Measuring computer system performance is no mean task. There is a lot of confusion among technicians and capacity planners about how to reasonably and intelligently approach the measurement problem and about which performance metrics to select. Still, understanding the characteristics and distributions of the various loads on a computer (or set of interrelated computers) is fundamental to effective management of the technology environment. These issues must be addressed regularly by organizations that provide information technology infrastructure capabilities.
Lilja’s book is conceptually well formulated, mathematically well grounded, and even intuitively appealing. The author focuses on the key issues in quantifying and characterizing the capacity and usage of a typical computer (or server, switch, or any other device based on computing or networking architectures). He begins by discussing the concept of measurement. The next several chapters intuitively develop the statistical tools and techniques that form the basis of computer performance measurement and capacity planning. The ideas of the accuracy, precision, and resolution of measurement tools are also presented. The author emphasizes the use of statistically valid tools and techniques to both collect and analyze the data. There is also a strong emphasis on interpreting the results of these analyses, which is a crucial aspect of the effective measurement of such performance.
Once the conceptual material is covered, the author turns to the more concrete aspects of the topic, beginning with a discussion of events, specific counting and measuring techniques, and benchmarking. Next, he discusses linear regression modeling and techniques for maximizing the results of the data collection efforts needed to construct and use effective models. This discussion naturally leads to an exposition on simulation modeling and associated strategies. The text concludes with the fundamentals of modeling using queueing theory.
This book is clearly an essential reference for professionals engaged in computer (or network) performance measurement and capacity planning. At first blush, it appears that it is not a book for the mathematically faint of heart, because it includes a healthy dose of mathematical formulations. However, that first impression is misleading. The first half of this book is an interesting intuitive overview of the principles and issues related to this topic. As such, it can be an excellent reference for information technology managers who need a conceptual understanding of this topic in order to handle performance issues in computers and networks.