Computing Reviews

Measuring computer performance :a practitioner’s guide
Lilja D., Cambridge University Press,New York, NY,2000. 261 pp.Type:Book
Date Reviewed: 06/20/02

The author states in the preface: “This book is intended to be used as the primary text in a one-semester course for advanced undergraduate and beginning graduate students in computer science and engineering who need to understand how to rigorously measure the performance of computer systems.” He goes on to claim, “This text will also be useful as a reference text for professional engineers and scientists who use computers in their daily work, or who design systems that incorporate computers as their primary control elements.” This book succeeds in the first of these objectives but is only marginally useful in the second.

A preface section marked “Organization” gives an accurate overview of the book’s content. The following are included: goals of performance analysis and measurement; performance metrics; means; modes; variability; sources and analysis of errors; measurement tools and techniques; perturbations due to measurement; benchmarks techniques; linear regression analysis; design of experiments; simulation; and queuing analysis. All but the last two chapters (on simulation and queuing analysis) are appropriate for a book on performance measurement. The last two chapters, tangential to the central topic, are too limited and too weak to be of much use, except as a teaser for these very big subjects. These chapters would have been more useful had the connection between these analytical/simulation methodologies and the analysis of measured data been made clearer, and if a few more important models had been given, (such as M/G/1 and GI/G/1) even if only in a cookbook form. As it is, students and performance neophytes are unlikely to get much out these chapters, and are even less likely to apply these ideas to the data reduction of measured data.

Another weakness that is likely to bewilder the student and annoy the professional is the weak index. Key technical terms such as “sample mean” and “population mean” are missing. References to these terms, however, do occur elsewhere. The student and the professional will have to dig back through previous pages to find these and other definitions. If the book is to be used as a student text (as intended by the author) then the instructor would be advised to make a first course in probability and statistics a prerequisite or provide an auxiliary text for the purpose, since only a student who already knows the statistical notions is likely to get through the statistical sections of this book without help.

As for its secondary intended purpose, as a reference text for professional engineers, I do not feel that it will be successful. For one thing, professional readers will have little patience for the use of technical terms without clear definitions, for the thin index and inadequate glossary, for the uneven level of the mathematics, and for the sometimes opaque mathematical notation. Most problematic is the paucity of bibliographic and reference material. Although the papers and books cited are all relevant and important, they are scattered throughout the book, and are simply not enough for the would-be professional user.

These criticisms aside, I do recommend serious consideration of this book for a one-semester, advanced undergraduate, first-year graduate course focused on performance measurement, based on the first eight chapters. This could possibly be followed by a more advanced course on modeling and analysis taught from a different text.

Reviewer:  B. Beizer Review #: CR126187 (0208-0398)

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy