The K2 towers at 8,611 meters in the Karakorum range of the western Himalayas. It is the second highest peak of the world and is considered the most difficult 8000er to climb. An expedition to the K2 typically lasts several months in the summer, when the weather is most favorable. Even in summer, snowstorms are frequent. An expedition requires thousands of pounds of equipment, including climbing gear, severe weather protection gear, tents, food, communication equipment, and pay and shoes for hundreds of porters. Planning such an expedition takes a significant amount of time in the life of a climber and requires dozens of participants in supporting roles. Once on site, many unexpected events, such as avalanches, porter strikes, or equipment failures, will force the climbers to adapt, find new solutions, or retreat. The success rate for expeditions to the K2 is currently less than 40%.
The United States National Airspace System (NAS) monitors and controls air traffic in the United States. The NAS includes more than 18,300 airports, 21 air route traffic control centers, and over 460 control towers. These add up to more than 34,000 pieces of equipment, including radar systems, communication switches, radios, computer systems, and displays. The current infrastructure is aging rapidly. The computers supporting the 21 air route traffic control centers, for example, are IBM 3083 mainframes that date back to the early 1980s. In 1996, the United States government initiated a program to modernize the NAS infrastructure, including improvements such as satellite navigation, digital controller/pilot communications, and a higher degree of automation in controlling the air routes, deciding the order in which aircraft land, and controlling ground traffic as aircraft move from and to the runways. Such a complex infrastructure, however, can only be modernized incrementally. Consequently, while new components offering new functionality are introduced, older components still need to be supported. For example, during the transition period, a controller will have to be able to use both analog and digital voice channels to communicate with pilots. Finally, the modernization of the NAS coincides with a dramatic increase in global air traffic, predicted to double within the next 10–15 years. The previous modernizing effort of the NAS, called the Advanced Automation System (AAS), was suspended in 1994 because of software-related problems, after missing its initial deadline by several years and exceeding its budget by several billions of dollars.
Both of the above examples discuss complex systems in which external conditions can trigger unexpected changes. Complexity puts the problem beyond the control of any single individual. Change forces participants to move away from well-known solutions and to invent new ones. In both examples, several participants need to cooperate and develop new techniques to address these challenges. Failure to do so results in failure to reach the goal.
This book is about conquering complex and changing software systems.
The application domain (mountain expedition planning, air traffic control, financial systems, word processing) usually includes many concepts that software developers are not familiar with. The solution domain (user interface toolkits, wireless communication, middleware, database management systems, transaction processing systems, wearable computers) is often immature and provides developers with many competing implementation technologies. Consequently, the system and the development project are complex, involving many different components, tools, methods, and people.
As developers learn more about the application domain from their users, they update the requirements of the system. As developers learn more about emerging technologies or about the limitations of current technologies, they adapt the system design and implementation. As quality control finds defects in the system and users request new features, developers modify the system and its associated work products. The result is continuous change.
Complexity and change represent challenges that make it impossible for any single person to control the system and its evolution. If controlled improperly, complexity and change defeat the solution before its release, even if the goal is in sight. Too many mistakes in the interpretation of the application domain make the solution useless for the users, forcing a retreat from the route or the market. Immature or incompatible implementation technologies result in poor reliability and delays. Failure to handle change introduces new defects in the system and degrades performance beyond usability.
This book reflects more than 10 years of building systems and of teaching software engineering project courses. We have observed that students are taught programming and software engineering techniques in isolation, often using small problems as examples. As a result, they are able to solve well-defined problems efficiently, but are overwhelmed by the complexity of their first real development experience, when many different techniques and tools need to be used and different people need to collaborate. Reacting to this state of affairs, the typical undergraduate curriculum now often includes a software engineering project course, organized as a single development project.
The tools: UML, Java, and Design Patterns
We wrote this book with a project course in mind. The book can be used, however, in other situations as well, such as short and intensive workshops or short-term R&D projects. We use examples from real systems and examine the interaction among state-of-the art techniques, such as UML (Unified Modeling Language), Java-based technologies, design patterns, design rationale, configuration management, and quality control. Moreover, we discuss project management related issues that are related to these techniques and their impact on complexity and change.
We teach software engineering following five principles:
Practical experience. We believe that software engineering education must be linked with practical experience. Students can understand complexity only by working with a complex system—that is, a system that no single student can completely understand.
Problem solving. We believe that software engineering education must be based on problem solving. Consequently, there are no right or wrong solutions, only solutions that are better or worse relative to stated criteria. Although we survey existing solutions to real problems and encourage their reuse, we also encourage criticism and the improvement of standard solutions.
Limited resources. If we have sufficient time and resources, we could perhaps build the ideal system. There are several problems with such a situation. First, it is not realistic. Second, even if we had sufficient resources, if the original problem rapidly changes during the development, we would eventually deliver a system solving the wrong problem. As a result, we assume that our problem-solving process is limited in terms of resources. Moreover, the acute awareness of scarce resources encourages a component-based approach and reuse of knowledge, design, and code. In other words, we support an engineering approach to software development.
Interdisciplinarity. Software engineering is an interdisciplinary field. It requires contributions from areas spanning electrical and computer engineering, computer science, business administration, graphic design, industrial design, architecture, theater, and writing. Software engineering is an applied field. When trying to understand and model the application domain, developers interact regularly with others, including users and clients, some of whom know little about software development. This requires viewing and approaching the system from multiple perspectives and terminologies.
Communication. Even if developers built software for developers only, they would still need to communicate among themselves. As developers, we cannot afford the luxury of being able to communicate only with our peers. We need to communicate alternatives, articulate solutions, negotiate trade-offs, and review and criticize others’ work. A large number of failures in software engineering projects can be traced to the communication of inaccurate information or to missing information. We must learn to communicate with all project participants, including, most importantly, the client and the end users.
These five principles are the basis for this book. They encourage and enable the reader to address complex and changing problems with practical and state-of-the-art solutions.
This book is based on object-oriented techniques applied to software engineering. It is neither a general software engineering book that surveys all available methods nor a programming book about algorithms and data structures. Instead, we focus on a limited set of techniques and explain their application in a reasonably complex environment, such as a multi-team development project that includes 20 to 60 participants. Consequently, the book also reflects our biases, our strengths, and our weaknesses. We hope, nevertheless, that all readers will find something they can use. The book is structured into 16 chapters organized into three parts, which can be taught as a semester-long course.
Part I, Getting Started, includes three chapters. In this part, we focus on the basic skills necessary for a developer to function in a software engineering context.
In Chapter 1, Introduction to Software Engineering, we describe the difference between programming and software engineering, the current challenges in our discipline, and basic definitions of concepts we use throughout the book.
In Chapter 2, Modeling with UML, we describe the basic elements of a modeling language, UML, used in object-oriented techniques. We present modeling as a technique for dealing with complexity. This chapter teaches the reader how to read and understand UML diagrams. Subsequent chapters teach the reader how to build UML diagrams to model various aspects of the system. We use UML throughout the book to model a variety of artifacts, from software systems to processes and work products.
In Chapter 3, Project Organization and Communication, we introduce basic concepts of project organization and communication. Developers and managers spend more than half of their time communicating with others, either face-to-face or via E-mail, groupware, video conference, or written documents. Whereas modeling deals with complexity, communication deals with change. We describe project organizations and discuss what constitutes effective communication.
In Part II, Dealing with Complexity, we focus on methods and technologies that enable developers to specify, design, and implement complex systems.
In Chapter 4, Requirements Elicitation, and Chapter 5, Analysis, we describe the definition of the system from the users’ point of view. During requirements elicitation, developers determine the functionality users need and a usable way of delivering it. During analysis, developers formalize this knowledge and ensure its completeness and consistency. We focus on how UML is used to deal with application domain complexity.
In Chapter 6, System Design: Decomposing the System, and Chapter 7, System Design: Addressing Design Goals, we describe the definition of the system from the developers’ point of view. During this phase, developers define the architecture of the system in terms of design goals and a subsystem decomposition. They address global issues, such as the mapping of the system onto hardware, the storage of persistent data, and global control flow. We focus on how developers can use architectural styles, components, and UML to deal with solution domain complexity.
In Chapter 8, Object Design: Reusing Pattern Solutions, Chapter 9, Object Design: Specifying Interfaces, and Chapter 10, Mapping Models to Code, we describe the detailed modeling and construction activities related to the solution domain. During this phase, developers identify and adapt design patterns and frameworks to realize specific subsystems. They refine and specify precisely the interfaces of classes using constraint languages such as UML’s Object Constraint Language. Finally, they map the detailed object design model to source code and database schema.
In Chapter 11, Testing, we describe the validation of system behavior against the system models. Testing detects faults in the system, including those introduced during changes to the system or its requirements. Testing activities include unit testing, integration testing, and system testing. We describe several testing techniques, such as whitebox, blackbox, path testing, state-based testing, and inspections, and discuss their application to object-oriented systems.
In Part III, Managing Change, we focus on methods and technologies that support the control, assessment, and implementation of changes throughout the development of a system.
In Chapter 12, Rationale Management, we describe the capture of design decisions and their justifications. The models developed during requirements elicitation, analysis, and system design help us deal with complexity by providing different perspectives on what the system should be doing and how it should do it. To be able to deal with change, we need also to know why the system is the way it is. Capturing design decisions, considered alternatives, and their argumentation enables us to access the rationale of the system.
In Chapter 13, Configuration Management, we describe techniques for modeling the project history. Configuration management complements rationale in helping us deal with change. Version management records the evolution of the system. Release management ensures consistency and quality across the components of a release. Change management ensures that modifications to the system are consistent with project goals.
In Chapter 14, Project Management, we describe techniques for initiating a software development project, tracking its progress, and dealing with risks and unplanned events. We focus on organizations, roles, and management activities that allow a large number of participants to collaborate and deliver a high-quality system within planned constraints.
In Chapter 15, Software Life Cycle, we describe software life cycles, such as Boehm’s Spiral Model and the Unified Software Development Process, that provide an abstract model of development activities. In this chapter, we also describe the Capability Maturity Model, which is used for assessing the maturity of organizations.
In Chapter 16, Methodologies: Putting It All Together, we describe methodologies and heuristics for applying the material covered in the other chapters to concrete situations. No matter how thorough the requirements elicitation or detailed the planning, projects of any realistic size encounter unexpected events and changes. Dealing with uncertainty makes real projects and systems look very different from projects and systems examined in textbooks. In this chapter, we describe several different methodologies, discuss issues that need to be addressed in every project, and present three case studies of actual projects.
The topics above are strongly interrelated. To emphasize their relationships, we selected an iterative approach. Each chapter consists of five sections. In the first section, we introduce the issues relevant to the topic with an illustrative example. In the second section, we describe briefly the activities of the topic. In the third section, we explain the basic concepts of the topic with simple examples. In the fourth section, we detail the technical activities with examples from real systems. Finally, we describe management activities and discuss typical trade-offs. In Chapters 4–10, we present a running case study of a complex multi-user game management system called ARENA. By repeating and elaborating on the same concepts in increasingly complex examples, we hope to provide the reader with an operational knowledge of object-oriented software engineering.
Building a large, complex system can be compared with climbing a big mountain. It is good to have a route description, but the route can never be completely mapped out, as new crevasses may open anytime. Even though we map out our software engineering knowledge in this book, changes will occur and methods that we believe in now may be out of date soon.
How can we teach students to cope with such rapidly changing conditions? For us, the most important thing to pass on to a student is not only knowledge of the map, but also the ability to negotiate the terrain. Although it is wise to study the description of a route, there is no substitute for the experience of actually traveling the route.
We wrote this book for a semester-long software engineering project course for senior or graduate students. We assume that students have experience with a programming language such as C, C++, C#, or Java. We expect that students have the necessary problem-solving skills to attack technical problems, but we do not expect that they have been exposed to the complex or changing situations typical of system development. This book can also be used for other types of courses, such as short, intensive professional courses.
Project and senior-level courses.
A project course should include all the chapters of the book, roughly in the order presented. An instructor may consider teaching more advanced project management concepts from Chapter 14, Project Management, early in the course so that students become familiar with planning and controlling.
Introductory-level course. An introductory course with homework should focus on the first three sections of each chapter. The fourth section and the case study can be used as material for homework and can simulate the building of a minisystem using paper for UML diagrams, documents, and code.
Short technical course. The book can also be used for a short, intensive course geared toward professionals. A technical course focusing on UML and object-oriented methods could use the chapter sequence 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, covering all development phases from requirements elicitation to testing. An advanced course would also include Chapter 12, Rationale Management, and Chapter 13, Configuration Management.
Short management course. The book can also be used for a short, intensive course geared toward managers. A management course focusing on managerial aspects such as communication, risk management, rationale, maturity models, and UML could use the chapter sequence 1, 2, 3, 14, 15, 16, 12, 13.
The second edition of this book started as a well-scoped project. Our goal was to add two new chapters and a case study to address the feedback we received from the first users of our book. The work was to last one year.
Two years later, we found ourselves with four new chapters and a complete overhaul of many of the existing chapters. Between adding detailed examples, keeping up with the latest developments in software engineering, maintaining the coherence of the book, and meeting the schedule, we decided to compromise on the schedule. We hope that the quality of the final product reflects the delay. We thank Alan Apt, our publisher, for his infinite patience. We made the following changes:
Running case study. We received many requests for a single running example through the book, so that the relationships among chapters become visible. Consequently, we threaded a running case study of ARENA into the technical chapters of the book. The material associated with the study is also available online at http://oose.globalse.org.
Expanded object design material. We expanded the coverage of design patterns and consolidated all material related to reuse into a new chapter, Chapter 8. Similarly, we expanded the coverage of OCL (Object Constraint Language), in terms of both concepts and examples, and included this material in a new chapter, Chapter 9. In both cases, we moved away from writing a reference manual and, instead, focused on providing knowledge about the application of these concepts.
Extended scope to selected implementation activities. We found that many students initially have difficulty relating the new material (e.g., requirements engineering, UML modeling) to the concepts they already know (e.g., programming). To address this issue, we extended the scope of the book to include selected implementation topics. A new chapter, Chapter 10, describes how models are mapped into source code.
Reorganized project management and software life cycle material. Software engineering courses often treat project management and software life cycle topics together and at the beginning of the course, leading to a top-down approach to teaching software engineering. In our experience, we found this material difficult to communicate to students if they have not yet been exposed to issues inherent in large projects. Consequently, we chose a bottom-up approach, revisiting this material incrementally and in broadening scopes. Chapter 3, in the early section of the course, focuses only on the basic concepts of project management from the developer’s point of view. Chapter 14 revisits and expands these concepts from the novice project manager’s point of view. Once project management issues are understood, Chapter 15 focuses on software life cycle issues and how to transfer process knowledge across projects. Both Chapter 14 and Chapter 15 take an ideal and academic perspective to the topic. To balance this ideal perspective and inject a dose of realism, we discuss methodological issues faced in real projects in Chapter 16.
Dr. Bernd Bruegge has been studying and teaching Software Engineering at Carnegie Mellon University for 20 years, where he received his masters and doctorate degrees. He received his Diplom from the University of Hamburg. He is now a university professor of Computer Science with a chair for Applied Software Engineering at the Technische Universität München and an adjunct faculty member of Carnegie Mellon University. He has taught object-oriented software engineering project courses on the text materials and website described in this book for 15 years. He won the Herbert A. Simon Excellence in Teaching Award at Carnegie Mellon University in 1995. Bruegge is also an international consultant and has used the techniques in this book to design and implement many real systems, including an engineering feedback system for DaimlerChrysler, an environmental modeling system for the U.S. Environmental Protection Agency, and an accident management system for a municipal police department, to name just a few.
Dr. Allen Dutoit is a Research Scientist at the Technische Universität München. He received his M.S. and Ph.D. from Carnegie Mellon University and his Diplôme d’Ingénieur from the Swiss Federal Institute of Technology in Lausanne. He has been teaching software engineering project courses with Professor Bruegge since 1993, both at Carnegie Mellon University and the Technische Universität München, where they used and refined the methods described in this book. Dutoit’s research covers several areas of software engineering and object-oriented systems, including requirements engineering, rationale management, distributed development, and prototype-based systems. He was previously affiliated with the Software Engineering Institute and the Institute for Complex Engineered Systems at Carnegie Mellon University.
