Название: Computer Performance ModelingАвтор: Alexandre Brandwajn, Thomas Begin
Издательство: World Scientific Publishing Company
Год: 2025
Страниц: 355
Язык: английский
Формат: pdf (true)
Размер: 16.8 MB
This compendium gives an overview of computer performance modelling techniques and an in-depth explanation of their principles and practical applications. The book emphasizes selected approximation methods for complex systems, including recent advances in multi-server models.
This text incorporates many years of experience doing research, teaching, and practicing computer performance modeling. This book can be used as a textbook for a graduate-level course in computer performance modeling, as well as a reference for professionals working in this field.
Computer performance modeling provides tools to predict the performance of computer systems before they are built, as well as the effect of system modifications before they are released to the users. Ideally, outside emergency situations, system designs and modifications whose performance has not been assessed and is thus unknown, should not find their way into the marketplace. Computer performance modeling includes all aspects of computer systems, from their hardware and software components to their overall architecture. When dealing with existing systems, its goals include the determination of critical performance factors and the impact of projected changes. For hypothetical systems, performance modeling during the system design phase allows the designers to make critical choices and separate good solutions from poor ones.
Loosely speaking, there are two categories of computer performance models: deterministic and probabilistic. In the first category, we assume that all system parameters, such as job (or task) execution times and sequencing, are known exactly. Such deterministic models are mostly used in the area of real-time scheduling and control systems and are outside the scope of this book. In the second category, many workload and some system operation characteristics are represented through probability distributions. The resulting probabilistic model can usually be described by a set of mathematical equations. This book is devoted to selected aspects of probabilistic computer performance models.
The first chapter of this text defines the notion of a model, discusses in general terms different classes of approaches used to solve a computer model, and stresses the importance of prediction as the ultimate practical goal of computer modeling. This chapter is recommended for all users of this text.
In our second chapter, we use a simple high-level model of a transaction processing system to introduce essential terminology, as well as a common practical way to derive balance equations. We also cover the important Little’s formula. Refinements of this basic model lead us to consider models with finite population and state-dependent service rates. We use this opportunity to develop some insight into the applicability of open models in practice. We use a model of systems where no queueing is allowed to have a first look at the computation of the state seen by an arriving customer from the standard steady-state probabilities. Unless one is using this text to teach only selected advanced topics in computer performance modeling, Chapter 2 is recommended for all users of this text.
The useful reference text benefits professionals, researchers, academics and graduate students in the fields of computer engineering, operations research and mathematical modeling.
Скачать Computer Performance Modeling