Performance measurement and modeling is essential for both real systems and systems being modeled in a research environment. Without good models and tests to verify that the systems match those models, it can be very difficult indeed to correctly provision the systems with servers, forecast load problems (both coverage load and transient peaks), and generally ensure that the system is stable and responsive. Real systems are often very complicated with a dozen server types, multiple queues, wildly varying response times, and the like. Individual components may be measured, but the interactions among components are much more difficult.

The preface states that the book is intended for someone “interested in the performance of [a] system” and who would like to model the system’s behavior, but is not for those who want to “become an expert in performance analysis.”

Queueing theory is often mentioned in conjunction with systems performance modeling, but this book uses queueing systems “only as an expository device for a coherent presentation.”

Ten chapters cover many facets of system performance modeling, including some basic queueing theory, birth-death processes, Markov chains, experimental validation of models, and so on. These chapters are quite short, with the longest at about seven pages of main text, but all chapters include a discussion of the topics as presented in other research papers. These research papers are picked from a pool of 40; the annotations for each paper range widely in size, from a short paragraph to more than a page long. This is an interesting presentation style that allows for a generally productive read. However, the papers really need to be easily available to readers, as the book’s text sometimes refers to specific results or specific figures in a paper, often without much explanation. (The two papers I looked for were easy enough to find online via CiteSeerX, but I did not check all 40.)

There is a chapter of exercises, one for each of the papers used in the pool, but no answers, and a bibliography that includes the papers in the pool (which are also all cited in the preface).

The writing is clear and lucid, the math is well explained, and the separation of text and annotations works nicely. However, a few more pages of exposition in each chapter would not be amiss--sometimes the reading feels a bit telegraphic.

The book would be a nice way to approach performance measurement in a classroom setting, but is less likely to be a satisfactory introduction for autodidacts, as the coverage of many topics is too brief. Furthermore, modeling real systems is far more complex than the relatively simple systems described, and it is not obvious how to apply the lessons from the text to real systems.