why?
5. When off shoring development, how could diff erences
in Hall’s context dimension of culture aff ect the con-
tribution of a team member to the successful devel-
opment of an information system? What about Hall’s
time or speed of messages dimensions?
6. What are Hofstede’s fi ve dimensions of cultural diff er-
ences? How could diff erences in them infl uence the eff ec-
tiveness of an information systems development team?
7. What are the common language or languages used
today in information systems development?
8. Compare and contrast user documentation and system
documentation.
9. Why is online documentation becoming more important?
10. What are the primary disadvantages of online docu-
mentation?
11. Compare and contrast reference documents, proce-
dures manuals, and tutorials.
12. What are fi ve types of documentation navigation
controls?
13. What are the commonly used sources of documentation
topics? Which is the most important? Why?
14. What are the commonly used sources of documenta-
tion navigation controls? Which is the most impor-
tant? Why?
15. What is the purpose of testing?
16. Describe how object orientation aff ects testing.
17. Compare and contrast the terms test, test plan, and
test case.
18. What is a stub and why is it used in testing?
19. What is the primary goal of unit testing?
20. How are the test cases developed for unit tests?
21. Compare and contrast black-box testing and white-
box testing.
22. What are the diff erent types of class tests?
23. What is the primary goal of integration testing?
24. How are the test cases developed for integration tests?
25. Describe the yo-yo problem. Why does it make inte-
gration testing diffi cult?
26. What is the primary goal of system testing?
27. How are the test cases developed for system tests?
28. What is the primary goal of acceptance testing?
29. How are the test cases developed for acceptance tests?
30. Compare and contrast alpha testing and beta testing.
EXERCISES
A. Diff erent views of plagiarism and collaborative learn-
ing were described as examples of diff erences among
diff erent cultures today. Using the Web, identify other
diff erences that could aff ect the success of an informa-
tion systems development team.
B. Besides Hall and Hofstede, both David Victor and Fons
Trompenaars have identifi ed a set of cultural dimen-
sions that could be useful in information systems devel-
opment. Using the Web, identify their dimensions.
C. If the registration system at your university does not
have a good online help system, develop one for one
screen of the user interface.
D. Examine and prepare a report on the online help
system for the calculator program in Windows (or a
similar one on the Mac or Unix). (You will probably
be surprised at the amount of help for such a simple
program.)
E. Compare and contrast the online help at two diff erent
websites that enable you to perform some function
(e.g., make travel reservations, order books).
F. Create an invariant test specifi cation for the class you
chose for the A Real Estate Inc. problem in exercise A
in Chapter 8.
G. Create a use-case test plan, including the specifi c class
plans and invariant tests, for a use case from the A
Real Estate Inc. exercises in the previous chapters.
H. Create an invariant test specifi cation for the class you
chose for the A Video Store problem in exercise B in
Chapter 8.
I. Create a use-case test plan, including the specifi c class
plans and invariant tests, for a use case from the A
Video Store exercises in the previous chapters.
J. Create an invariant test specifi cation for the class you
chose for the gym problem in exercise C in Chapter 8.
480 Chapter 12 Construction
K. Create a use-case test plan, including the specifi c class
plans and invariant tests for a use case from the health
club exercises in previous chapters.
L. Create an invariant test specifi cation for the class you
chose for Picnics R Us in exercise D in Chapter 8.
M. Create a use-case test plan, including the specifi c
class plans and invariant tests, for a use case from the
Picnics R Us exercises in the previous chapters.
N. Create an invariant test specifi cation for the class you
chose for the Of-the-Month Club (OTMC) in exercise
E in Chapter 8.
O. Create a use-case test plan, including the specifi c
class plans and invariant tests, for a use case from the
Of-the-Month Club (OTMC) exercises in the previ-
ous chapters.
MINICASES
1. Pete is a project manager on a new systems develop-
ment project. Th is project is Pete’s fi rst experience as a
project manager, and he has led his team successfully
to the programming phase of the project. Th e project
has not always gone smoothly, and Pete has made a
few mistakes, but he is generally pleased with the pro-
gress of his team and the quality of the system being
developed. Now that programming has begun, Pete
has been hoping for a little break in the hectic pace of
his workday.
Prior to beginning programming, Pete recognized
that the time estimates made earlier in the project
were too optimistic. However, he was fi rmly com-
mitted to meeting the project deadline because of his
desire for his fi rst project as project manager to be a
success. In anticipation of this time pressure problem,
Pete arranged with the Human Resources department
to bring in two new college graduates and two college
interns to beef up the programming staff . Pete would
have liked to fi nd some staff with more experience,
but the budget was too tight, and he was committed
to keeping the project budget under control.
Pete made his programming assignments, and
work on the programs began about two weeks ago.
Now, Pete has started to hear some rumbles from the
programming team leaders that might signal trouble.
It seems that the programmers have reported several
instances where they wrote programs, only to be
unable to fi nd them when they went to test them.
Also, several programmers have opened programs
that they had written, only to fi nd that someone had
changed portions of their programs without their
knowledge.
a. Is the programming phase of a project a time for
the project manager to relax? Why or why not?
b. What problems can you identify in this situation?
c. What advice do you have for the project manager?
How likely does it seem that Pete will achieve his
desired goals of being on time and within budget
if nothing is done?
2. Th e systems analysts are developing the test plan for
the user interface for the Holiday Travel Vehicles
system. As the salespeople are entering a sales invoice
into the system, they will be able to enter an option
code into a text box or to select an option code from
a drop-down list. A combo box was used to imple-
ment this, because it was felt that the salespeople
would quickly become familiar with the most com-
mon option codes and would prefer entering them
directly to speed up the entry process.
It is now time to develop the test for validating
the option code fi eld during data entry. If the cus-
tomer did not request any dealer-installed options
for the vehicle, the salesperson should enter “none”;
the fi eld should not be blank. Th e valid option codes
are four-character alphabetic codes and should be
matched against a list of valid codes.
Prepare a test plan for the test of the option code
fi eld during data entry.
481
C H A P T E R 1 3
Installation and Operations
This chapter examines the activities needed to install an information system and suc-
cessfully convert an organization to using it. It also discusses post-implementation activities,
such as system support, system maintenance, and project assessment. Installing the system
and making it available for use from a technical perspective are relatively straightforward.
However, the training and organizational issues surrounding the installation are more com-
plex and challenging because they focus on people, not computers.
OBJECTIVES
■ Be familiar with the system installation process.
■ Understand diff erent types of conversion strategies and when to use them.
■ Understand several techniques for managing change.
■ Be familiar with post-installation processes.
INTRODUCTION
“It must be remembered that there is nothing more diffi cult to plan, more doubtful of success,
nor more dangerous to manage than the creation of a new system. For the initiator has the ani-
mosity of all who would profi t by the preservation of the old institution and merely lukewarm
defenders in those who would gain by the new.”
—Niccolò Machiavelli, Th e Prince, 1513
Although written almost 500 years ago, Machiavelli’s comments are still true today. Managing
the change to a new system—whether it is computerized or not—is one of the most diffi cult
tasks in any organization. Because of the challenges involved, most organizations begin
developing their conversion and change management plans while the programmers are still
developing the soft ware. Leaving conversion and change management planning to the last
minute is a recipe for failure.
In many ways, using a computer system or set of work processes is much like driving on a
dirt road. Over time, with repeated use, the road begins to develop ruts in the most commonly
used parts of the road. Although these ruts show where to drive, they make change diffi cult.
As people use a computer system or set of work processes, those systems or work processes
begin to become habits or norms; people learn them and become comfortable with them.
Th ese systems or work processes then begin to limit people’s activities and make it diffi cult
for them to change because they begin to see their jobs in terms of these processes rather than
of the fi nal business goal of serving customers.
482 Chapter 13 Installation and Operations
One of the earliest models for managing organizational change was developed by Kurt
Lewin.1 Lewin argued that change is a three-step process: unfreeze, move, refreeze (Figure
13-1). First, the project team must unfreeze the existing habits and norms (the as-is system) so
that change is possible. Most of system development to this point has laid the groundwork for
unfreezing. Users are aware of the new system being developed, some have participated in an
analysis of the current system (and so are aware of its problems), and some have helped design
the new system (and so have some sense of the potential benefi ts of the new system). Th ese activ-
ities have helped to unfreeze the current habits and norms.
Th e second step is to help the organization move to the new system via a migration plan.
Th e migration plan has two major elements. One is technical, which includes how the new
system will be installed and how data in the as-is system will be moved into the to-be system;
this is discussed in the conversion section of this chapter. Th e second component is organi-
zational, which includes helping users understand the change and motivating them to adopt
it; this is discussed in the change management section of this chapter.
Th e third step is to refreeze the new system as the habitual way of performing the
work processes—ensuring that the new system successfully becomes the standard way of
performing the business function it supports. Th is refreezing process is a key goal of the
post-implementation activities discussed in the fi nal section of this chapter. By providing
ongoing support for the new system and immediately beginning to identify improvements
for the next version of the system, the organization helps solidify the new system as the
new habitual way of doing business. Post-implementation activities include system support,
which means providing help desk and telephone support for users with problems; system
maintenance, which means fi xing bugs and improving the system aft er it has been installed;
and project assessment, evaluating the project to identify what went well and what could be
improved for the next system development project.
Change management is the most challenging of the three components because it focuses
on people, not technology, and because it is the one aspect of the project that is the least con-
trollable by the project team. Change management means winning the hearts and minds of
potential users and convincing them that the new system actually provides value.
Maintenance is the most costly aspect of the installation process, because the cost of
maintaining systems usually greatly exceeds the initial development costs. It is not unusual
for organizations to spend 60 to 80 percent of their total IS development budget on mainte-
nance. Although this might sound surprising initially, think about the soft ware you use. How
many soft ware packages do you use that are the very fi rst version? Most commercial soft ware
1 Kurt Lewin, “Frontiers in Group Dynamics,” Human Relations 1, no. 5 (1947): 5–41; Kurt Lewin, “Group Decision
and Social Change,” in E. E. Maccoby, T. M. Newcomb, and E. L. Hartley (eds.), Readings in Social Psychology (New
York: Holt, Rinehart, & Winston, 1958), pp. 197–211.
FIGURE 13-1
Implementing
Change
As-Is
system
To-Be
system
Unfreeze
Analysis and
design
Refreeze
Support and
maintenance
Transition
Move
Migration plan:
■ Technical conversion
■ Change management
Cultural Issues and Information Technology Adoption 483
packages become truly useful and enter widespread use only in their second or third version.
Maintenance and continual improvement of soft ware is ongoing, whether it is a commercially
available package or soft ware developed in-house. Would you buy soft ware if you knew that no
new versions were going to be produced? Of course, commercial soft ware is somewhat diff erent
from custom in-house soft ware used by only one company, but the fundamental issues remain.
Project assessment is probably the least commonly performed part of system devel-
opment but is perhaps the one that has the most long-term value to the IS department.
Project assessment enables project team members to step back and consider what they did
right and what they could have done better. It is an important component in the individual
growth and development of each member of the team, because it encourages team members
to learn from their successes and failures. It also enables new ideas or new approaches to
system development to be recognized, examined, and shared with other project teams to
improve their performance.
CULTURAL ISSUES AND INFORMATION
TECHNOLOGY ADOPTION2
Cultural issues are one of the things that are typically identifi ed as at least partially to blame
when there is a failure in an organization. Cultural issues have been studied at both organi-
zational and national levels. In previous chapters, we discussed the eff ect that cultural issues
can have on designing the human–computer interaction and physical architecture layers
(see Chapters 10 and 11) and the management of programmers (Chapter 12). Th e cultural
dimensions identifi ed by Hall and Hofstede included speed of messages, context, time, power
distance, uncertainty avoidance, individualism versus collectivism, masculinity versus fem-
ininity, and long- versus short-term orientation.3 In this chapter, we describe how these
dimensions can aff ect the successful deployment of an information system that supports a
global information supply chain.
Hall’s fi rst dimension, speed of messages, has implications for the development of
documentation (see Chapter 12) and training approaches (see later in this chapter).
In a culture that values “deep” content, so that members of the culture can take their
time to thoroughly understand the new system, simply providing an online help system
is not going to be suffi cient to ensure the successful adoption of the new information
system. However, in a culture that prefers “fast” messages, an online help system could
be suffi cient.
Hall’s second dimension, context, also aff ects the adoption and deployment of a new sys-
tem. In high-context cultures, it is expected that the new information system will be placed
into the entire context of the enterprise-wide system. Members of this type of society expect
to be able to understand exactly where the system fi ts into the fi rm’s overall picture. Again,
like the speed of messages dimension, this aff ects the training approach used and the docu-
mentation developed.
Hall’s third dimension, time, can also eff ect the adoption and deployment of a new
system. In a polychronic time culture, the training could need to be spread out over a longer
2 A good summary of cultural issues and information systems is Dorothy E. Leidner and Timothy Kayworth, “A
Review of Culture in Information Systems Research: Toward a Th eory of Information Technology Culture Confl ict,”
MIS Quarterly 30, no. 2 (2006): 357–399.
3 See Geert Hofstede, Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations across
Nations, 2nd Ed. (Th ousand Oaks, CA: Sage, 2001); Geert Hofstede, Gert Jan Hofstede, and Michael Minkov, Cultures
and Organizations: Soft ware of the Mind, 3rd Ed. (New York: McGraw-Hill, 2010); Edward T. Hall, Beyond Culture
(New York: Anchor Books, 1981).
484 Chapter 13 Installation and Operations
period of time, when compared to a monochronic time culture. In a monochromic time
culture, interruptions would be considered rude. Consequently, training could be accom-
plished in a small set of intense sessions. However, with a polychronic time culture, because
interruptions may occur frequently, maximum fl exibility in setting up the training sessions
may be necessary.
Hofstede’s fi rst dimension, power distance, addresses how power issues are dealt
with in the culture. For example, if a superior in an organization has an incorrect belief
about an important issue, can a subordinate point out this error? In some cultures, the
answer is a resounding no. Consequently, this dimension could have major ramifi cations
for the successful deployment of an information system. For example, in a culture with
a high power distance, the deployment of a new information system is dependent on the
impression of the most important stakeholder (see Chapter 2). Th erefore, much care must
be taken to ensure that this stakeholder is pleased with the system. Otherwise, it might
never be used.
Hofstede’s second dimension, uncertainty avoidance, is based on the degree to which
the culture depends on rules for direction, how well individuals in the culture handle stress,
and the importance of employment stability. For example, in a high-uncertainty-avoidance
culture, the use of detailed procedures manuals (see Chapter 12) and good training (see later
in this chapter) can reduce the uncertainty in adopting the new system.
Hofstede’s third dimension, individualism versus collectivism, is based on the level of
emphasis the culture places on the individual or the collective. Th e relationship between the
individual and the group is important for the success of an information system. Depending
on the culture’s orientation, the success of an information system being transitioned into
production can depend on whether the focus of the information system will benefi t the indi-
vidual or the group.
Hofstede’s fourth dimension, masculinity versus femininity, addresses how well mas-
culine and feminine characteristics are valued by the culture. Some of the diff erences that
could aff ect the adoption of an information system include employee motivational issues.
In a masculine culture, motivation would be based on advancement, earnings, and training,
whereas in a feminine culture, motivations would include friendly atmosphere, physical
conditions, and cooperation. Depending on how the culture views this dimension, diff erent
motivations might need to be used to increase the likelihood of the information system being
successfully deployed.
Th e fi fth dimension, long- versus short-term orientation, deals with how the culture views
the past and the future. In East Asia, long-term thinking is highly respected, whereas in North
America and Europe, short-term profi ts and the current stock price seem to be the only things
that matter. Based on this dimension, all the political concerns raised previously in this text
become very important. For example, if the local culture views success only in a short-term
manner, then any new information system that is deployed to support one department of
an organization may give that department a competitive advantage over other departments
in the short run. If only short-run measures are used to judge the success of a department,
then it would be in the interest of the other departments to fi ght the successful deployment
of the information system. However, if a longer-run perspective is the norm, then the other
departments could be convinced to support the new information system because they could
have new supportive information systems in the future.
Obviously, when reviewing these dimensions, we can see they interact with each other.
Th e most important thing to remember from an IT perspective is that we must be careful
not to view the local user community through our eyes; in a global economy, we must take
into consideration the local cultural concerns for the information system to be deployed in a
successful manner.
Conversion 485
CONVERSION4
Conversion is the technical process by which a new system replaces an old system. Users are
moved from using the as-is business processes and computer programs to the to-be business
processes and programs. Th e migration plan specifi es what activities will be performed when
and by whom and includes both technical aspects (such as installing hardware and soft ware
and converting data from the as-is system to the to-be system) and organizational aspects
(such as training and motivating the users to embrace the new system). Conversion refers to
the technical aspects of the migration plan.
Th ere are three major steps to the conversion plan before commencement of operations:
Install hardware, install soft ware, and convert data (Figure 13-2). Although it may be possible
to do some of these steps in parallel, usually they must be done sequentially at any one location.
Th e fi rst step in the conversion plan is to buy and install any needed hardware. In many
cases, no new hardware is needed, but sometimes the project requires new hardware such as
servers, client computers, printers, and networking equipment. It is critical to work closely
with vendors who are supplying needed hardware and soft ware to ensure that the deliveries
are coordinated with the conversion schedule so that the equipment is available when it is
needed. Nothing can stop a conversion plan in its tracks as easily as the failure of a vendor to
deliver needed equipment.
Once the hardware is installed, tested, and certifi ed as being operational, the second step
is to install the soft ware. Th is includes the to-be system under development and, sometimes,
additional soft ware that must be installed to make the system operational. At this point, the
system is usually tested again to ensure that it operates as planned.
Th e third step is to convert the data from the as-is system to the to-be system. Data conversion
is usually the most technically complicated step in the migration plan. Oft en, separate programs
must be written to convert the data from the as-is system to the new formats required in the to-be
system and store it in the to-be system fi les and databases. Th is process is oft en complicated by the
fact that the fi les and databases in the to-be system do not exactly match the fi les and databases in
the as-is system (e.g., the to-be system may use several tables in a database to store customer data
that were contained in one fi le in the as-is system). Formal test plans are always required for data
conversion eff orts (see Chapter 12).
Conversion can be thought of along three dimensions: the style in which the conversion is
done (conversion style), what location or work groups are converted at what time (conversion
4 Th e material in this section is related to the Enhanced Unifi ed Process’s Transition phase and the Deployment
workfl ow (see Figure 1-18).
FIGURE 13-2
Elements of a
Migration Plan
Commence
operations
Conversion Plan
(Technical Issues)
Install hardware
Install software
Convert data
Change Management Plan
(Organizational Issues)
Revise management policies
Conduct training
Assess costs and benefits
Motivate adoption
486 Chapter 13 Installation and Operations
FIGURE 13-3
Conversion
Strategies
PhasedPilot
Modular
Whole system
Direct
Parallel
Simultaneous
Location
Modules
Style
location), and what modules of the system are converted at what time (conversion modules).
Figure 13-3 shows the potential relationships among these three dimensions.
Conversion Style
Th e conversion style is the way users are switched between the old and new systems. Th ere
are two fundamentally diff erent approaches to the style of conversion: direct conversion and
parallel conversion.
Direct Conversion With direct conversion (sometimes called cold turkey, big bang, or
abrupt cutover), the new system instantly replaces the old system. Th e new system is turned
on, and the old system is immediately turned off . Th is is the approach that we are likely to use
when we upgrade commercial soft ware (e.g., Microsoft Word) from one version to another;
we simply begin using the new version and stop using the old version.
Direct conversion is the simplest and most straightforward. However, it is the most risky
because any problems with the new system that have escaped detection during testing can
seriously disrupt the organization.
Parallel Conversion With parallel conversion, the new system is operated side by side with
the old system; both systems are used simultaneously. For example, if a new accounting
system is installed, the organization enters data into both the old system and the new system
and then carefully compares the output from both systems to ensure that the new system is
performing correctly. Aft er some time period (oft en one to two months) of parallel operation
and intense comparison between the two systems, the old system is turned off and the organ-
ization continues using the new system.
Th is approach is more likely to catch any major bugs in the new system and prevent the
organization from suff ering major problems. If problems are discovered in the new system, the
system is simply turned off and fi xed and then the conversion process starts again. Th e problem
with this approach is the added expense of operating two systems that perform the same function.
Conversion Location
Conversion location refers to the parts of the organization that are converted when the conversion
occurs. Oft en, parts of the organization are physically located in diff erent offi ces (e.g., Toronto,
Atlanta, Los Angeles). In other cases, location refers to diff erent organizational units located in
Conversion 487
diff erent parts of the same offi ce complex (e.g., order entry, shipping, purchasing). Th ere are at
least three fundamentally diff erent approaches to selecting the way diff erent organizational loca-
tions are converted: pilot conversion, phased conversion, and simultaneous conversion.
Pilot Conversion With a pilot conversion, one or more locations or units or work groups
within a location are selected to be converted fi rst as part of a pilot test. Th e locations partici-
pating in the pilot test are converted (using either direct or parallel conversion). If the system
passes the pilot test, then the system is installed at the remaining locations (again using either
direct or parallel conversion).
Pilot conversion has the advantage of providing an additional level of testing before the
system is widely deployed throughout the organization, so that any problems with the sys-
tem aff ect only the pilot locations. However, this type of conversion obviously requires more
time before the system is installed at all organizational locations. Also, it means that diff erent
organizational units are using diff erent versions of the system and business processes, which
can make it diffi cult for them to exchange data.
Phased Conversion With phased conversion, the system is installed sequentially at diff er-
ent locations. A fi rst set of locations is converted, then a second set, then a third set, and
so on, until all locations are converted. Sometimes there is a deliberate delay between the
diff erent sets (at least between the fi rst and the second), so that any problems with the sys-
tem are detected before too much of the organization is aff ected. In other cases, the sets are
converted back to back so that as soon as those converting one location have fi nished, the
project team moves to the next and continues the conversion.
Phased conversion has the same advantages and disadvantages of pilot conversion. In
addition, it means that fewer people are required to perform the actual conversion (and any
associated user training) than if all locations were converted at once.
Simultaneous Conversion Simultaneous conversion, as the name suggests, means that all
locations are converted at the same time. Th e new system is installed and made ready at all
locations; at a preset time, all users begin using the new system. Simultaneous conversion is
oft en used with direct conversion, but it can also be used with parallel conversion.
Simultaneous conversion eliminates problems with having diff erent organizational units
using diff erent systems and processes. However, it also means that the organization must have
suffi cient staff to perform the conversion and train the users at all locations simultaneously.
Conversion Modules
Although it is natural to assume that systems are usually installed in their entirety, this is not
always the case.
Whole-System Conversion A whole-system conversion, in which the entire system is
installed at one time, is the most common. It is simple and the easiest to understand. How-
ever, if the system is large and/or extremely complex (e.g., an enterprise resource-planning
system such as SAP or PeopleSoft ), the whole system can prove too diffi cult for users to learn
in one conversion step.
Modular Conversion When the modules5 within a system are separate and distinct, organ-
izations sometimes choose to convert to the new system one module at a time—i.e., using
modular conversion. Modular conversion requires special care in developing the system
5 In this case, a module is typically a component or a package, i.e., a set of collaborating classes.
488 Chapter 13 Installation and Operations
(and usually adds extra cost). Each module either must be written to work with both the old
and new systems or object wrappers (see Chapter 7) must be used to encapsulate the old sys-
tem from the new. When modules are tightly integrated, this is very challenging and therefore
is seldom done. However, when there is only a loose association between modules, module
conversion is easier. For example, consider a conversion from an old version of Microsoft
Offi ce to a new version. It is relatively simple to convert from the old version of Word to the
new version without simultaneously having to change from the old to the new version of
Microsoft Excel.
Modular conversion reduces the amount of training required to begin using the new
system. Users need training only in the new module being implemented. However, modular
conversion does take longer and has more steps than does the whole-system process.
Selecting the Appropriate Conversion Strategy
Each of the three dimensions in Figure 13-3 is independent, so that a conversion strategy can
be developed to fi t in any one of the boxes in this fi gure. Diff erent boxes can also be mixed
and matched into one conversion strategy. For example, one commonly used approach is
to begin with a pilot conversion of the whole system using parallel conversion in a hand-
ful of test locations. Once the system has passed the pilot test at these locations, it is then
installed in the remaining locations using phased conversion with direct cutover. Th ere are
three important factors to consider in selecting a conversion strategy: risk, cost, and the time
required (Figure 13-4).
Risk Aft er the system has passed a rigorous battery of unit, system, integration, and accept-
ance testing, it should be bug free . . . maybe. Because humans make mistakes, nothing built by
people is ever perfect. Even aft er all these tests, there might still be a few undiscovered bugs.
Th e conversion process provides one last step in which to catch these bugs before the system
goes live and the bugs have the chance to cause problems.
Parallel conversion is less risky than is direct conversion because it has a greater chance
of detecting bugs that have gone undiscovered in testing. Likewise, pilot conversion is less
risky than is phased conversion or simultaneous conversion because if bugs do occur, they
occur in pilot test locations whose staff are aware that they might encounter bugs. Because
potential bugs aff ect fewer users, there is less risk. Likewise, converting a few modules at a
time lowers the probability of a bug because there is more likely to be a bug in the whole
system than in any given module.
How important the risk is depends on the system being implemented—the combination
of the probability that bugs remain undetected in the system and the potential cost of those
undetected bugs. If the system has indeed been subjected to extensive methodical testing,
including alpha and beta testing, then the probability of undetected bugs is lower than if the
testing was less rigorous. However, there still might have been mistakes made in the analysis
Risk High Low Low Medium High High Medium
Cost Low High Medium Medium High Medium High
Time Short Long Medium Long Short Short Long
Conversion Style Conversion Location Conversion Modules
Direct Parallel Pilot Phased Simultaneous Whole-System Modular
Characteristic Conversion Conversion Conversion Conversion Conversion Conversion Conversion
FIGURE 13-4 Characteristics of Conversion Strategies
Change Management 489
process, so that although there might be no soft ware bugs, the soft ware might fail to properly
address the business needs.
Assessing the cost of a bug is challenging, but most analysts and senior managers can
make a reasonable guess at the relative cost of a bug. For example, the cost of a bug in an
automated stock market trading program or a heart–lung machine keeping someone alive
is likely to be much greater than a bug in a computer game or word processing program.
Th erefore, risk is likely to be a very important factor in the conversion process if the system
has not been as thoroughly tested as it might have been or if the cost of bugs is high. If the
system has been thoroughly tested or the cost of bugs is not that high, then risk becomes less
important to the conversion decision.
Cost As might be expected, diff erent conversion strategies have diff erent costs. Th ese costs
can include things such as salaries for people who work with the system (e.g., users, trainers,
system administrators, external consultants), travel expenses, operation expenses, communi-
cation costs, and hardware leases. Parallel conversion is more expensive than direct cutover
because it requires that two systems (the old and the new) be operated at the same time.
Employees must then perform twice the usual work because they have to enter the same data
into both the old and the new systems. Parallel conversion also requires the results of the two
systems to be completely crosschecked to make sure there are no diff erences between the two,
which entails additional time and cost.
Pilot conversion and phased conversion have somewhat similar costs. Simultaneous con-
version has higher costs because more staff are required to support all the locations as they
simultaneously switch from the old to the new system. Modular conversion is more expensive
than whole-system conversion because it requires more programming. Th e old system must
be updated to work with selected modules in the new system, and modules in the new system
must be programmed to work with selected modules in both the old and new systems.
Time Th e fi nal factor is the amount of time required to convert between the old and the
new system. Direct conversion is the fastest because it is immediate. Parallel conversion
takes longer because the full advantages of the new system do not become available until
the old system is turned off . Simultaneous conversion is fastest because all locations are
converted at the same time. Phased conversion usually takes longer than pilot conversion
because once the pilot test is complete all remaining locations are usually (but not always)
converted simultaneously. Phased conversion proceeds in waves, oft en requiring several
months before all locations are converted. Likewise, modular conversion takes longer than
whole-system conversion because the models are introduced one aft er another.
CHANGE MANAGEMENT6
In the context of a systems development project, change management is the process of
helping people to adopt and adapt to the to-be system and its accompanying work processes
without undue stress. Th ere are three key roles in any major organizational change. Th e fi rst
is the sponsor of the change—the person who wants the change. Th is person is the business
sponsor who fi rst initiated the request for the new system (see Chapter 2). Usually, the
6 Th e material in this section is related to the Enhanced Unifi ed Process’s Transition and Production phases and the
Confi guration and Change Management workfl ow (see Figure 1-18). Many books have been written on change man-
agement. Some of our favorites are the following: Patrick Connor and Linda Lake, Managing Organizational Change,
2nd Ed. (Westport, CT: Praeger, 1994); Douglas Smith, Taking Charge of Change (Reading, MA: Addison-Wesley,
1996); Daryl Conner, Managing at the Speed of Change (New York: Villard Books, 1992); Mary Lynn Manns and
Linda Rising, Fearless Change: Patterns for Introducing New Ideas (Boston: Addison-Wesley, 2005).
490 Chapter 13 Installation and Operations
sponsor is a senior manager of the part of the organization that must adopt and use the new
system. It is critical that the sponsor be active in the change management process because a
change that is clearly being driven by the sponsor, not by the project team or the IS organi-
zation, has greater legitimacy. Th e sponsor has direct management authority over those who
adopt the system.
Th e second role is that of the change agent—the person(s) leading the change eff ort.
Th e change agent, charged with actually planning and implementing the change, is usu-
ally someone outside of the business unit adopting the system and therefore has no direct
management authority over the potential adopters. Because the change agent is an outsider,
he or she has less credibility than do the sponsor and other members of the business unit.
Aft er all, once the system has been installed, the change agent usually leaves and thus has
no ongoing impact.
Th e third role is that of potential adopters, or targets of the change—the people who
actually must change. Th ese are the people for whom the new system is designed and who will
ultimately choose to use or not use the system.
In the early days of computing, many project teams simply assumed that their job ended
when the old system was converted to the new system at a technical level. Th e philosophy was
“build it and they will come.” Unfortunately, that happens only in the movies. Resistance to
change is common in most organizations. Th erefore, the change management plan is an impor-
tant part of the overall installation plan that glues together the key steps in the change man-
agement process. Successful change requires that people want to adopt the change and are able
to adopt the change. Th e change management plan has four basic steps: revising management
policies, assessing the cost and benefi t models of potential adopters, motivating adoption, and
enabling people to adopt through training (see Figure 13-2). However, before we can discuss the
change management plan, we must fi rst understand why people resist change.
Understanding Resistance to Change7
People resist change—even change for the better—for very rational reasons. What is good for
the organization is not necessarily good for the people who work there. For example, con-
sider an order-processing clerk who used to receive orders to be shipped on paper shipping
documents but now uses a computer to receive the same information. Rather than typing
shipping labels with a typewriter, the clerk now clicks on the print button on the computer
and the label is produced automatically. Th e clerk can now ship many more orders each day,
which is a clear benefi t to the organization. Th e clerk, however, probably doesn’t really care
how many packages are shipped. His or her pay doesn’t change; it’s just a question of which
the clerk prefers to use, a computer or typewriter. Learning to use the new system and work
processes—even if the change is minor—requires more eff ort than continuing to use the
existing, well-understood system and work processes.
So why do people accept change? Simply put, every change has a set of costs and benefi ts
associated with it. If the benefi ts of accepting the change outweigh the costs of the change,
then people change. Sometimes the benefi t of change is avoidance of the pain that might be
experienced if the change were not adopted (e.g., if you don’t change, you are fi red, so one of
the benefi ts of adopting the change is that you still have a job).
In general, when people are presented with an opportunity for change, they perform
a cost–benefi t analysis (sometime consciously, sometimes subconsciously) and decide the
extent to which they will embrace and adopt the change. Th ey identify the costs of and
7 Th is section benefi ted from conversations with Dr. Robert Briggs, research scientist at the Center for the Manage-
ment of Information at the University of Arizona.
Change Management 491
benefi ts from the system and decide whether the change is worthwhile. However, it is not
that simple, because most costs and benefi ts are not certain. Th ere is some uncertainty as
to whether a certain benefi t or cost will actually occur; so both the costs of and benefi ts
from the new system need to be weighted by the degree of certainty associated with them
(Figure 13-5). Unfortunately, most humans tend to overestimate the probability of costs and
underestimate the probability of benefi ts.
Th ere are also costs and, sometimes, benefi ts associated with the actual transition pro-
cess itself. For example, suppose we found a nicer house or apartment than our current one.
Even if we liked it better, we might decide not to move simply because the cost of moving
outweighed the benefi ts from the new house or apartment itself. Likewise, adopting a new
computer system might require us to learn new skills, which could be seen as a cost to some
people or as a benefi t to others, if they perceived that those skills would somehow provide
other benefi ts beyond the use of the system itself. Once again, any costs and benefi ts from
the transition process must be weighted by the certainty with which they will occur (see
Figure 13-5).
Taken together, these two sets of costs and benefi ts (and their relative certainties) aff ect
the acceptance of change or resistance to change that project teams encounter when install-
ing new systems in organizations. Th e fi rst step in change management is to understand the
factors that inhibit change—the factors that aff ect the perception of costs and benefi ts and
certainty that they will be generated by the new system. It is critical to understand that the
real costs and real benefi ts are far less important than the perceived costs and perceived bene-
fi ts. People act on what they believe to be true, not on what is true. Th us, any understanding
of how to motivate change must be developed from the viewpoint of the people expected to
change, not from the viewpoint of those leading the change.
Revising Management Policies
Th e fi rst major step in the change management plan is to change the management policies that
were designed for the as-is system to new management policies designed to support the to-be
system. Management policies provide goals, defi ne how work processes should be performed, and
determine how organizational members are rewarded. No computer system will be successfully
adopted unless management policies support its adoption. Many new computer systems bring
changes to business processes; they enable new ways of working. Unless the policies that provide
the rules and rewards for those processes are revised to refl ect the new opportunities that the
system permits, potential adopters cannot easily use it.
FIGURE 13-5
The Costs and
Benefi ts of Change
As-Is
System
Restraining
Factors
Enabling
Factors
Costs of
Transition
X
Certainty of
Costs
Occurring
Benefits of
Transition
X
Certainty of
Benefits
Occurring
To-Be
System
Transition
Restraining
Factors
Enabling
Factors
Costs of
To-Be System
X
Certainty of
Costs
Occurring
Benefits of
To-Be System
X
Certainty of
Benefits
Occurring
492 Chapter 13 Installation and Operations
Management has three basic tools for structuring work processes in organizations.8 Th e
fi rst are the standard operating procedures (SOPs) that become the habitual routines for how
work is performed. Th e SOPs are both formal and informal. Formal SOPs defi ne proper
behavior. Informal SOPs are the norms that have developed over time for how processes are
actually performed. Management must ensure that the formal SOPs are revised to match the
to-be system. Th e informal SOPs will then evolve to refi ne and fi ll in details absent in the
formal SOPs.
Th e second aspect of management policy is defi ning how people assign meaning to
events. What does it mean to “be successful” or “do good work”? Policies help people
understand meaning by defi ning measurements and rewards. Measurements explicitly defi ne
meaning because they provide clear and concrete evidence about what is important to the
organization. Rewards reinforce measurements because “what gets measured gets done” (an
overused but accurate saying). Measurements must be carefully designed to motivate desired
behavior.
A third aspect of management policy is resource allocation. Managers can have clear
and immediate impacts on behavior by allocating resources. Th ey can redirect funds and
staff from one project to another, create an infrastructure that supports the new system, and
invest in training programs. Each of these activities has both a direct and symbolic eff ect. Th e
direct eff ect comes from the actual reallocation of resources. Th e symbolic eff ect shows that
management is serious about its intentions. Th ere is less uncertainty about management’s
long-term commitment to a new system when potential adopters see resources being com-
mitted to support it.
Assessing Costs and Benefi ts
Th e next step in developing a change management plan is to develop two clear and concise
lists of costs and benefi ts provided by the new system (and the transition to it) compared with
the as-is system. Th e fi rst list is developed from the perspective of the organization, which
should fl ow easily from the business case developed during the feasibility study and refi ned
over the life of the project (see Chapter 2). Th is set of organizational costs and benefi ts should
be distributed widely so that everyone expected to adopt the new system should clearly under-
stand why the new system is valuable to the organization.
Th e second list of costs and benefi ts is developed from the viewpoints of the diff erent
potential adopters expected to change, or stakeholders in the change. For example, one set of
potential adopters may be the frontline employees, another may be the fi rst-line supervisors,
and yet another might be middle management. Each of these potential adopters, or stake-
holders, may have a diff erent set of costs and benefi ts associated with the change—costs and
benefi ts that can diff er widely from those of the organization. In some situations, unions may
be key stakeholders that can make or break successful change.
Many systems analysts naturally assume that frontline employees are the ones whose set
of costs and benefi ts are the most likely to diverge from those of the organization and thus are
the ones who most resist change. However, they usually bear the brunt of problems with the
current system. When problems occur, they oft en experience them fi rsthand. Middle manag-
ers and fi rst-line supervisors are the most likely to have a divergent set of costs and benefi ts
and, therefore, resist change because new computer systems oft en change how much power
they have. For example, a new computer system may improve the organization’s control over
8 Th is section builds on the work of Anthony Giddons, Th e Constitution of Society: Outline of the Th eory of Structure
(Berkeley: University of California Press, 1984). A good summary of Giddons’s theory that has been revised and
adapted for use in understanding information systems is an article by Wanda Orlikowski and Dan Robey, “Infor-
mation Technology and the Structuring of Organizations,” Information Systems Research 2, no. 2 (1991): 143–169.
Change Management 493
a work process (a benefi t to the organization) but reduce the decision-making power of middle
management (a clear cost to middle managers).
An analysis of the costs and benefi ts for each set of potential adopters, or stakeholders,
will help pinpoint those who will likely support the change and those who might resist the
change. Th e challenge at this point is to try to change the balance of the costs and benefi ts
for those expected to resist the change so that they support it (or at least do not actively resist
it). Th is analysis could uncover some serious problems that have the potential to block the
successful adoption of the system. It may be necessary to reexamine the management policies
and make signifi cant changes to ensure that the balance of costs and benefi ts is such that
important potential adopters are motivated to adopt the system.
Figure 13-6 summarizes some of the factors that are important to successful change.
Th e fi rst and most important reason is a compelling personal reason to change. All change is
made by individuals, not organizations. If there are compelling reasons for the key groups of
individual stakeholders to want the change, then the change is more likely to be successful.
Factors such as increased salary, reduced unpleasantness, and—depending on the individu-
als—opportunities for promotion and personal development can be important motivators.
However, if the change makes current skills less valuable, individuals might resist the change
because they have invested a lot of time and energy in acquiring those skills, and anything that
diminishes those skills may be perceived as diminishing the individual (because important
skills bring respect and power).
Th ere must also be a compelling reason for the organization to need the change; other-
wise, individuals become skeptical that the change is important and are less certain it will,
in fact, occur. Probably the hardest organization to change is an organization that has been
successful because individuals come to believe that what worked in the past will continue to
work. By contrast, in an organization that is on the brink of bankruptcy, it is easier to convince
individuals that change is needed. Commitment and support from credible business sponsors
and top management are also important in increasing the certainty that the change will occur.
Th e likelihood of successful change is increased when the cost of the transition to indi-
viduals who must change is low. Th e need for signifi cantly diff erent new skills or disruptions
in operations and work habits can create resistance. A clear migration plan developed by a
credible change agent who has support from the business sponsor is an important factor in
increasing the certainty about the costs of the transition process.
Motivating Adoption
Th e single most important factor in motivating a change is providing clear and convincing
evidence of the need for change. Simply put, everyone who is expected to adopt the change
must be convinced that the benefi ts from the to-be system outweigh the costs of changing.
Th ere are two basic strategies to motivating adoption: informational and political.
Both strategies are oft en used simultaneously. With an informational strategy, the goal is to
convince potential adopters that the change is for the better. Th is strategy works when the
cost–benefi t set of the target adopters has more benefi ts than costs. In other words, there
really are clear reasons for the potential adopters to welcome the change.
Using this approach, the project team provides clear and convincing evidence of the costs
and benefi ts of moving to the to-be system. Th e project team writes memos and develops pres-
entations that outline the costs and benefi ts of adopting the system from the perspective of the
organization and from the perspective of the target group of potential adopters. Th is information
is disseminated widely throughout the target group, much like an advertising or public relations
campaign. It must emphasize the benefi ts and increase the certainty in the minds of potential
adopters that these benefi ts will actually be achieved. In our experience, it is always easier to sell
painkillers than vitamins; that is, it is easier to convince potential adopters that a new system will
494 Chapter 13 Installation and Operations
Factor Examples Effects Actions to Take
Benefi ts of to-be
system
Compelling
personal
reason(s) for
change
Increased pay, fewer
unpleasant aspects,
opportunity for pro-
motion, most existing
skills remain valuable
If the new system provides clear
personal benefi ts to those who
must adopt it, they are more
likely to embrace the change.
Perform a cost–benefi t analysis
from the viewpoint of the
stakeholders, make changes
where needed, and actively
promote the benefi ts.
Certainty of
benefi ts
Compelling
organizational
reason(s) for
change
Risk of bankruptcy,
acquisition,
government
regulation
If adopters do not understand
why the organization is
implementing the change,
they are less certain that the
change will occur.
Perform a cost–benefi t analysis
from the viewpoint of the
organization and launch
a vigorous information
campaign to explain the
results to everyone.
Demonstrated top
management
support
Active involvement,
frequent mentions in
speeches
If top management is not seen
to actively support the change,
there is less certainly that the
change will occur.
Encourage top management
to participate in the
information campaign.
Committed and
involved business
sponsor
Active involvement,
frequent visits to users
and project team,
championing
If the business sponsor (the
functional manager who
initiated the project) is not
seen to actively support the
change, there is less certainty
that the change will occur.
Encourage the business
sponsor to participate in the
information campaign and
play an active role in the
change management plan.
Credible top
management and
business sponsor
Management and
sponsor who do what
they say instead of
being members of the
“management fad of
the month” club
If the business sponsor and top
management have credibility
in the eyes of the adopters,
the certainty of the claimed
benefi ts is higher.
Ensure that the business
sponsor and/or top
management has credibility
so that such involvement
will help; if there is no
credibility, involvement will
have little effect.
Costs of
transition
Low personal costs
of change
Few new skills needed The cost of the change is
not borne equally by all
stakeholders; the costs are
likely to be higher for some.
Perform a cost–benefi t analysis
from the viewpoint of the
stakeholders, make changes
where needed, and actively
promote the low costs.
Certainty of costs Clear plan for
change
Clear dates and
instructions for change,
clear expectations
If there is a clear migration
plan, it will likely lower the
perceived costs of transition.
Publicize the migration plan.
Credible change
agent
Previous experience
with change, does
what he/she promises
to do
If the change agent has
credibility in the eyes of the
adopters, the certainty of the
claimed costs is higher.
If the change agent is not
credible, then change will
be diffi cult.
Clear mandate for
change agent
from sponsor
Open support for
change agent when
disagreements occur
If the change agent has a clear
mandate from the business
sponsor, the certainty of the
claimed costs is higher.
The business sponsor must
actively demonstrate
support for the change
agent.
FIGURE 13-6 Major Factors in Successful Change
remove a major problem (or other source of pain) than that it will provide new benefi ts (e.g.,
increase sales). Th erefore, informational campaigns are more likely to be successful if they stress
reducing or eliminating problems rather than focusing on providing new opportunities.
Th e other strategy for motivating change is a political strategy. With a political strategy,
organizational power, not information, is used to motivate change. Th is approach is oft en
Change Management 495
used when the cost–benefi t set of the target adopters has more costs than benefi ts. In other
words, although the change might benefi t the organization, there are no reasons for the
potential adopters to welcome the change.
Th e political strategy is usually beyond the control of the project team. It requires some-
one in the organization who holds legitimate power over the target group to infl uence the
group to adopt the change. Th is may be done in a coercive manner (e.g., adopt the system
or you’re fi red) or in a negotiated manner, in which the target group gains benefi ts in other
ways that are linked to the adoption of the system (e.g., linking system adoption to increased
training opportunities). Management policies can play a key role in a political strategy by
linking salary to certain behaviors desired with the new system.
In general, for any change that has true organizational benefi ts, about 20 to 30 percent of
potential adopters will be ready adopters. Th ey recognize the benefi ts, quickly adopt the sys-
tem, and become proponents of the system. Another 20 to 30 percent are resistant adopters.
Th ey simply refuse to accept the change and they fi ght it, either because the new system has
more costs than benefi ts for them personally or because they place such a high cost on the
transition process itself that no amount of benefi ts from the new system can outweigh the
change costs. Th e remaining 40 to 60 percent are reluctant adopters. Th ey tend to be apathetic
and will go with the fl ow to either support or resist the system, depending on how the project
evolves and how their coworkers react to the system. Figure 13-7 illustrates the actors who are
involved in the change management process.
Th e goal of change management is to actively support and encourage the ready adopters
and help them win over the reluctant adopters. Th ere is usually little that can be done about
the resistant adopters because their set of costs and benefi ts may be divergent from those of
the organization. Unless there are simple steps that can be taken to rebalance their costs and
benefi ts or the organization chooses to adopt a strongly political strategy, it is oft en best to
ignore this small minority of resistant adopters and focus on the larger majority of ready and
reluctant adopters.
Enabling Adoption: Training
Potential adopters might want to adopt the change, but unless they are capable of adopting
it, they won’t. Careful training enables adoption by providing the skills needed to adopt
the change. Training is probably the most self-evident part of any change management
initiative. How can an organization expect its staff members to adopt a new system if they are
not trained? However, we have found that training is one of the most commonly overlooked
parts of the process. Many organizations and project managers simply expect potential adop-
ters to fi nd the system easy to learn. Because the system is presumed to be so simple, it is taken
for granted that potential adopters should be able to learn with little eff ort. Unfortunately, this
is usually an overly optimistic assumption.
Every new system requires new skills, either because the basic work processes have changed
or because the computer system used to support the processes is diff erent. Th e more radical the
changes to the business processes, the more important it is to ensure the organization has the new
The sponsor wants The change agent leads Potential adopters are the people
the change to occur. the change effort. who must change.
20–30 percent are ready adopters.
20–30 percent are resistant adopters.
40–60 percent are reluctant adopters.
Sponsor Change Agent Potential Adopters
FIGURE 13-7
Actors in the
Change
Management
Process
496 Chapter 13 Installation and Operations
skills required to operate the new business processes and supporting information systems. In
general, there are three ways to get these new skills. One is to hire new employees who have the
needed skills that the existing staff does not. Another is to outsource the processes to an organi-
zation that has the skills that the existing staff does not. Both these approaches are controversial
and are usually considered only when the new skills needed are likely to be the most diff erent
from the set of skills of the current staff . In most cases, organizations choose the third alternative:
training existing staff in the new business processes and the to-be system. Every training plan
must consider what to train and how to deliver the training.
What to Train What training should you provide to the system users? It’s obvious: how to
use the system. Th e training should cover all the capabilities of the new system so that users
understand what each module does, right? Wrong. Training for business systems should
focus on helping the users to accomplish their jobs, not on how to use the system. Th e system
is simply a means to an end, not the end in itself. Th is focus on performing the job (i.e., the
business processes), not using the system, has two important implications. First, the training
must focus on the activities around the system as well as on the system itself. Th e training
must help the users understand how the computer fi ts into the bigger picture of their jobs.
Th e use of the system must be put in context of the manual business processes as well as of
those that are computerized, and it must also cover the new management policies that were
implemented along with the new computer system.
Second, the training should focus on what the user needs to do, not what the system
can do. Th is is a subtle—but very important—distinction. Most systems provide far more
capabilities than the users will need to use (e.g., when was the last time you wrote a macro in
Microsoft Word?). Rather than attempting to teach the users all the features of the system,
training should instead focus on the much smaller set of activities that users perform on a
regular basis and ensure that users are truly expert in those. When the focus is on the 20 per-
cent of functions that the users will use 80 percent of the time (instead of attempting to cover
all functions), users become confi dent about their ability to use the system. Training should
mention the other little-used functions but only so that users are aware of their existence and
know how to learn about them when their use becomes necessary.
One source of guidance for designing training materials is the use cases. Th e use cases
outline the common activities that users perform and thus can be helpful in understanding
the business processes and system functions that are likely to be most important to the users.
How to Train Th ere are many ways to deliver training. Th e most commonly used approach
is classroom training, in which many users are trained at the same time by the same instructor.
Th is has the advantage of training many users at one time with only one instructor and creates
a shared experience among the users.
It is also possible to provide one-on-one training, in which one trainer works closely with
one user at a time. Th is is obviously more expensive, but the trainer can design the training
program to meet the needs of individual users and can better ensure that the users really do
understand the material. Th is approach is typically used only when the users are very impor-
tant or when there are very few users.
Another approach that is becoming more common is to use some form of computer-based
training (CBT), in which the training program is delivered via computer, either on CD or over
the Web. CBT programs can include text slides, audio, and even video and animation. CBT
is typically more costly to develop but is cheaper to deliver because no instructor is needed to
actually provide the training.
Figure 13-8 summarizes four important factors to consider in selecting a training
method: cost to develop, cost to deliver, impact, and reach. CBT is typically more expensive
Post-Implementation Activities 497
Cost to develop Low to Medium Medium High
Cost to deliver High Medium Low
Impact High Medium to High Low to Medium
Reach Low Medium High
One-on-One Classroom Computer-Based
Training Training Training
FIGURE 13-8
Selecting a
Training Method
to develop than one-on-one or classroom training, but it is less expensive to deliver. One-
on-one training has the most impact on the user because it can be customized to the user’s
precise needs, knowledge, and abilities, whereas CBT has the least impact. However, CBT has
the greatest reach—the ability to train the most users over the widest distance in the shortest
time—because it is much simpler to distribute than classroom and one-on-one training, sim-
ply because no instructors are needed.
Figure 13-8 suggests a clear pattern for most organizations. If there are only a few users to
train, one-on-one training is the most eff ective. If there are many users to train, many organ-
izations turn to CBT. We believe that the use of CBT will increase in the future. Quite oft en,
large organizations use a combination of all three methods. Regardless of which approach
is used, it is important to leave the users with a set of easily accessible materials that can be
referred to long aft er the training has ended (usually a quick reference guide and a set of
manuals, whether on paper or in electronic form).
POST-IMPLEMENTATION ACTIVITIES9
Th e goal of post-implementation activities is the institutionalization of the use of the new
system—i.e., to make it the normal, accepted, routine way of performing the business processes.
Post-implementation activities attempt to refreeze the organization aft er the successful tran-
sition to the new system. Although the work of the project team naturally winds down aft er
implementation, the business sponsor and sometimes the project manager are actively involved
in refreezing. Th ese two—and, ideally, many other stakeholders—actively promote the new
system and monitor its adoption and usage. Th ey usually provide a steady fl ow of information
about the system and encourage users to contact them to discuss issues.
In this section, we examine three key post-implementation activities: system support
(providing assistance in the use of the system), system maintenance (continuing to refi ne
and improve the system), and project assessment (analyzing the project to understand what
activities were done well—and should be repeated—and what activities need improvement in
future projects).
System Support
Once the project team has installed the system and performed the change management activ-
ities, the system is offi cially turned over to the operations group. Th is group is responsible for
operating the system, whereas the project team was responsible for developing the system.
Members of the operations group are usually closely involved in the installation activities
because they are the ones who must ensure that the system actually works. Aft er the system
is installed, the project team leaves but the operations group remains.
9 Th e material in this section is related to the Enhanced Unifi ed Process’s Production Phase and the Operations and
Support workfl ow (see Figure 1-18).
498 Chapter 13 Installation and Operations
Providing system support means helping the users to use the system. Usually, this means
providing answers to questions and helping users understand how to perform a certain func-
tion; this type of support can be thought of as on-demand training.
Online support is the most common form of on-demand training. Th is includes the doc-
umentation and help screens built into the system, as well as separate websites that provide
answers to frequently asked questions (FAQs), which enable users to fi nd answers without
contacting a person. Obviously, the goal of most system support is to provide suffi ciently good
online support so that the user doesn’t need to contact a person, because providing online sup-
port is much less expensive than is providing a person to answer questions.
Most organizations provide a help desk that provides a place for a user to talk with a
person who can answer questions (usually over the phone but sometimes in person). Th e
help desk supports all systems, not just one specifi c system, so it receives calls about a wide
variety of soft ware and hardware. Th e help desk is operated by level-1 support staff who have
very broad computer skills and are able to respond to a wide range of requests, from network
problems and hardware problems to problems with commercial soft ware and problems with
the business application soft ware developed in-house.
Th e goal of most help desks is to have the level-1 support staff resolve 80 percent of the
help requests they receive on the fi rst call. If the issue cannot be resolved by level 1 support
staff , a problem report (Figure 13-9) is completed (oft en using a special computer system
designed to track problem reports) and passed to a level-2 support staff member.
Th e level-2 support staff members are people who know the application system well and can
provide expert advice. For a new system, they are usually selected during the implementation
phase and become familiar with the system as it is being tested. Sometimes the level-2 support
staff members participate in training during the change management process to become more
knowledgeable about the system, the new business processes, and the users themselves.
Th e level-2 support staff works with users to resolve problems. Most problems are suc-
cessfully resolved by the level-2 staff . However, sometimes, particularly in the fi rst few months
aft er the system is installed, the problem turns out to be a bug in the soft ware that must be
fi xed. In this case, the problem report becomes a change request that is passed to the system
maintenance group (see the next section).
System Maintenance
System maintenance is the process of refi ning the system to make sure it continues to meet
business needs. More money and eff ort are devoted to system maintenance than to the initial
development of the system, simply because a system continues to change and evolve as it is used.
Most beginning systems analysts and programmers work fi rst on maintenance projects; usually
only aft er they have gained some experience are they assigned to new development projects.
• Time and date of the report
• Name, e-mail address, and telephone number of the
support person taking the report
• Name, e-mail address, and telephone number of the
person who reported the problem
• Software and/or hardware causing problem
• Location of the problem
• Description of the problem
• Action taken
• Disposition (problem fi xed or forwarded to system
maintenance)
FIGURE 13-9
Elements of a
Problem Report
Post-Implementation Activities 499
Every system is “owned” by a project manager in the IS group (Figure 13-10). Th is individ-
ual is responsible for coordinating the system’s maintenance eff ort for that system. Whenever
a potential change to the system is identifi ed, a change request is prepared and forwarded to
the project manager. Th e change request is a smaller version of the system request discussed in
Chapter 2. It describes the change requested and explains why the change is important.
Changes can be small or large. Change requests that are likely to require a signifi cant
eff ort are typically handled in the same manner as system requests: Th ey follow the same
process as the project described in this book, starting with project identifi cation in Chapter 2
and following through installation in this chapter. Minor changes typically follow a smaller
version of this same process. Th ere is an initial assessment of feasibility and of costs and ben-
efi ts, and the change request is prioritized. Th en a systems analyst (or a programmer/analyst)
performs the analysis, which might include interviewing users, and prepares an initial design
before programming begins. Th e new (or revised) program is then extensively tested before
the system is converted from the old system to the revised one.
2. Change Request
with Feasibility,
Costs, and Benefits
3. Priority
4. Change Request
5. Design
6. Changed
System
1. Potential
Change
Analyst
Users
Change Committee
Programmer
Project
Manager
Problem Reports
Results: Passed Open items:
Test ID: Requirement addressed:
Objective:
Test cases
Interface ID
Script
Expected results notes
Actual results notes
1.
2.
3.
4.
5.
6.
Data Field Value Entered
Problem Reports
ORD56-3.5 ZIP code/postal code blank
ORD56-3.5 ZIP code/postal code 9021
ORD56-3.5 ZIP code/postal code 90210
ORD56-3.5 ZIP code/postal code C1A58
ORD56-3.5 ZIP code/postal code CAA 2C6
ORD56-3.5
Test 3 and 6 are valid U.S. and Canadian codes that match tested city. All others should be rejected.
Test 3 and 6 accepted. Tests 1, 2, 4, and 5 were rejected with correct message.
ZIP code/postal code C1A 2C6
12 Verify ordering information
Ensure that the information entered by the customer on the place-order form is valid
Changes to Other Systems
Results: Passed Open items:
Test ID: Requirement addressed:
Objective:
Test cases
Interface ID
Script
Expected results notes
Actual results notes
1.
2.
3.
4.
5.
6.
Data Field Value Entered
Change Request
ORD56-3.5 ZIP code/postal code blank
ORD56-3.5 ZIP code/postal code 9021
ORD56-3.5 ZIP code/postal code 90210
ORD56-3.5 ZIP code/postal code C1A58
ORD56-3.5 ZIP code/postal code CAA 2C6
ORD56-3.5
Test 3 and 6 are valid U.S. and Canadian codes that match tested city. All others should be rejected.
Test 3 and 6 accepted. Tests 1, 2, 4, and 5 were rejected with z message.
ZIP code/postal code C1A 2C6
12 Verify ordering information
Ensure that the information entered by the customer on the place-order form is valid
Software or Network Changes
FIGURE 13-10 Processing a Change Request
500 Chapter 13 Installation and Operations
Change requests typically come from fi ve sources. Th e most common source is problem
reports from the operations group that identify bugs in the system that must be fi xed. Th ese
are usually given immediate priority because a bug can cause signifi cant problems. Even a
minor bug can cause major problems by upsetting users and reducing their acceptance of and
confi dence in the system.
Th e second most common source of change requests is enhancement to the system from
users. As users work with the system, they oft en identify minor changes in the design that can
make the system easier to use or identify additional functions that are needed. Such enhancements
are important in satisfying the users and are oft en key in ensuring that the system changes as the
business requirements change. Enhancements are oft en given second priority aft er bug fi xes.
Th e third source of change requests is other system development projects. For example,
if the doctor in the appointment problem decided that he or she would like to have a Web-
based appointment system that would allow patients to directly interact with the current
appointment system, it is likely that other systems, such as billing, would have to be modifi ed
to ensure that the two systems would work together. Th ese changes required by the need to
integrate two systems are generally rare but are becoming more common as system integra-
tion eff orts become more common.
Th e fourth source of change requests is those that occur when underlying soft ware or net-
works change. For example, new versions of Windows oft en require an application to change
the way the system interacts with Windows or enables application systems to take advantage of
new features that improve effi ciency. Although users might never see these changes (because
most changes are inside the system and do not aff ect its user interface or functionality), these
changes can be among the most challenging to implement because analysts and programmers
must learn about the new system characteristics, understand how application systems use (or
can use) those characteristics, and then make the needed programming changes.
Th e fi fth source of change requests is senior management. Th ese change requests are
oft en driven by major changes in the organization’s strategy or operations. Th ese signifi cant
change requests are typically treated as separate projects, but the project manager responsible
for the initial system is oft en placed in charge of the new project.
Project Assessment
Th e goal of project assessment is to understand what was successful about the system and the
project activities (and, therefore, should be continued in the next system or project) and what
needs to be improved. Project assessment is not routine in most organizations, except for
military organizations, which are accustomed to preparing aft er-action reports. Nonetheless,
assessment can be an important component in organizational learning because it helps organ-
izations and people understand how to improve their work. It is particularly important for
junior staff members because it helps promote faster learning. Th ere are two primary parts to
project assessment—project team review and system review.
Project Team Review A project team review focuses on the way the project team carried out
its activities. Each project member prepares a short two- to three-page document that reports
and analyzes his or her performance. Th e focus is on performance improvement, not pen-
alties for mistakes made. By explicitly identifying mistakes and understanding their causes,
project team members will, it is hoped, be better prepared for the next time they encounter a
similar situation—and less likely to repeat the same mistakes. Likewise, by identifying excel-
lent performance, team members will be able to understand why their actions worked well
and how to repeat them in future projects.
Th e project manager, who meets with the team members to help them understand how
to improve their performance, assesses the documents prepared by each team member. Th e
Post-Implementation Activities 501
project manager then prepares a summary document that outlines the lessons learned from
the project. Th is summary identifi es what actions should be taken in future projects to improve
performance but is careful not to identify team members who made mistakes. Th e summary is
widely circulated among all project managers to help them understand how to manage their
projects better. Oft en, it is also circulated among regular staff members who did not work on
the project so that they, too, can learn from other projects.
System Review Th e focus of the system review is to understand the extent to which the
proposed costs and benefi ts from the new system identifi ed during feasibility analysis were
actually recognized from the implemented system. Project team review is usually conducted
immediately aft er the system is installed while key events are still fresh in team members’
minds, but system review is oft en undertaken several months aft er the system is installed
because it oft en takes a while before the system can be properly assessed.
System review starts with the system request and feasibility analysis prepared at the start
of the project. Th e detailed analyses prepared for the expected business value (both tangible
and intangible) as well as the economic feasibility analysis are reexamined, and a new analysis
is prepared aft er the system has been installed. Th e objective is to compare the anticipated
business value against the actual realized business value from the system. Th is helps the
organization assess whether the system actually provided the value it was planned to provide.
Whether or not the system provides the expected value, future projects can benefi t from an
improved understanding of the true costs and benefi ts.
A formal system review also has important behavior implications for project initiation.
Because everyone involved with the project knows that all statements about business value
and the fi nancial estimates prepared during project initiation will be evaluated at the end of
the project, they have an incentive to be conservative in their assessments. No one wants to
be the project sponsor or project manager for a project that goes radically over budget or fails
to deliver promised benefi ts.
How do you avoid bugs in the commercial software you
buy? Here are six tips:
1. Know your software: Find out if the few programs you
use day in and day out have known bugs and patches,
and track the websites that offer the latest information
on them.
2. Back up your data: This dictum should be tattooed on
every monitor. Stop reading right now and copy the
data you can’t afford to lose onto a second hard disk
or Web server. We’ll wait.
3. Don’t upgrade—yet: It’s tempting to upgrade to the
latest and greatest version of your favorite software,
but why chance it? Wait a few months, check out
other users’ experiences with the upgrade on Usenet
newsgroups or the vendor’s own discussion forum,
and then go for it. But only if you must.
4. Upgrade slowly: If you decide to upgrade, allow your-
self at least a month to test the upgrade on a separate
system before you install it on all the computers in
your home or offi ce.
5. Forget the betas: Installing beta software on your pri-
mary computer is a game of Russian roulette. If you
really have to play with beta software, get a second
computer.
6. Complain: The more you complain about bugs and
demand remedies, the more costly it is for vendors to
ship buggy products. It’s like voting—the more people
participate, the better the results.
Based upon material from “Software Bugs Run Rampant,” PC World
17, no. 1 (January 1999): 46.
13-1 Beating Buggy SoftwarePRACTICAL
TIP
502 Chapter 13 Installation and Operations
KEY TERMS
Change agent
Change management
Change request
Classroom training
Collectivism
Computer-based training
(CBT)
Context
Conversion
Conversion location
Conversion modules
Conversion strategy
Conversion style
Cost
Direct conversion
Femininity
Frequently asked
questions (FAQ)
Help desk
Individualism
Informational strategy
Institutionalization
Level 1 support
Level 2 support
Long-term orientation
Management policies
Masculinity
Measurements
Migration plan
Modular conversion
Modules
Monochronic time
On-demand training
One-on-one training
Online support
Operations group
Parallel conversion
Perceived benefi ts
Perceived costs
Phased conversion
Pilot conversion
Political strategy
Polychronic time
Post-implementation
Potential adopter
Power distance
Problem report
Project assessment
Project team review
Ready adopters
Real benefi ts
Real costs
Refreeze
Reluctant adopters
Resistant adopters
Resource allocation
Rewards
Risk
Short-term orientation
Simultaneous conversion
Speed of messages
Sponsor
Standard operating
procedure (SOP)
System maintenance
System request
System review
System support
Time
Training
Transition process
Uncertainty avoidance
Unfreeze
Whole-system
conversion
APPLYING THE CONCEPTS AT PATTERSON
SUPERSTORE
In this chapter, we see how the fi rst phase of the Integrated Health Clinic Delivery
System transitions from development into production for use by the user commu-
nity. Making this transition involved providing training to the users, including Clinic
employees and clients. Ruby and the team conducted an assessment of the develop-
ment process and each member’s contribution. Th ey also developed a plans for con-
tinued maintenance of the system.
You can fi nd the rest of the case at: www.wiley.com/go/dennis/casestudy
CHAPTER REVIEW
Aft er reading and studying this chapter, you should be able to:
Describe how Hall’s and Hofstede’s cultural dimensions can eff ect the adoption of an information system.
Describe the technical and managerial issues related to system conversion.
Discuss the three dimensions of system conversion.
Discuss why people resist and accept change.
Describe the major steps in a change management plan.
Discuss the diff erent strategies to motivate adoption of a new system.
Discuss why training is crucial to the acceptance of a new system.
Describe the diff erent post-implementation activities that take place aft er the successful deployment of the
information system.
Exercises 503
QUESTIONS
1. What are the three basic steps in managing organiza-
tional change?
2. What are the cultural issues of which developers
should be aware?
3. What are the major components of a migration plan?
4. Compare and contrast direct conversion and parallel
conversion.
5. Compare and contrast pilot conversion, phased conver-
sion, and simultaneous conversion.
6. Compare and contrast modular conversion and
whole-system conversion.
7. Explain the trade-off s among selecting between the
types of conversion in questions 4, 5, and 6.
8. What are the three key roles in any change manage-
ment initiative?
9. Why do people resist change? Explain the basic model
for understanding why people accept or resist change.
10. What are the three major elements of management
policies that must be considered when implementing
a new system?
11. Compare and contrast an information change man-
agement strategy with a political change management
strategy. Is one better than the other?
12. Explain the three categories of adopters you are likely
to encounter in any change management initiative.
13. How should you decide what items to include in your
training plan?
14. Compare and contrast three basic approaches to training.
15. What is the role of the operations group in system
development?
16. Compare and contrast two major ways of providing
system support.
17. How is a problem report diff erent from a change
request?
18. What are the major sources of change requests?
19. Why is project assessment important?
20. How is project team review diff erent from system review?
21. What do you think are three common mistakes that
novice analysts make in migrating from the as-is to
the to-be system?
22. Some experts argue that change management is more
important than any other part of system development.
Do you agree or not? Explain.
23. In our experience, change management planning
oft en receives less attention than conversion planning.
Why do you think this happens?
EXERCISES
A. Suppose you are installing a new accounting package
in your small business. What conversion strategy
would you use? Develop a conversion plan (i.e., tech-
nical aspects only).
B. Suppose you are installing a new room reservation
system for your university that tracks which courses
are assigned to which rooms. Assume that all the
rooms in each building are “owned” by one college
or department and only one person in that college or
department has permission to assign them. What con-
version strategy would you use? Develop a conversion
plan (i.e., technical aspects only).
C. Suppose you are installing a new payroll system in a
very large multinational corporation. What conver-
sion strategy would you use? Develop a conversion
plan (i.e., technical aspects only).
D. Consider a major change you have experienced in
your life (e.g., taking a new job, starting a new school).
Prepare a cost–benefi t analysis of the change in terms
of both the change and the transition to the change.
E. Suppose you are the project manager for a new library
system for your university. Th e system will improve
the way students, faculty, and staff can search for
books by enabling them to search over the Web,
rather than using only the current text-based system
available on the computer terminals in the library.
Prepare a cost–benefi t analysis of the change in terms
of both the change and the transition to the change for
the major stakeholders.
F. Prepare a plan to motivate the adoption of the system
in exercise E.
G. Prepare a training plan that includes both what you
would train and how the training would be delivered
for the system in exercise E.
H. Suppose you are leading the installation of a new DSS
to help admissions offi cers manage the admissions
process at your university. Develop a change manage-
ment plan (i.e., organizational aspects only).
I. Suppose you are the project leader for the development
of a new Web-based course registration system for your
504 Chapter 13 Installation and Operations
university that replaces an old system in which students
had to go to the coliseum at certain times and stand in
line to get permission slips for each course they wanted
to take. Develop a migration plan (including both tech-
nical conversion and change management).
J. Suppose you are the project leader for the development
of a new airline reservation system that will be used
by the airline’s in-house reservation agents. Th e sys-
tem will replace the current command-driven system
designed in the 1970s that uses terminals. Th e new
system uses PCs with a Web-based interface. Develop a
migration plan (including both conversion and change
management) for your telephone operators.
K. Develop a migration plan (including both conver-
sion and change management) for the independent
travel agencies that use the airline reservation system
described in exercise J.
L. For the A Real Estate Inc problem in Chapters 4
through 12:
1. Prepare a plan to motivate adoption of the system.
2. Prepare a training plan that includes both what you
would train and how the training would be delivered.
3. Prepare a change management plan.
4. Develop a migration plan.
M. For the A Video Store problem in Chapters 4 through 12:
1. Prepare a plan to motivate adoption of the system.
2. Prepare a training plan that includes both what
you would train and how the training would be
delivered.
3. Prepare a change management plan.
4. Develop a migration plan.
N. For the gym problem in Chapters 4 through 12:
1. Prepare a plan to motivate adoption of the system.
2. Prepare a training plan that includes both what you
would train and how the training would be delivered.
3. Prepare a change management plan.
4. Develop a migration plan.
O. For the Picnics R Us problem in Chapters 4 through
12:
1. Prepare a plan to motivate adoption of the system.
2. Prepare a training plan that includes both what you
would train and how the training would be delivered.
3. Prepare a change management plan.
4. Develop a migration plan.
P. For Of-the-Month Club problem in Chapters 4
through 12:
1. Prepare a plan to motivate adoption of the system.
2. Prepare a training plan that includes both what
you would train and how the training would be
delivered.
3. Prepare a change management plan.
4. Develop a migration plan.
MINICASES
1. Nancy is the IS department head at MOTO Inc., a human
resources management fi rm. Th e IS staff at MOTO Inc.
completed work on a new client management soft ware
system about a month ago. Nancy was impressed with
the performance of her staff on this project because the
fi rm had not previously undertaken a project of this scale
in-house. One of Nancy’s weekly tasks is to evaluate and
prioritize the change requests that have come in for the
various applications used by the fi rm.
Right now, Nancy has fi ve change requests for
the client system on her desk. One request is from a
system user who would like some formatting changes
made to a daily report produced by the system.
Another request is from a user who would like the
sequence of menu options changed on one of the
system menus to more closely refl ect the frequency of
use for those options. A third request came in from
the billing department.
Th is department performs billing through a billing
soft ware package. A major upgrade of this soft ware is
being planned, and the interface between the client sys-
tem and the bill system need to be changed to accom-
modate the new soft ware’s data structures. Th e fourth
request seems to be a system bug that occurs whenever
a client cancels a contract (a rare occurrence, fortu-
nately). Th e last request came from Susan, the com-
pany president. Th is request confi rms the rumor that
MOTO Inc. is about to acquire another new business.
Th e new business specializes in the temporary place-
ment of skilled professional and scientifi c employees
and represents a new business area for MOTO Inc.
Th e client management soft ware system will need to
be modifi ed to incorporate the special client arrange-
ments that are associated with the acquired fi rm.
How do you recommend that Nancy prioritize these
change requests for the client/management system?
Minicases 505
2. Sky View Aerial Photography off ers a wide range of
aerial photographic, video, and infrared imaging ser-
vices. Th e company has grown from its early days of
snapping pictures of client houses to its current status
as a full-service aerial image specialist. Sky View now
maintains numerous contracts with various govern-
mental agencies for aerial mapping and surveying
work. Sky View has its offi ces at the airport, where it
keeps its fl eet of specially equipped aircraft . Sky View
contracts with several freelance pilots and photogra-
phers for some of its aerial work and also employs
several full-time pilots and photographers.
Th e owners of Sky View Aerial Photography
recently contracted with a systems development
consulting fi rm to develop a new information system
for the business. As the number of contracts, aircraft ,
fl ights, pilots, and photographers increased, the com-
pany experienced diffi culty keeping accurate records
of its business activity and the utilization of its fl eet
of aircraft . Th e new system will require all pilots and
photographers to swipe an ID badge through a reader
at the beginning and conclusion of each photo fl ight,
along with recording information about the aircraft
used and the client served on that fl ight. Th ese records
are to be reconciled against the actual aircraft utili-
zation logs maintained and recorded by the hangar
personnel.
Th e offi ce staff was eagerly awaiting the installation
of the new system. Th eir general attitude was that the
system would reduce the number of problems and
errors that they encountered and would make their
work easier. Th e pilots, photographers, and hangar
staff were less enthusiastic, being unaccustomed to
having their activities monitored in this way.
a. Discuss the factors that might inhibit the acceptance
of this new system by the pilots, photographers, and
hangar staff .
b. Discuss how an informational strategy could be
used to motivate adoption of the new system at Sky
View Aerial Photography.
c. Discuss how a political strategy could be used to
motivate adoption of the new system at Sky View
Aerial Photography.
3. For the Holiday Travel Vehicles problem described in
Chapters 5 through 12:
a. Prepare a plan to motivate adoption of the system.
b. Prepare a training plan that includes both what you
would train and how the training would be delivered.
c. Prepare a change management plan.
d. Develop a migration plan.
4. For the Professional and Scientifi c Staff Management
problem described in Chapters 4, and 6 through 11:
a. Prepare a plan to motivate adoption of the system.
b. Prepare a training plan that includes both what you
would train and how the training would be delivered.
c. Prepare a change management plan.
d. Develop a migration plan.
A
Abstract class, 22, 164, 258
Abstraction, 258
in object-oriented systems, 281
Accelerometers, 397, 398
Acceptance tests, 472, 477
Access control, 361
Access control requirements, defi ned, 445
Acknowledgment messages, 386
Action–object order, 384
Actions
in activity diagrams, 131, 132, 317
in behavioral state machines, 222
Action statements, 316
Activities
in activity diagrams, 131, 132, 317
in behavioral state machines, 222
black-hole, 136
miracle, 136
Activity-based costing, 94
Activity coordination, 458
Activity diagrams, 119, 120
actions in, 131, 317
activities in, 131, 317
control fl ows in, 131, 317
control nodes in, 132–136
decision node in, 134, 317
fi nal-activity node in, 132–133, 317
fi nal-fl ow node in, 133, 317
fork node in, 135, 317
guard condition, 134
guidelines for creating, 136–137
initial node in, 132, 317
joint node in, 136, 317
merge node in, 134, 317
object fl ows in, 131, 317
object nodes in, 131, 317
steps in creating, 137–140
swimlanes in, 136, 317
syntax for, 132, 317
of UML, 316
Activity elimination, 95
Actors
average, 59
defi ned, 59, 206, 217
in interaction diagrams, 205
primary, 142
simple, 59
specialized, 122
in use-case diagrams, 121–123
Adjusted usecase points (UCP), 62
Adoption, motivating, 493–495
Aesthetics, 370–371
Aggregation association, 181–183
Aggregation relationships
in factoring process, 257
in structural modeling, 166
Agile development, 12–15
criticisms, 13
extreme programming (XP), 13–14
principles, 12
scrum, 14–15
A-kind-of
in factoring, 257
in object-oriented systems, 21
in structural model, 165
Alexander, Christopher, 170
Algorithm specifi cations, 316–319
Alpha testing, 477
Alternative fl ows, 144
Alternative matrix, 274–275, 439
Alternative requirements documentation techniques,
see Requirements documentation techniques,
alternative
Amazon.com, 408, 445
Ambler, S. W., 30, 136, 207, 218, 226
Analysis models, 4
balancing, 242–257
evolving into design models, 257–262
verifying and validating, 242–257
Analysis paralysis, 28
Analysis patterns, 294
I N D E X
507
508 Index
B
Backyard recycling technique, 431
Bar-code readers, 389
Batch processing, 387–389
Batch reports, 393
Behavior, 282
Behavioral modeling, 202–235
behavioral state machines for, 221–229
crude analysis for, 229–232
interaction diagrams for, 204–221
verifying and validating, 233–234
Behavioral models, 6, 203
balancing functional model and, 243–251
balancing structural model and, 251–254
Behavioral state machine, 221–229, 473
actions in, 222
activities in, 222
creating, 226–229
elements of, 222–226
events, 222
guidelines for creating, 226
states, 221
transitions in, 222
Behavioral view, 24
Behavior diagrams, 34
Behaviors, 204
in object-oriented systems, 20
Bellin, D., 167
Benchmark, 440
Benchmarking, informal, 94
Beta testing, 477
Bias minimization, 393–394
Binding technique, 23
Black-box testing, 471, 472
Black hole states, 226
Booch, Grady, 24, 25, 34
Bottom-up interviews, 98, 99
Brainstorming, 167–169
Breadth of information
(requirements analysis), 109
Break-even point
defi ned, 51
determination, 47, 50–51
graphing, 47
Brief description (use cases), 142
Broad and shallow menu, 384
Brynjolfsson, Erik, 429
Business analyst, 18
Business-modeling workfl ow (Unifi ed Process), 28
Business need, 43, 272
Analysis phase, 3–4. See also Systems development
life cycle (SDLC)
analysis strategy, 3
requirements gathering, 4
system proposal, 4
Analysis workfl ow (Unifi ed Process), 28–29
API, see Application program interface
Appelo, J., 76
Application logic, 419
Application program interface (API),
59, 297, 469
Application service providers (ASP), 270
Application soft ware, 420
Application system, 336–337
Approval committee, 3, 43
Architectural components, 419
Architecture-centric OOSAD, 24
Architecture design, 4
Artifact, 432, 433
As-is system, 3, 86
ASP, see Application service providers
Assemblies, 166
Association
class, 180
in communication diagram, 217
Association relationships
in structural models, 166
in use-case descriptions, 142
in use-case diagrams, 122, 123–124
Asymmetric encryption algorithm, 446
Attributes, 165, 204, 282
derived, 176, 300
multivalued, 333
in object-oriented systems, 20
private, 176
protected, 176
public, 176
visibility of, 176
Attribute sets, 333
Audit fi les, 330
Augmented reality (AR), 404
Authentication, 446
Authentication requirements, 445–446
defi ned, 445
Authentication testing, 477
Authorization testing, 477
Availability and reliability
requirements, 443–444
Average actors, 59
Avison, D., 91
Index 509
Data Access and Manipulation (DAM), 260
in factoring, 257
in object orientation, 282
in object-oriented systems, 19, 20
in structural models, 164–165
utility, 260
Class and method design, 280–321
constraints, 304–314
contracts, 304–314
design criteria, 286–293
method specifi cation, 314–319
object design activities, 293–304
object orientation, characteristics of, 282–286
verifying and validating, 319–321
Class cohesion, 290
ideal, 291
mixed-domain, 291
mixed-instance, 291
mixed-role, 291
Class diagrams
aggregation association, 181–183
class, 176–179
generalization association, 181
object diagrams, 184
relationships, 179–181
simplifying, 184
Class library, 297
Class-Responsibility-Collaboration (CRC) cards
collaborations, 173
elements of, 173–174
role-playing, with use cases, 174–175
Class–Responsibility–Collaboration (CRC) cards
responsibilities, 172
Classroom training, 496
Class tests, 473
Client, 259
Client-based architectures, 420–421
Client computers, 419
Client-server architectures, 421–422
Client-server tiers, 422–424
Closed-ended questions, 97, 98
Cloud computing, 426–428
Cloud services, 270
Clustering, 354–355
Coad, Peter, 286, 289
COBIT (Control Objectives for Information and related
Technology), 88
Coding practices, 14
Coding standards, 78
Cognitive map, 405
Business process, 426. See also specifi c types of
business process
Business process documentation, 140–152. See also
Use-case description
use cases, types of, 141
Business process identifi cation, 121–129
use-case diagram, creating, 127–129
use-case diagram, elements of, 121–126
use cases, identifying, 126–127
Business process modeling, 129–140
activity diagram creation, 137–140
activity diagrams, 131–136
guidelines for activity diagrams creation, 136–137
Business requirements, 44, 87
Business value, 44
C
CA, see Certifi cate authority
Caching computational results, 300
Capability list, 477
Capability Maturity Model compliance, 88
Capacity requirements, 443
CASE, see Computer-aided soft ware engineering
Cash fl ow
determination, 48–49
method, 49
Cassimally, Hakim, 429
Certainty of benefi ts, 494
Certainty of costs, 494
Certifi cate authority (CA), 446
Champion, 52
Change agents, 17
Change control, 458
Change management, 482, 489–497
adoption, motivating, 493–495
analyst, 19
cost and benefi ts, 492–493
management policies, revising, 491–492
resistance to change, 490–491
training, 495–497
Change management workfl ow, confi guration
and, 30, 33–34
Change request, 498, 499, 500
Check box, 391
Check digit check, 392
Class(es)
abstract, 164
in class diagrams, 176–179
concrete, 164
container, 260
510 Index
system, 5
test designing, 467–477
Construction phase (Unifi ed Process), 27
Constructor operation, 176
Container classes, 260
Content awareness, 369–370
Context, 408, 460, 483
Contract, 173, 259, 294, 304
Elements of, 306–314
fi xed-price, 271
time-and-arrangements, 271
value-added, 271
Control and security (server-based architecture), 425
Control fl ows, 131, 132, 317
Controllers, 259
Control nodes
decision node, 134
fi nal-activity node, 132–133
fi nal-fl ow node, 133
fork node, 135
initial node, 132
joint node, 136
merge node, 134
Conversion, 485–489. See also specifi c types of
conversion
location, 486–487
modules, 487–488
selecting, 488–489
style of, 486
Conversion location, 486–487
phased, 487
pilot, 487
simultaneous, 487
style of, 486
Conversion modules
modular conversion, 487–488
whole-system conversion, 487
Conversion strategy
cost, 489
risk, 488–489
time, 489
Conversion style
direct, 486
parallel, 486
Costs, 109
certainty of, 494
in conversion strategy, 489
development, 47, 48
of development, 424–425
of infrastructure, 424
Cohesion
class, 290
defi ned, 289
generalization/specialization, 291
ideal class, 290
method, 289–290
Collaborations, 173, 258–259
Collectivism, 461
Color
aesthetics, 370
cultural meanings of, 407
Columnar data stores, 334
Combo box, 391
Command language, 384
Common object lists, 169
Communication diagrams
creating, 219
elements of, 216–218
examples, 219–221
guidelines for creating, 218–219
Communication path, 433
Compatibility, 46
Completeness check, 392
Complex actors, 59
Complex systems, 16–17
Component, 297
Computer-aided soft ware engineering (CASE)
benefi ts of, 77
defi ned, 77
repository, 77
tools, 77
Computer-based training (CBT), 496, 497
Concept mapping, 110
Concept maps, 110–112
Conceptual model, 163
Concrete class, 21, 164, 258
Confi guration and change management workfl ow
(Unifi ed Process), 30, 33–34
Confi rmation message, 386
Confl ict management, 76
Connascence, 292–293
Consistency, 369, 371–372
Consistency check, 392
Constantine, L. L., 30
Constraints, 304
types of, 306
Construction, 456–477
defi ned, 456
documentation development, 462–467
programming management, 457–461
Index 511
Data entry operator, 389
Data management layer, 261
Data management layer design, 326–363
data access and manipulation classes,
designing, 357–360
mapping problem domain objects to object
persistence formats, 337–346
nonfunctional requirements and,
360–361
object persistence formats, 327–337
RDBMS-based object storage, 346–357
verifying and validating, 361–362
Data storage, 419
size of, 356–357
DBMS, see Database management system
Decision node, 132, 134, 317
Decision support systems (DSS), 336
Decomposition, 166
Default value, 390
Delay message, 386
DeMarco, T., 71
Dennis, Alan, 105
Denormalization, 351–354
Dependency
partial, 349
transitive, 351
Dependency relationship, 262, 263
Deployment diagrams, 432–434
Deployment engineering workfl ow, 27
Deployment workfl ow (Unifi ed Process), 29, 32
Depth of information (requirements analysis), 109
Derived attributes, 176, 300
Design, 240–275
acquisition strategy, selecting, 273–275
and balancing of analysis models, 242–257
classic, avoiding, 241
custom development, 268–269
evolving analysis models into design models,
257–262
optimization, 298–300
outsourcing, 270–272
packaged soft ware, 269–270
packages and package diagrams, 262–268
restructuring, 297–298
selecting, 272–273
strategies, 268–273
Design models, 29
evolving analysis models into, 257–262
packages and package diagrams, 262–268
Design patterns, 294, 295
intangible, 48
materials, 94
operational, 47, 48
in requirements analysis, 109
of transition, 494
Costs and benefi ts analysis
assigning values to, 48
in change management, 492–493
fi nancial calculations for, 51
identifying, 47–48
Coupling
defi ned, 286
inheritance, 289
interaction, 287–288
CRC cards, see Class-Responsibility-Collaboration
(CRC) cards
Critical path method (CPM), 58
Critical task, 58
Critical thinking skills, 92
CRUDE (create, read, update, delete, or execute), 126
CRUDE analysis, 229–232
CRUDE matrix, 243, 245
Cultural and political requirements
customization requirements, 447
legal requirements, 448
synopsis, 449
Cultural diff erences, 407–410
Cultural issues, 406–410
and information technology, 483–484
in programming management, 460–461
Cultural requirements, 88
Custom development, 268–269
Customization, 269
Customization requirements, 447
D
DAM classes, see Data Access and Manipulation classes
Data Access and Manipulation (DAM) classes, 260, 338
designing, 357–360
Data access logic, 419
Data access speed optimization, 351–356
clustering, 354–355
denormalization, 351–354
indexing, 355–356
Database, 327
Database and fi le specifi cations, 4
Database checks, 392
Database management system (DBMS), 327
Data capture at source, 389390
Data-centered methodology, 5
512 Index
costs and benefi ts, identifying, 47–48
net present value (NPV), 49–50
return on investment (ROI), 50
Edit checks, 391
EIS, see Executive information systems
E-JAD, see Electronic JAD
Elaboration phase (Unifi ed Process), 27
Electronic brainstorming, 169
Electronic distribution, 104
Electronic JAD, 102
Encapsulation, 282, 468
in object-oriented systems, 20–21
in testing and object orientation, 468
Enchanted objects, 429
Encryption, 445–446
defi ned, 445
End-user DBMS, 327
Engineering workfl ows. See also Workfl ows
analysis workfl ow, 28–29
business-modeling workfl ow, 28
deployment workfl ow, 29
design workfl ow, 29
implementation workfl ow, 29
requirements workfl ow, 28
testing workfl ow, 29
English-language messages, 406
Enhanced Unifi ed Process, 31, 33
Enterprise DBMS, 327
Enterprise resource planning (ERP), 269
Environmental factors (EF), 59, 62
Environmental factor value (EFactor), 59, 62
Environment and infrastructure management, 76–79.
See also Project management
CASE tools, 77
documentation, 78–79
standards, 77–78
Environment workfl ow (Unifi ed Process), 30, 32
Error(s), 153
Error correction, 153
Error message, 386
Essential use case, 141, 372
Estimates, refi ning, 69–70
Estimation, defi ned, 58
Event
in behavioral state machines, 222, 223
in method specifi cation, 314
Event driven languages, 314
Evolutionary work breakdown structures, 63–67
e-waste, 431
Exceptional fl ows, 144
Design phase. See also Systems development life
cycle (SDLC)
architecture design, 4
database and fi le specifi cations, 4
design strategy, 4
program design, 4
Design prototype, 11
Design strategy, 4
Design workfl ows, 27, 29
Destructor operation, 179
Detail report, 395
Detail use case, 141
Development
costs, 47, 48, 424–425
incremental, 24–25
iterative, 24–25
parallel, 8
phased, 9
waterfall, 7
Digital signatures, 446
Direct conversion, 486
Direct manipulation (navigation control), 385–386
Document analysis, 106–107
Documentation, 78–79
development, 462–467
procedures manuals, 463
reference documents, 463
standards, 78
topics, 463, 465
tutorials, 463
Documentation navigation controls, 463
Documentation structure designing, 463–465
Document data stores, 334
Doing responsibilities, 172
Drop-down list box, 391
Drop-down menu, 385
DSS, see Decision support systems
Duration analysis, 93–94
Dynamic binding, 468
in object-oriented systems, 22–23, 283
Dynamic model, 204
E
Ease of development, 425
Ease of learning, 371
Ease of use, 371
Economic feasibility, 46–51. See also Feasibility analysis
break-even point determination, 50–51
cash fl ow determination, 48–49
costs and benefi ts, assigning values to, 48
Index 513
Functional model
balancing behavioral models and, 243–251
balancing structural model and, 242–243
and structural model, relationships, 244
Functional modeling
interrelationships, 156
verifi cation and validation, 154–156
verifi cation and validation through walkthroughs,
154–155
Functional quality, 88
Functional requirements, 87, 90
G
Games, 400
Gamifi cation, 400–401
Gantt chart, 56–57
Generalization association, 181
Generalization relationship, 122, 144, 165, 257
Generalization/specialization cohesion, 291
Generic sequence diagram, 204
Globalization, 89
Glocalization, 407
Gradual refi nement, 3
Grammar order, consistent, 384
Graphical displays and reports, 393
Graphical user interface (GUI), 368, 425
Graphs, 395
Green data centers, 431
Green IT, 431–432
Grid computing, 426
Ground rules (JAD sessions), 103
Group cohesiveness, 76
Guard condition, 134, 217, 222
H
Hall, Edward, 407, 408, 409, 460, 483, 484
Haptic feedback, 396
Hardcoded value, 471
Hardware and operating system, 360
Hardware and soft ware specifi cation, 438–440
Hardware components, primary, 419
Has-parts
in factoring, 257
in structural model, 166
Health and Human Services Health Insurance Portability
and Accountability Act (HIPAA), 428
Help desk, 498
Help message, 386
Heuristic evaluation, 381
Execution occurrence, 206
Executive information systems (EIS), 336
Extend relationship, 122, 144
Extent, 332
External nonfunctional dimensions, 88
External trigger, 142
Extreme programming (XP), 13–14
F
Facilitator, 101, 104
Factoring, 257–258, 298
Familiarity
with functional area, 46
with technology, 46
Fan-out, 300
Fat client, 421
Faults, 153
Feasibility analysis, 3, 43, 45–53. See also Project
management
economic feasibility, 46–51
organizational feasibility, 51–53
technical feasibility, 45–46
Feminine cultures, 409
Field labels, 370
Final-activity node, 132–133, 317
Final-fl ow node, 132, 133, 317
Final state, 222, 223
Financial awards, 75
First-line supervisors, 492
First mover, 43
First normal form (1NF), 347, 349
Fitzgerald, G., 91
Fixed-price contract, 271
Flow of events
alternative or exceptional fl ows, 144
normal, 144
subfl ows, 144
in use-case description, 144–145
Foreign key, 330, 331
Fork node, 132, 135, 317
Formal usability testing, 381–382
Format check, 392
Foundation layer, 260
Frame, 209, 217, 223
Framework, 297
Frequently asked questions (FAQ), 498
Friedman, T. L., 89, 407
Functional decomposition, 144
Functionality, 44
Functional lead, 74
514 Index
Infrastructure design, 432–438
deployment diagrams, 432–434
network model, 434–438
Infrastructure management workfl ow (Unifi ed
Process), 32
Inheritance, 144, 469
confl ict, 284, 285
multiple, 285
in object orientation, 284–286
in object-oriented systems, 21–22
single, 284
Inheritance coupling, 289
In-house experience, 272
Initial node, 132, 317
Initial state, 222, 223
Input design, 387–392
basic principles, 387–390
input validation, 391–392
types of inputs, 390, 391
Input validation, 391–392
Installation process, 5, 481–501
change management, 489–497
conversion, 485–489
cultural issues in, 483–484
post-implementation activities, 497–501
Instance sequence diagrams, 204
Instantiation, 184
Institutionalization, 497
Intangible benefi ts, 47, 48
Intangible costs, 48
Intangible value, 44
Integration of information, 109
Integration testing, 468, 472
Integration tests, 475–476
Interaction, 169
Interaction coupling, 287–288
Interaction diagrams, 204–221
communication diagrams, 216–221
messages, 204
objects in, 204
operations in, 204
sequence diagrams, 204–215
Interaction testing, 472, 475
Interactive evaluation, 381
Interface actions, 377
Interface capabilities, 425
Interface design, 4
Interface design prototyping, 377–380
selecting, 379
storyboard, 377–379
History fi les, 330
Hofstede, Geert, 407, 408, 409, 460, 461, 483, 484
Holland, Ian M., 287
Hot keys, 385
Human-computer interaction layer, 261
Human-computer interaction layer design, 367–410
games, 400–402
gamifi cation, 400–402
immersive environments, 404–406
input design, 387–392
international and cultural issues, 406–410
mobile computing, 395–398
multidimensional information visualization design,
402–404
navigation design, 383–387
nonfunctional requirements and, 410
output design, 392–395
social media and, 398–400
user interface design, 368–372, 395–398, 400–402
user interface design process, 372–383
Hybrid clouds, 426
I
Ideal class cohesion, 290
Image map, 385
Immersive environments, 404–406
Impedance mismatch, 336
Implementation phase, 4–5. See also Systems
development life cycle (SDLC)
construction, 5
installation, 5
support plan, 5
Implementation workfl ow (Unifi ed Process), 29
Importance level (use-cases), 142
Inception phase (Unifi ed Process), 26–27
Incidents, 169
Include relationship, 122, 144
Incremental development, 24–25
Indexing, 355–356
Individualism, 461
versus collectivism, 409, 484
Informal benchmarking, 94, 147
Informational strategy, 493
Information hiding, 282
in object-oriented systems, 20–21
in testing and object orientation, 468
Information load, 393
Infrastructure analyst, 19
Infrastructure as a Service (IaaS), 427
Infrastructure cost, 424
Index 515
conducting session, 103–104
designing, 103
electronic, 102
ground rules, 103
participant selection, 102–103
post-session report, 104
preparing for session, 103
problem management in, 105
for RAD-based methodologies, 9
Joint node, 136
Jones, Capers, 101
K
Karner, Gustav, 58
Keystrokes minimization, 390
Key-value data stores, 334
KISS principle, 13, 146
Knowing responsibilities, 172
Krug, Steve, 382, 383, 396, 408
L
Languages (navigation control), 384
Larman, C., 15, 147
Law of Demeter, 287, 288
Layers, 259–262. See also Design
data management, 260–261
foundation, 260
human–computer interaction, 261
physical architecture, 261–262
problem domain, 260
Layout (user interface design), 369
Legal requirements, 448
Lencioni, P., 72
Lewin, Kurt, 482
Lieberherr, Karl J., 287
Lifeline, 205, 206
Linked list, 328
Lister, T., 71
Load tests, 477
Local area network (LAN), 420
Locations, 435
Logical models, 120
Long-versus short-term orientation, 461, 484
Lookup fi les, 328
M
Magnetic stripe readers, 389
Maintainability requirements, 441, 442
Maintenance oracle, 154
user interface prototypes, 379
windows layout diagram, 377
Interface evaluation, 380–382
formal usability testing, 381–382
heuristic evaluation, 381
interactive evaluation, 381
walkthrough evaluation, 381
Interface evaluation, 380–382
Interface icons, 377
Interface metaphor, 376
Interface objects, 376–377
Interface standards design, 376–377
Interface templates, 376
Interfi le clustering, 354
Internal nonfunctional dimensions, 88
International issues, 406–410
Internet of Th ings (IoT), 428–431
Interpersonal skill, 74
development of, 102
Interrelationships
behavioral models, 234
functional modeling, 156
Interview(s), 96–100
bottom-up, 98, 99
closed-ended questions, 97
conducting, 99
open-ended questions, 97
post-interview follow-up, 100
preparing for, 99
probing question, 97
as requirements-gathering technique, 96–100, 101
schedule, 96, 97
structured, 98
top-down, 98
unstructured, 98
Interview notes, 100
Interview report, 100, 101
Intrafi le clustering, 354
Invariants, 306
class diagram, 308
on CRC card, 307
in text fi le, 308
Iterative development, 24–25
Iterative workplans, 63–67
J
Jacobson, Ivar, 24, 25, 34
Jelled team, 71–72
Join node, 132, 317
Joint application development (JAD), 100–104, 105
516 Index
Middleware, 421
Migration plan, 482
Milestones, project, 55, 57
Miracle states, 226
Mission-critical systems, 445
agile for, 13
and Scrum, 15
XP for, 14
Mistakes
implementation, 459
preventing, 383
recovery from, 383
Mobile computing, 395–398
Mobile devices, 396, 439
Model–View–Controller (MVC)
architecture, 259
Modular conversion, 487–488
Module, 257
Monochronic time, 408, 461, 484
Motivation, 75–76
Multidimensional information visualization
design, 402–404
Multilingual requirements, 406–407
Multiple inheritance, 285
Multiple layout, 369
Multiplicity, 180
Multitenancy, 426
Multivalued attributes, 333
MVC architecture, see Model-View-Controller
(MVC) architecture
N
Narrow and deep menu, 384
Natural language, 384
Navigation controls
consistency in, 371
direct manipulation, 385–386
languages, 384
menus, 384–385
Navigation design, 383–387
basic principles, 383–384
documentation, 387
grammar order, consistent, 384
messages, 386–387
preventing mistakes, 383
recovery from mistakes, 383
types of controls, 384–386
Navigation terms identifi cation, 465–467
Net present value (NPV), 47, 49–50
defi ned, 51
Management information systems (MIS), 336
Management policies, 491–492
Manual systems, 89
Masculinity versus femininity, 409, 484
Master fi les, 328
Materials costs, 94
McAfee, Andrew, 429
McEwen, Adrian, 429
Measurements, 492
Media, 394–395
Meeting, scrum, 15
Menu bar, 385
Menus (navigation control), 384–385
Merge node, 132, 134, 317
Message passing, 315
Messages
defi ned, 206, 217
in interaction diagrams, 204
navigation design, 386–387
in object orientation, 282
in object-oriented systems, 20
Method(s), 165, 204, 257, 259
in object orientation, 282
in object-oriented systems, 20
Method cohesion, 289–290
classical, 290
coincidental, 290
communicational, 290
functional, 290
logical, 290
procedural, 290
sequential, 290
temporal, 290
Methodology(-ies), 5–17
agile development, 12–15
criteria for selecting, 15–17
data-centered, 5
defi ned, 5
object-oriented, 5
process-centered, 5
rapid application development (RAD), 8–12
sequencing of SDLC phases, 5
structured design, 6–8
Method specifi cation, 314–319
algorithm specifi cations, 316–319
events, 314
general information, 314
message passing, 315
Meyers, Glenford, 290
Middle managers, 492
Index 517
patterns for, 169–172
in structural modeling, 166–172
textual analysis for, 166–167
Object Management Group (OMG), 34, 119
Object nodes, 131, 132, 317
Object orientation
classes, 282
dynamic binding, 283
encapsulation, 282
information hiding, 282
inheritance, 284–286
messages, 282
methods, 282
objects, 282
polymorphism, 282–284
and testing, 468–469
Object-oriented database, 332–333, 335
Object-oriented database management systems
(OODBMS), 332
mapping problem domain objects to, 338–341
Object-oriented development process
and products, 469
Object-oriented methodology, 5
Object-oriented programming language (OOPL), 333
Object-oriented systems
attributes in, 20
behaviors in, 20
classes in, 19, 20
dynamic binding in, 22–23
encapsulation in, 20–21
information hiding in, 20–21
inheritance in, 21–22
messages in, 20
methods in, 20
objects in, 19, 20
polymorphism in, 22
Object-oriented systems analysis and design
(OOSAD), 23–25
architecture-centric, 24
benefi ts of, 25
incremental development, 24–25
iterative, 24–25
use-case driven, 24
Object persistence formats
application system, type of, 336–337
criteria for fi les, 337
data types supported, 336
future needs, 337
mapping problem domain objects to, 337–346
NoSQL data stores, 333–334, 335
Network, 419
Network diagram, 57–58
Network model, 434–438
Node, 58, 432, 433
Nonfunctional requirements, 87, 88
cultural, 88, 90
cultural and political requirements, 447–448
and data management layer design, 360–361
and human-computer interaction layer
design, 410
operational, 88, 90
operational requirements, 441–442
performance, 88, 90
performance requirements, 442–444
and physical architecture layer design, 440–449
political, 88, 90
security, 88, 90
security requirements, 444–447
synopsis, 448–449
Normal fl ow of events, 144
Normalization process, 298, 347
NoSQL data stores, 333–334, 335
n-tiered architecture, 422
advantage of, 423
disadvantage of, 424
Null values, 347
Number box, 390
O
Object(s)
defi ned, 206, 217
in interaction diagrams, 204, 205
in object orientation, 282
in object-oriented systems, 19, 20
temporary, 205
Object–action order, 384
Object-based language, 301–303
Object Constraint Language (OCL), 304, 305
Object design activities, 293–304
adding specifi cations, 293
mapping problem-domain classes to implementation
languages, 300–304
opportunities for reuse, 294–297
optimizing design, 298–300
restructuring design, 297–298
Object diagrams, 184
Object fl ows, 131, 132, 317
Object identifi cation
brainstorming for, 167–169
common object lists for, 169
518 Index
technical environment requirements, 441–442
Operation call messages, 207
Operations and support workfl ow (Unifi ed Process), 32
Optical character recognition, 389
ORDBMS, see Object-relational database management
systems
Ordered sequential access fi les, 328
Organizational feasibility, 51–53
Organizational management, 52
Outcome analysis, 95
Output design, 392–395
basic principles, 392–394
media, 394–395
types of output, 394
Outsourcing, 270–272
Overview information, 142
Overview use case, 141
P
Package(s)
in class diagrams, 184
communication diagram, 218
in design model, 262–268
in use-case diagram, 127
Package diagrams
creating, 266
dependency relationship in, 262, 263
in design model, 262–268
guidelines for creating, 264–265
syntax for, 263
verifi cation and validation of, 266–268
Packaged soft ware, 269–270
Page-Jones, Meilir, 291
Paper-based documentation, 462
Paperless offi ce, 432
Parallel conversion, 486
Parallel development, 8
Parallelization, process, 94
Parkinson’s Law, 440
Partial dependency, 349
Partitions, 258–259
Patterns, 294
for object identifi cation, 169–172
Perceived benefi ts, 491
Perceived costs, 491
Pereira, Arun, 407, 408
Performance requirements, 88, 360, 410
availability and reliability requirements, 443–444
capacity requirements, 443
speed requirements, 442, 443
synopsis, 448
Object persistence formats (continued)
object-oriented database, 332–333, 335
object-relational databases, 332, 335
random access fi les, 328, 335
relational database, 330–332
selecting, 335–337
sequential access fi les, 327, 335
storage formats, existing, 337
strengths of fi les, 335
weaknesses of fi les, 335–336
Object recognition, 404
Object-relational database management systems
(ORDBMS), 332
mapping problem domain objects to, 341–344
using DAM classes, 358
Object-relational databases, 332, 335
Object storage optimization, RDBMS-based, 346–357
data access speed, optimizing, 351–356
data storage size, 356–357
storage effi ciency, optimizing, 347–351
Object wrapper, 270
Observation, 108
Occlusion, 403
OCL, see Object Constraint Language
OMG, see Object Management Group
On-demand training, 498
One-on-one training, 496
Online documentation, 463
Online support, 498
Online versus batch processing, 387–389
On-screen list box, 391
OODBMS, see Object-oriented database management
systems
OOPL, see Object-oriented programming language
OOSAD, see Object-oriented systems analysis and design
Open-ended questions, 97, 98
OPEN process (Object-oriented Process, Environment,
and Notation), 31
Operating system, 438
Operation, 165, 204
constructor, 176
destructor, 176
query, 176
update, 176
Operational costs, 47, 48
Operational requirements, 88, 360, 410
maintainability requirements, 441, 442
portability requirements, 441, 442
synopsis, 448
system integration requirements, 441, 442
Index 519
Post-session report (JAD session), 104
Potential adopters, 490
Power distance, 408, 484
Precondition, 306
Presentation logic, 419
Presenters, 153
Present value (PV), defi ned, 51
Primary actor, 142
Primary insurance carrier, 179
Primary key, 330, 349
Private attribute, 176
Private clouds, 426
Probing question, 97, 98
Problem analysis, 92
Problem-domain classes to implementation languages,
mapping, 300–304
in object-based language, 301–303
in single-inheritance language, 301
in traditional language, 304
Problem domain layer, 260
Problem domain models, 120
Problem domain objects to object persistence
formats, 337–346
to OODBMS format, mapping, 338–341
to ORDBMS format, mapping, 341–344
to RDBMS format, mapping, 344–346
Problem management (JAD sessions), 105
Problem report, 498
Procedural standards, 78
Procedures manuals, 463
Process-centered methodology, 5
Process integration, 94
Process models, 120
Process parallelization, 94
Production phase (Unifi ed Process), 31–32
Program design, 4
Program Evaluation and Review Technique
(PERT), 57–58
Program log, 458
Programmers, 18, 457–461
Programming management
activity coordination, 458
cultural issues, 460–461
programmers, assigning, 457–458
schedule management, 458–459
Project, 42
Project assessment, 497
project team review, 500–501
system review, 501
Project binder, 78
Performance testing, 477
Person, 164, 165
Person-hours multiplier (PHM), 63
Phase(s)
construction, 27
elaboration, 27
inception, 26–27
production, 31–32
transition, 27–28
of Unifi ed Process, 26–28
Phased conversion, 487, 489
Phased development, 9, 10
Physical architecture layer, 261–262
Physical architecture layer design, 418–449
architectural components, 419
client-based architectures, 420–421
client–server architectures, 421–422
client–server tiers, 422–424
cloud computing, 426–428
Green IT, 431–432
hardware and soft ware specifi cation, 438–440
infrastructure design, 432–438
Internet of Th ings (IoT), 428–431
nonfunctional requirements and, 440–449
selecting, 424–425
server-based architectures, 420
ubiquitous computing, 428
verifying and validating, 449
Physical models, 120
Pilot conversion, 487, 489
Pink, D. H., 74, 76
Planning phase. See also Systems development life
cycle (SDLC)
project initiation, 3
project management, 3
Platform as a Service (PaaS), 427
Pointer, 328
Political and cultural requirements, 361
Political requirements, 88
Political strategy, 494, 495
Polychronic time, 408, 461, 483
Polymorphism, 22, 282–284, 468
Pop-up menu, 385
Portability requirements, 441, 442
Portfolio management, 53
Postcondition, 306
Post-implementation activities, 482
project assessment, 500–501
system maintenance, 498–500
system support, 497–498
520 Index
phased development, 9
prototyping, 9–11
throwaway prototyping, 11–12
Rational Soft ware, 34
Raw data, 356
RDBMS, see Relational database management systems
Ready adopters, 495
Real benefi ts, 491
Real costs, 491
Real-time reports, 393
Real use case, 141, 373, 387
Recorders, 153
Redefi nition, 284, 285
Reference documents, 463
Referential integrity, 330, 331
Refi nement
in factoring, 258
Regular meetings, 458
Reich, Robert, 74
Relational database, 330–332
Relational database management systems (RDBMS), 330
data access speed, optimizing, 351–356
data storage size, 356–357
mapping problem domain objects to, 344–346
referential integrity, referencing, 330
storage effi ciency, optimizing, 347–351
Relationships
aggregation, 166
association, 142, 166
extend, 144
generalization, 144, 165–166
include, 144
sets, 333
Reliability, system, 17
Reluctant adopters, 495
Repeating groups (fi elds), 333
Reporting structure, 73
Report usage, 393
Request for information (RFI), 274
Request for proposal (RFP), 274
Request for quote (RFQ), 274
Requirements. See also specifi c types of requirements
business, 44, 87
functional, 87
gathering, 4
system, 87
Requirements analysis strategies, 92–95
activity-based costing, 94
activity elimination, 95
duration analysis, 93–94
Project charter, 76
Project eff ort estimation, 58–63
Project identifi cation, 43–45
system request, 44
Project initiation, 3
Project management, 3, 41–80, 42, 273
environment and infrastructure management, 76–79
feasibility analysis in, 45–53
project eff ort estimation, 58–63
project identifi cation in, 43–45
project selection, 53–54
staffi ng in, 71–76
traditional tools for, 54–58
workplan, creating/managing, 63–71
Project management tools, traditional, 54–58
Gantt chart, 56–57
network diagram, 57–58
work breakdown structure (WBS), 55–56
Project management workfl ow (Unifi ed Process), 29–30, 33
Project manager, 3, 19, 42
Project plan, 3
Project size, 46
Project skills, 272–273
Project sponsor, 3, 42, 43
Project team(s), 274
Project team review, 500–501
Protected attribute, 176
Prototyping, 9–11
throwaway, 11–12
Public attribute, 176
Public clouds, 426
Public key, 446
Public key infrastructure (PKI), 446
Pull approaches (social media), 399
Push approaches (social media), 399
Q
Query operation, 178
Questionnaires, 104–106
administration of, 106
designing, 105
participants selection, 104–105
R
RAD, see Rapid application development
Radio button, 391
Random access fi les, 328, 335
Range check, 392
Rapid application development (RAD), 8–12
Index 521
S
Sans serif fonts, 370
Sarbanes-Oxley Act, 88, 428
Scalability, 421, 424, 425
Scenario, 141
Schedlbauer, Martin, 130
Schedule
adjusting for missed dates, 70
management, 458–459
short time, 17
visibility, 17
Schell, Jesse, 402
Scope creep, 67, 459
Scope management, 67–68
Scribes, 101, 153
Scrum, 14–15
Second normal form (2NF), 349, 351
Security requirements, 88, 361, 410
access control requirements, 445
authentication requirements, 445–446
encryption, 445–446
synopsis, 449
system value, 444–445
virus control requirements, 447
Security testing, 477
Selection box, 390
Self-delegation, 207
Sequence diagrams
creating, 209–210
elements of, 204–207
examples, 210–215
guidelines for creating, 207–209
Sequential access fi le, 327, 335
ordered, 328
unordered, 328
Serif fonts, 370
Server-based architectures, 420
Server object, 173
Servers, 259, 419
Service-oriented architectures, 426
Short time schedules, 17
Signature of method, 294
Simone, S. S., 167
Simple actors, 59
Simultaneous conversion, 487
Singh, Nitish, 407, 408
Single inheritance, 284
Single inheritance language, 301, 302
Slider, 391
Smalltalk, 259, 260
informal benchmarking, 94
outcome analysis, 95
problem analysis, 92
root cause analysis, 92–93
technology analysis, 95
Requirements determination, 86–91
defi ning requirement, 87–89
determining requirements, 89–91
purpose, 87
real-world problems with, 91
requirements defi nition creation, 91
requirements defi nition report, 89
Requirements documentation techniques, alternative
concept maps, 110–112
user stories, 112
Requirements-gathering techniques, 95–110
combining, 109–110
document analysis, 106–107
interviews, 96–100, 101
joint application development (JAD), 100–104
observation, 108
questionnaires, 104–106
selection of, 108–110
Requirements workfl ow (Unifi ed Process), 28
Resistance to change, 490–491
Resistant adopters, 495
Resource allocation, 492
Responsibilities
CRC cards, 172
doing, 172
knowing, 172
Return message, 207
Return on investment (ROI), 47, 50
defi ned, 51
Reuse, 469
Rewards, 492
RFI, see Request for information
RFP, see Request for proposal
RFQ, see Request for quote
Risk
assessment, 70, 71
in conversion strategy, 488–489
management, 70–71
Role-playing CRC cards, 110, 141
with use cases, 174–175
Root cause analysis, 92–93
Rose, David, 430
Round-robin approach, 169
Round-trip engineering, 77
Rumbaugh, James, 24, 25, 34
522 Index
Structural model, 163, 164
balancing behavioral model and, 251–254
balancing functional models and, 242–243
and functional model, relationships, 244
Structural modeling, 163–197
attributes, 165
class diagrams, 176–185
classes, 164–165
CRC cards, 172–175
creating, 185–194
object identifi cation, 166–172
operations, 164–165
primary purposes of, 164
relationships, 165–166
verifying and validating, 194–197
Structured design, 6–8
parallel development, 8
waterfall development, 7
Structured English, 316
Structure diagrams (UML), 34
Structured interviews, 98
Structured query language (SQL), 332, 419
Stubs, 471
Subclass
in behavioral state machines, 225
in generalization relationships, 165
in object-oriented systems, 21
Subfl ows, 144
Subject boundary, 122, 125–126
Subject–Verb–Direct-Object–Preposition–Indirect object
(SVDPI), 186
Submenus, 384
Substitutability, 166
Summary report, 395
Superclass
generalization relationships, 165
in object-oriented systems, 21
Supporting workfl ows. See also Workfl ows
confi guration and change management
workfl ow, 30
environment workfl ow, 30
project management workfl ow, 29–30
Support plan, 5
Swimlane, 132, 136, 317
Symmetric encryption algorithm, 446
Synopsis
cultural and political requirements, 449
operational requirements, 448
performance requirements, 448
security requirements, 449
Smart cards, 389
Snyder, Alan, 289
Social media, 398–400
Social networking platforms, 261
Soft ware as a Service (SaaS), 427
Soft ware quality, 88
Soft ware testing, 467
SOP, see Standard operating procedures
Source data automation, 389
Space, 370
Special issues, 44
Specialized actor, 122
Specifi cation requirement standards, 78
Speed of messages, 408, 461, 483
Speed requirements, 442, 443
Sponsors, 489, 490
SQL, see Structured query language
Staffi ng, 71–76. See also Project management
confl ict management, 76
jelled team, 71–72
motivation, 75–76
staffi ng plan, 73–74
Stakeholder analysis, 52
Stakeholders, 142
Standard operating procedures (SOP), 492
Standards
coding, 78
documentation, 78
environment and infrastructure management, 77–78
procedural, 78
specifi cation requirement, 78
user interface design, 78
State, 221
black hole, 226
defi ned, 222, 223
fi nal, 222
initial, 222
miracle, 226
State symbol, 222
Static binding, 23
Static model, 176
Static structure diagram, 184
Steering committee, 3
Stereotype, 375, 432
Storage effi ciency, optimizing, 347–351
Storage formats, 337
Storyboard, 377–379
Story cards, 112
Strategic alignment, 52
Stress tests, 477
Index 523
Technical feasibility, 45–46
Technical lead, 74
Technical risk analysis, 46
Technical skills, 74
Technical writer, 18
Technology
analysis, 95
familiarity with, 16
Temporal trigger, 142
Temporary object, 205
Testing workfl ow (Unifi ed Process), 29
Test planning, 469–471
Tests, designing, 467–477
acceptance tests, 477
integration tests, 475–476
and object orientation, 468–469
system tests, 476–477
test planning, 469–471
unit tests, 471–475
Test specifi cations, 471
Test workfl ow (Unifi ed Process), 32
Text box, 390
Textual analysis, 166–167
Th ick client, 421
Th in client, 421
Th ird normal form (3NF), 351
Th ree-tiered architecture, 422
Th rowaway prototyping, 11–12, 110
Tidwell, Jenifer, 396, 397
Time-and-arrangements contract, 271
Timeboxing, 68–69
Time dimension, 461
Time frame, 273
Time in conversion strategy, 489
Timesharing, 427
To-be system, 3, 86
Tool bar, 385
Top-down interviews, 98, 99
Total cost of ownership, 422
Touchscreen, 397
Traceability of artifacts, 457
Trade-off s, 54
Traditional language, 304
Training (Change management), 495–497
Training plan, 5
Transaction fi les, 328
Transaction processing, 336, 387
Transition phase (Unifi ed Process), 27–28
Transition process, 222, 223, 375, 491
Transitive dependency, 351
System complexity, 16–17
System documentation, 462
System integration requirements, 441, 442
System interface testing, 475
System maintenance, 497, 498–500
System proposal, 4, 113
System reliability, 17
System request, 3, 42, 44, 45, 499
System requirements, 87
System review, 501
Systems analyst, 18
business analyst, 18
change management analyst, 19
infrastructure analyst, 18–19
primary objective of, 2
project manager, 19
roles and skills, 17–19
Systems development life cycle (SDLC), 2–5
analysis phase, 3–4
defi ned, 1
design phase, 4
implementation phase, 4–5
planning phase, 3
Systems integration, 270
System specifi cation, 4
System support, 497–498
System tests, 472, 476–477
System users, 52, 53
System value, 444–445
System value estimates, defi ned, 445
T
Table scan, 354
Tab menu, 385
Tangible benefi ts, 47, 48
Tangible value, 44
Task, 54
Task information, 54
Task lists, 112
Teams
autonomy for, 76
complexity with, 73
dysfunctional, 72
jelled, 72
leaders of, 76
scrum, 15
Technical complexity factors (TCF), 59, 62
Technical environment requirements,
441–442
Technical factor value (TFactor), 59, 62
524 Index
overview information in, 142
process documentation with, 140–152
relationships, 142–144
Use-case diagrams, 121
actors, 121–123
association, 123–124
creating, 127–129
subject boundary, 125–126
use case in, 124
Use-case driven OOSAD, 24
Use-case ID number, 142
Use-case name, 142
Use-case point estimation
for appointment system, 61
worksheet, 60
Use-case points, 58
Use-case type, 142
User documentation, 462
User eff ort minimization, 372
User experience, 371, 401
User interface design. See also Human-computer
interaction layer design
aesthetics, 370–371
consistency, 371–372
content awareness, 369–370
games/gamifi cation and, 400–402
and immersive environments, 404–406
international and cultural issues and, 406–410
layout, 369
mobile computing and, 395–398
principles for, 368–372
social media and, 398–400
user eff ort, minimizing, 372
user experience, 371
User interface design process, 372–383
common sense approach to, 382–383
interface design prototyping, 377–380
interface evaluation, 380–382
interface standards design, 376–377
navigation structure design, 375–376
use scenario development, 373–374
User interface design standards, 78
User interface prototypes, 379
User interface testing, 472, 475
User involvement, 109
User participation, 53
User requirements, clarity of, 16
User stories, 112
Use scenarios, 372
Utility classes, 260
Trigger
external, 142
temporal, 142
Turnaround document, 395
Tutorials, 463
Two-tiered architecture, 422
Type of information
(requirements analysis), 108–109
U
Ubiquitous computing, 428–431
UML, see Unifi ed Modeling Language
Unadjusted Actor Weight Total (UAW), 59
Unadjusted use-case points (UUCP), 59
Unadjusted use-case weight total (UUCW), 59
Uncertainty avoidance, 409, 484
Unifi ed Modeling Language (UML), 34–36, 119
objective of, 34
UML 2.5 diagram summary, 35
Unifi ed Process, 25–34
documentation in, 79
enhanced, 31, 33, 64
extensions to, 30–34
phases, 26–28
workfl ows, 28–30
Unit tests, 471–475
Unordered sequential access fi le, 328
Unstructured interviews, 98
Update anomaly, 347
Update operation, 179
Usability testing, 476
Use case, 24, 120
behavioral models, 203
complex, 59
defi ned, 59, 120
detail, 141
essential, 141
identifying, 126–127
overview, 141
real, 141
role-playing CRC cards with, 174–175
simple, 59
testing, 472, 475
types of, 141
in use-case diagrams, 122, 124
Use-case description, 121
creating, 146–152
elements of, 141–145
fl ow of events, 144
guidelines for creating, 145–146
Index 525
Waterfall development, 7
WBS, see Work breakdown structure
Web services, 270, 426
White-box testing, 472, 473
White space, 370
Wholes, 166
Whole-system conversion, 487
Windows layout diagram, 372, 377
Windows navigation diagram (WND), 372, 375
Workaround, 270
Work breakdown structure (WBS), 55–56
evolutionary, 63–67
Workfl ow modifi cations and extensions
confi guration and change management workfl ow,
33–34
deployment workfl ow, 32
environment workfl ow, 32
project management workfl ow, 33
test workfl ow, 32
Workfl ows
engineering, 28–29
supporting, 29–30
in Unifi ed Process, 28–30
Workplan creation and management, 63–71. See also
Project management
estimates, refi ning, 69–70
evolutionary work breakdown structures, 63–67
iterative workplans, 63–67
risk management, 70–71
scope management, 67–68
timeboxing, 68–69
X
XP, see Extreme programming
Y
Yourdon, Edward, 154, 286, 289
Yo-yo problem, 476
V
Validation
of analysis models, 242–257
of behavioral models, 233–234
of class and method design, 319–321
of data management layer design, 361–362
of functional modeling, 154–156
of package diagrams, 266–268
of physical architecture layer design, 449
of structural modeling, 194–197
Validation of input, 391–392
Value-added contract, 271
Verifi cation
of analysis models, 242–257
of behavioral models, 233–234
of class and method design, 319–321
of data management layer design, 361–362
of functional modeling, 154–156
of package diagrams, 266–268
of physical architecture layer design, 449
of structural modeling, 194–197
Version 2.5 (UML), 34, 35
Virtualization, 426
Virtual memory, 426
Virtual reality (VR), 404
Virus control, 477
Virus control requirements, 447
defi ned, 445
Visibility
of attribute, 176
of methods, 282
schedule, 17
Visualization, 147
Volume tests, 477
Volumetrics, 356
W
Walkthrough, 108
evaluation, 381
verifi cation and validation through, 154–155
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.
http://www.wiley.com/go/eula
Cover
Title Page
Copyright Page
Preface
Acknowledgments
Contents
Chapter 1 Introduction to Systems Analysis and Design������������������������������������������������������������
Introduction�������������������
The Systems Development Life Cycle�����������������������������������������
Planning���������������
Analysis���������������
Design�������������
Implementation���������������������
Systems Development Methodologies����������������������������������������
Structured Design������������������������
Rapid Application Development (RAD)������������������������������������������
Agile Development������������������������
Selecting the Appropriate Development Methodology��������������������������������������������������������
Typical Systems Analyst Roles and Skills�����������������������������������������������
Business Analyst�����������������������
Systems Analyst����������������������
Infrastructure Analyst�����������������������������
Change Management Analyst��������������������������������
Project Manager����������������������
Basic Characteristics of Object-Oriented Systems�������������������������������������������������������
Classes and Objects��������������������������
Methods and Messages���������������������������
Encapsulation and Information Hiding�������������������������������������������
Inheritance������������������
Polymorphism and Dynamic Binding���������������������������������������
Object-Oriented Systems Analysis and Design (OOSAD)����������������������������������������������������������
Use-Case Driven����������������������
Architecture-Centric���������������������������
Iterative and Incremental��������������������������������
Benefits of Object-Oriented Systems Analysis and Design��������������������������������������������������������������
The Unified Process��������������������������
Phases�������������
Workflows����������������
Extensions to the Unified Process����������������������������������������
The Unified Modeling Language������������������������������������
Applying the Concepts at Patterson Superstore
Chapter Review���������������������
Chapter 2 Project Management�����������������������������������
Introduction�������������������
Project Identification�����������������������������
System Request���������������������
Feasibility Analysis���������������������������
Technical Feasibility����������������������������
Economic Feasibility���������������������������
Organizational Feasibility���������������������������������
Project Selection������������������������
Traditional Project Management Tools�������������������������������������������
Work Breakdown Structures��������������������������������
Gantt Chart������������������
Network Diagram����������������������
Project Effort Estimation��������������������������������
Creating and Managing the Workplan�����������������������������������������
Evolutionary Work Breakdown Structures and Iterative Workplans���������������������������������������������������������������������
Managing Scope���������������������
Timeboxing�����������������
Refining Estimates�������������������������
Managing Risk��������������������
Staffing the Project���������������������������
Characteristics of a Jelled Team���������������������������������������
Staffing Plan��������������������
Motivation�����������������
Handling Conflict������������������������
Environment and Infrastructure Management������������������������������������������������
CASE Tools�����������������
Standards����������������
Documentation��������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
PART ONE ANALYSIS MODELING���������������������������������
Chapter 3 Requirements Determination�������������������������������������������
Introduction�������������������
Requirements Determination���������������������������������
Defining a Requirement�����������������������������
Requirements Definition������������������������������
Determining Requirements�������������������������������
Creating a Requirements Definition�����������������������������������������
Real-World Problems with Requirements Determination����������������������������������������������������������
Requirements Analysis Strategies���������������������������������������
Problem Analysis�����������������������
Root Cause Analysis��������������������������
Duration Analysis������������������������
Activity-Based Costing�����������������������������
Informal Benchmarking����������������������������
Outcome Analysis�����������������������
Technology Analysis��������������������������
Activity Elimination���������������������������
Requirements-Gathering Techniques����������������������������������������
Interviews�����������������
Joint Application Development (JAD)������������������������������������������
Questionnaires���������������������
Document Analysis������������������������
Observation������������������
Selecting the Appropriate Techniques�������������������������������������������
Alternative Requirements Documentation Techniques��������������������������������������������������������
Concept Maps�������������������
User Stories�������������������
The System Proposal��������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter review���������������������
Chapter 4 Business Process and Functional Modeling���������������������������������������������������������
Introduction�������������������
Business Process Identification with Use Cases and Use-Case Diagrams���������������������������������������������������������������������������
Elements of Use-Case Diagrams������������������������������������
Identifying the Major Use Cases��������������������������������������
Creating a Use-Case Diagram����������������������������������
Business Process Modeling with Activity Diagrams�������������������������������������������������������
Elements of an Activity Diagram��������������������������������������
Guidelines for Creating Activity Diagrams������������������������������������������������
Creating Activity Diagrams���������������������������������
Business Process Documentation with Use Cases and Use-Case Descriptions������������������������������������������������������������������������������
Types of Use Cases�������������������������
Elements of a Use-Case Description�����������������������������������������
Guidelines for Creating Use-Case Descriptions����������������������������������������������������
Creating Use Case Descriptions�������������������������������������
Verifying and Validating the Business Processes and Functional Models����������������������������������������������������������������������������
Verification and Validation through Walkthroughs�������������������������������������������������������
Functional Model Verification and Validation���������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Chapter 5 Structural Modeling������������������������������������
Introduction�������������������
Structural Models������������������������
Classes, Attributes, and Operations������������������������������������������
Relationships��������������������
Object Identification����������������������������
Textual Analysis�����������������������
Brainstorming��������������������
Common Object Lists��������������������������
Patterns���������������
Crc Cards����������������
Responsibilities and Collaborations������������������������������������������
Elements of a CRC Card�����������������������������
Role-Playing CRC Cards with Use Cases��������������������������������������������
Class Diagrams���������������������
Elements of a Class Diagram����������������������������������
Simplifying Class Diagrams���������������������������������
Object Diagrams����������������������
Creating Structural Models Using CRC Cards and Class Diagrams��������������������������������������������������������������������
Campus Housing Example�����������������������������
Library Example����������������������
Verifying and Validating the Structural Model����������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Chapter 6 Behavioral Modeling������������������������������������
Introduction�������������������
Behavioral Models������������������������
Interaction Diagrams���������������������������
Objects, Operations, and Messages����������������������������������������
Sequence Diagrams������������������������
Communication Diagrams�����������������������������
Behavioral State Machines��������������������������������
States, Events, Transitions, Actions, and Activities�����������������������������������������������������������
Elements of a Behavioral State Machine���������������������������������������������
Creating a Behavioral State Machine������������������������������������������
Crude Analysis���������������������
Verifying and Validating the Behavioral Model����������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
PART TWO DESIGN MODELING�������������������������������
Chapter 7 Moving on to Design������������������������������������
Introduction�������������������
Verifying and Validating the Analysis Models���������������������������������������������������
Balancing Functional and Structural Models�������������������������������������������������
Balancing Functional and Behavioral Models�������������������������������������������������
Balancing Structural and Behavioral Models�������������������������������������������������
Summary��������������
Evolving the Analysis Models into Design Models������������������������������������������������������
Factoring����������������
Partitions and Collaborations������������������������������������
Layers�������������
Packages and Package Diagrams������������������������������������
Guidelines for Creating Package Diagrams�����������������������������������������������
Creating Package Diagrams��������������������������������
Verifying and Validating Package Diagrams������������������������������������������������
Design Strategies������������������������
Custom Development�������������������������
Packaged Software
Outsourcing������������������
Selecting a Design Strategy����������������������������������
Selecting an Acquisition Strategy����������������������������������������
Alternative Matrix�������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Chapter 8 Class and Method Design����������������������������������������
Introduction�������������������
Review of the Basic Characteristics of Object Orientation����������������������������������������������������������������
Classes, Objects, Methods, and Messages����������������������������������������������
Encapsulation and Information Hiding�������������������������������������������
Polymorphism and Dynamic Binding���������������������������������������
Inheritance������������������
Design Criteria����������������������
Coupling���������������
Cohesion���������������
Connascence������������������
Object Design Activities�������������������������������
Adding Specifications����������������������������
Identifying Opportunities for Reuse������������������������������������������
Restructuring the Design�������������������������������
Optimizing the Design����������������������������
Mapping Problem-Domain Classes to Implementation Languages�����������������������������������������������������������������
Constraints and Contracts��������������������������������
Types of Constraints���������������������������
Elements of a Contract�����������������������������
Method Specification���������������������������
General Information��������������������������
Events�������������
Message Passing����������������������
Algorithm Specifications�������������������������������
Example��������������
Verifying and Validating Class and Method Design�������������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter review���������������������
Chapter 9 Data Management Layer Design���������������������������������������������
Introduction�������������������
Object Persistence Formats���������������������������������
Sequential and Random Access Files�����������������������������������������
Relational Databases���������������������������
Object-Relational Databases����������������������������������
Object-Oriented Databases��������������������������������
NoSQL Data Stores������������������������
Selecting an Object Persistence Format���������������������������������������������
Mapping Problem Domain Objects to Object Persistence Formats�������������������������������������������������������������������
Mapping Problem Domain Objects to an OODBMS Format���������������������������������������������������������
Mapping Problem Domain Objects to an ORDBMS Format���������������������������������������������������������
Mapping Problem Domain Objects to a RDBMS Format�������������������������������������������������������
Optimizing RDBMS-Based Object Storage
Optimizing Storage Efficiency������������������������������������
Optimizing Data Access Speed�����������������������������������
Estimating Data Storage Size�����������������������������������
Designing Data Access and Manipulation Classes�����������������������������������������������������
Nonfunctional Requirements and Data Management Layer Design������������������������������������������������������������������
Verifying and Validating the Data Management Layer���������������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Chapter 10 Human–Computer Interaction Layer Design���������������������������������������������������������
Iintroduction��������������������
Principles for User Interface Design�������������������������������������������
Layout�������������
Content Awareness������������������������
Aesthetics�����������������
User Experience����������������������
Consistency������������������
Minimizing User Effort�����������������������������
User Interface Design Process������������������������������������
Use Scenario Development�������������������������������
Navigation Structure Design����������������������������������
Interface Standards Design���������������������������������
Interface Design Prototyping�����������������������������������
Interface Evaluation���������������������������
Common Sense Approach to User Interface Design�����������������������������������������������������
Navigation Design������������������������
Basic Principles�����������������������
Types of Navigation Controls�����������������������������������
Messages���������������
Navigation Design Documentation��������������������������������������
Input Design�������������������
Basic Principles�����������������������
Types of Inputs����������������������
Input Validation�����������������������
Output Design��������������������
Basic Principles�����������������������
Types of Outputs�����������������������
Media������������
Mobile Computing and User Interface Design�������������������������������������������������
Social Media and User Interface Design���������������������������������������������
Games, Multidimensional Information Visualizations, and Immersive Environments
Games, Gamification, and User Interface Design�����������������������������������������������������
Multidimensional Information Visualization Design��������������������������������������������������������
User Interface Design and Immersive Environments�������������������������������������������������������
International and Cultural Issues and User Interface Design������������������������������������������������������������������
Multilingual Requirements��������������������������������
Color������������
Cultural Differences���������������������������
Nonfunctional Requirements And Human-Computer Interaction Layer Design�����������������������������������������������������������������������������
Applying The Concepts At Patterson Superstore����������������������������������������������������
Chapter review���������������������
Chapter 11 Physical Architecture Layer Design����������������������������������������������������
Introduction�������������������
Elements of the Physical Architecture Layer��������������������������������������������������
Architectural Components�������������������������������
Server-Based Architectures���������������������������������
Client-Based Architectures���������������������������������
Client–Server Architectures����������������������������������
Client–Server Tiers��������������������������
Selecting a Physical Architecture����������������������������������������
Cloud Computing����������������������
Ubiquitous Computing and the Internet of Things������������������������������������������������������
Green IT���������������
Infrastructure Design����������������������������
Deployment Diagram�������������������������
Network Model��������������������
Hardware and System Software Specifications��������������������������������������������������
Nonfunctional Requirements and Physical Architecture Layer Design������������������������������������������������������������������������
Operational Requirements�������������������������������
Performance Requirements�������������������������������
Security Requirements����������������������������
Cultural and Political Requirements������������������������������������������
Synopsis���������������
Verifying and Validating the Physical Architecture Layer���������������������������������������������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
PART THREE CONSTRUCTION, INSTALLATION, AND OPERATIONS������������������������������������������������������������
Chapter 12 Construction������������������������������
Introduction�������������������
Managing Programming���������������������������
Assigning Programmers����������������������������
Coordinating Activities������������������������������
Managing the Schedule����������������������������
Cultural Issues����������������������
Developing Documentation�������������������������������
Types of Documentation�����������������������������
Designing Documentation Structure����������������������������������������
Writing Documentation Topics�����������������������������������
Identifying Navigation Terms�����������������������������������
Designing Tests����������������������
Testing and Object Orientation�������������������������������������
Test Planning��������������������
Unit Tests�����������������
Integration Tests������������������������
System Tests�������������������
Acceptance Tests�����������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Chapter 13 Installation and Operations���������������������������������������������
Introduction�������������������
Cultural Issues and Information Technology Adoption����������������������������������������������������������
Conversion�����������������
Conversion Style�����������������������
Conversion Location��������������������������
Conversion Modules�������������������������
Selecting the Appropriate Conversion Strategy����������������������������������������������������
Change Management������������������������
Understanding Resistance to Change�����������������������������������������
Revising Management Policies�����������������������������������
Assessing Costs and Benefits�����������������������������������
Motivating Adoption��������������������������
Enabling Adoption: Training����������������������������������
Post-Implementation Activities�������������������������������������
System Support���������������������
System Maintenance�������������������������
Project Assessment�������������������������
Applying the Concepts at Patterson Superstore����������������������������������������������������
Chapter Review���������������������
Index������������
EULA
2015-04-08T19:16:28+0000
Preflight Ticket Signature
Use of web-based sources
is not appropriate,
do use 12 point fonts preferred, single spaced with no links or citations of any kind
1. What is a systems requirement as seen in Chapter 3 and its importance in building a Systems Proposal?
2. Chapter 4 is about understanding business processes and how we can model them to help build systems. How do we do this, what “things” do we use to construct the beginning process?
An example of a very good set of responses would be:
Q: What is and how does Cloud computing work?
A.
Statement: Cloud computing is a current buzz word phrase that describes how the current technology of mainframe to terminal hardware architecture has evolved with the advent of smart phones and more useful (processor and memory enabled) technology. The current state of technology has allows us to augment our mobile life style environment with almost instant access to huge databases of materials we may find fun or necessary to our business and personal lives. It does this using the internet and the associated backbone via ISP and NSP infrastructure. The connectivity we have is the backbone of our society today.
B.
Proof or value added or examples to the above statement is: (Minimum of three (3))
1. Mainframe technology has gained significant capability to engage millions of users simultaneously and integrate their multiple databases into a searchable field of evidence for decision making.
2. The infrastructure of the communication’s world has granted us the ability to connect in many places within our cultural environment at a price point we can afford as well as the security level we are willing to allow. This connectivity is the one of the three keys to our level of success today.
3. The massive databases we can access almost at will brings us more than just data, but information and relationships of how the interaction of the information. It is a decision process enabler that we use. However, it is our experiences that allow us a use of the information so we can make the best possible decision in a given time frame that makes this system valuable.
Examples of the above technology infrastructure are: IBM, HP, CISCO, DELL and other massive mainframe sites located in Data Centers around the country and world allow the processing of millions of data request every second, Connectivity is granted by every communications vendors at high speed and bandwidth, such as AT&T, Verizon, Cable and Wireless etc… This is done/underwritten with government funded research and endowments via organizations like DARPA and DOE, and the FCC.
C.
Conclusion: This is a highly reliable (up-time) set of systems that is robust enough for today’s demands and flexible enough for us to evolve it as our uses expand and change over time. The Cloud is not new except as a term to characterize what we use and how we use technology and information today. The figures attached demonstrate the complexity as well as the simplicity of it. The entire Cloud based technology is a challenge to keep running and the security aspects of keeping each person’s or organization’s data clean and secure is difficult.
The above answer would get you approximately an 9 out of 10 points. I want depth and that you show me some analytical thought processes.
The Template is:
1. Statement about the question (Do not restate the question)
2. Three (3) examples, proofs or value added elements/statements to back up your original comment statement (see item 1).
a.
b.
c.
3. Conclusion
The template/format is not optional!
Expert paper writers are just a few clicks away
Place an order in 3 easy steps. Takes less than 5 mins.