Part I introduces software performance and the authors' software performance engineering (SPE) methodology. Although the book continues to drill down deeper into SPE and associated factors in subsequent chapters, this part of the book is my favorite. What makes it my favorite is the context in which the authors cast performance as a function of resource requirements (workload) and configuration (capacity). Although this is not a new concept, the ensuing discussion leading to SPE modeling strategies and models and the 9-step SPE process exposes the challenges and provides a sense that OO performance can be managed through careful systems analysis, modeling and design early in the life cycle. I also like the way SPE is aligned to the "Unified Software Process", which is the RUP thinly disguised. The chapter that discusses this has some realm gems, including performance patterns and anti-patterns. In addition, the SPE is also aligned to UML, with an excellent discussion on extending the UML and some example scenarios that show how to specify time, concurrency and other performance characteristics.

The SPE models given in Part II cover the full spectrum of system types, including distributed and web-based systems. The material is highly technical and requires close attention. It is also clearly written and will provide the design team (not to mention the post-implementation support team) with analytical techniques and an effective analysis approach to performance management. The highlights of this part of the book were the way middleware overhead is taken into account, scenarios and modeling hints.

Data collection is the topic of Part III and is covered in detail. The chapters I most liked in this part included resource estimation techniques and software measurement and instrumentation. Part IV will be the focal point for designers and architects in that it addresses performance solutions. Chapter topics include: performance-oriented design, performance patterns and performance anti-patterns(excellent material!) and implementation solutions. The latter covers performance tuning, as well as language-dependent and -independent solutions for OO software. The two languages discussed are c++ and java.

Part V is devoted to applications and provides a thorough discussion of web and realtime embedded system performance issues. The format of the two chapters in this part include: performance issues, SPE models, environments and constraints. Realtime systems are given a more in-depth treatment because of the special nature of these systems. I liked the case study that used a telephone switching system for realtime systems because I have professional interests in that area.

The actual steps for implementing SPE is the subject of Part VI, which covers the process, associated life cycle activities, post-deployment performance management and how to integrate SPE into your organization's software process. Part VII's two appendices are valuable. Appendix A covers UML notation and Appendix B covers SPE modeling notations. Both are useful as quick reference guides.

This is a comprehensive book that, while focused on a narrow topic, covers all issues and factors in minute detail. The book complements two other outstanding works, Software Reliability Engineered Testing by John D. Musa, and Testing Object-Oriented Systems by Robert V. Binder. Although the latter books are more focused on testing, the material dovetails nicely with the SPE approach given in this book.