Besides its case studies on applying queueing theory to the analysis of computer performance, this book on queueing circuits also includes a quick user guide to Perl Pretty Damn Quick (PDQ), a reference manual, and numerous examples. (PDQ is a front-end Perl language module to native C code for the numerical solution of queueing circuits.) It consists of three parts, a few short appendices, a nice bibliography, and an adequate index.

Part 1 is an introduction to the metrology of computer performance analysis. Part 2 is a discussion of the fundamentals and principles of queueing theory that uses high-school algebra and intuition rather than a more formal approach. Part 3 is a collection of case studies in quantitative computer performance analysis using queueing circuits. The short appendices comprise ancillary reference material and a Perl PDQ reference manual. Exercises and solutions are included.

The book should be read from cover to cover, even if its organization is somewhat suboptimal--this flaw may taint its value as a professional reference. The English prose is straightforward, but the editing is sloppy and there are many instances of missing words, wrong concordances, sentence fragments, and other minor flaws. The book is written to be a field manual for enlightened practitioners of performance analysis, or possibly a school textbook. Gunther deliberately avoids using calculus when presenting the tools of the trade; the outcome of his decision is open to discussion. It is a laudable achievement to make performance analysis theoretical foundations available to anyone who is interested and can master high-school algebra. A drawback is the nonaxiomatic, intuition-driven narrative that falls into the trap of too many forward and backward references, resulting in a tangled presentation that is perhaps more suited to hypertext than print. Readers should ideally have some prior understanding of probability and stochastic processes--for instance, the assumptions required of Poisson arrivals (constant probability, no clustering, and independence) are discussed in a fragmented manner, so that their overall significance is difficult to comprehend without other sources of knowledge.

The book makes it clear that performance modeling is about insight and demands simplicity rather than complexity. Artful practitioners are able to recognize the few key elements of a very complex system in order to apply analytical abstractions that are appropriate for the quantitative prediction of performance and behavior. This book makes the tools to build these abstractions available. It also clearly acknowledges that no tool is adequate when misapplied. Accordingly, the case studies encompass many areas of interest, including operating system workloads, multiprocessor computers, computational clusters, client/server analysis, Web application scalability, virtualization, and even a very interesting analysis of Amdahl’s law in terms of queueing behavior.

Gunther’s deliberate avoidance of calculus--even though the subject sometimes screams for its application--seems inappropriate in a college-level textbook for a computer science major. Nevertheless, it may be suitable for other areas of concentration where computer systems performance analysis remains a subject too often neglected.