Program Behavior: Models and Measurements [antikvár]

PrefaceThe study of program behavior has lately increased in intensity. There havebeen many recent research papers on the subject, and even a special issue ofIEEE Computer (November, 1976), not to mention frequent applications ofexisting program models in system analysis and simulation. But to my knowledge,there is no up-to-date attempt to draw the concepts together into a survey of thefield. I do not claim that all aspects of program behavior receive adequate cover-age here, but this book does attempt to meet this need.There are probably good reasons why there has been no previous book on thissubject-the field is hardly stabilized, and the current best models are still defi-cient in many ways. When teaching a rapidly changing technical discipline, it isusually thought desirable to emphasize the basic mathematical techniques thatwill not be made obsolete by advancing technology. Here, quite a lot of space isdevoted to the methodology of measurement and statistical characterization (or,in effect, "how to create and validate your own program behavior model"). Thesetechniques will outlive many current models. On the other hand, one should notunderestimate the difficulties in actually applying the techniques herein to anew model.This book is intended for two principle classes of readers: those interested inapplying program behavior models in their systems modeling efforts, and thoseinterested in program behavior, per se, for research or study purposes. I haveattempted to survey the current state of the art in modeling and in measuringprograms. Some results are new, having never before been published, but moreof them are gathered from the computer science and the statistical literature.The subtitle Models and Measurements indicates a primary themethat modelsrequire measurements for their validation, and conversely, that measurementsrequire models for their interpretation and application. Program behavior islargely an empirical study.The organization of the book is as follows: Chapter 1 motivates programbehavior modeling and measurement, and introduces several philosophical dis-tinctions to be used later. The principle of locality, the most basic programbehavior effect, is also introduced. Chapter 2 deals with those features of pagingand paging systems relevant to later discussions: hardware, paging, algorithms,and a more detailed examination of the properties of locality. Chapter 3 presentsseveral basic models, and chapter 4 elaborates on one of the most importantmodelsthe LRU stack model and its extensions. Chapter 5 covers techniquesfor program measurement and model validation, including an introduction tospectral analysis, with many examples. Some of the more advanced of these