P e r f o r m a n c e A n a l y s i s of T r a n s a c t i o n P r o - Each of these provides a nice overview of the cessing S y s t e m s , Wilbur H. Highleyman, design of the relevant sub-system, and then proPrentice Hall, 1989. 424 pp. $55.40. vides a series of appropriate performance models. Throughout there are worked-out examAs its title indicates, this book has the very specific purpose of applying performance evalu- ples using the performance models, using realisation tools to the study of on-line transaction tic parameter values. The examples also discuss processing systems. The book provides both an the implications of the conclusions of the perforoverview of the relevant mathematical methods mance model, describing how performance could from performance evaluation, and an application be improved, and principles of good system deof those methods to transaction processing sys- sign demonstrated by the results of the example. The last two chapters provide a different pertems. spective. One discusses the performance reAfter an introductory chapter laying out the port, describing both how to organize the report principles of transaction processing systems, the and how the performance analyst should interact author has a brief chapter showing how queueing with other systems personnel in creating and valtheoretic models can be Used to model computer idating the performance model. The final chapsystem performance. This is intended to moter is an extensive case study and reproduces the tivate the reader more than to provide specific actual performance report produced by the auanalysis tools. thor for a client. This case study will certainly The next chapter gives the background in give the reader a feel for the scale of actual perqueueing theory required for the remainder of formance studies: there are forty input paramthe book. The chapter starts with a statement eters and over forty equations in the analysis. of the Pollaczek-Khintchine formula for M/G/1 This chapter ends with a comparison of the requeues, for which a derivation is given in an sults of running a benchmark with the results of appendix. The author also presents informal the analytic model, showing excellent agreement. derivations of Little's Law and the queue length The book ends with several appendices, giving distribution of an M / M / 1 queue. Several other detailed derivations of some of the more compliqueueing models are discussed, including multi- cated models, and a bibliography of books and ple server models, priority queues, finite length journal articles. queues, and queues with finite populations. The author also discusses series of queues. Unfor- P e r f o r m a n c e M e a s u r e m e n t o f C o m p u t e r tunately, this chapter is really too brief to give S y s t e m s , Phillip McKerrow, Addison-Wesley, anovice an understanding of the tools necessary 1988. 260 pp. $35.50. for Performance evaluation (for example, general Measurement of computer system perfornetworks of queues are not discussed) and con- mance usually occupies at most a chapter or tains some unfortunate errors (for example, stat- two of performance evaluation texts, but here ing that the variance of the sum of random vari- it is given the book-length treatment it deserves. ables is the sum of the variances, without stat- The author begins the book with an introductory ing the requirement that the random variables chapter discussing the purposes and goals of perbeing summed are independent). This limited formance measurement, which of course varyn background forces the author to present all his from one study to another. He then surveys the performance models in the form of single queue kinds of measurement tools available, and sets or series queueing models, or else to quote results out his philosophy of measurement methodology from the literature with little explanation. (which includes references to Aristotle and the The next five chapters present performance Renaissance world view), which is expanded in a models for specific sub-systems of a transaction later chapter. processing system: communications, processThe main portion of the book is occupied with ing, data-base, applications, and fault-tolerance. discussions of the types of performance tools Performance Evaluation Review Vol. 18 #4, April 1991

/lp/association-for-computing-machinery/book-review-performance-measurement-of-computer-systems-by-phillip-PIhpbfLJpb