Approximate techniques for modeling the performance of complex systems

• Authors: John Spragins
• Venue: Computer Languages
• Year: 1979

This survey paper discusses a variety of approximate technique for analyzing and predicting the performance of complex systems, such as computer networks. Such techniques tend to the more readily usable by the persons in charge of design, development, selection, or tuning of such systems than the great majority of the techniques available in the current literature. The discussion is motivated by the author's experience in industry for nine years, during which time he saw numerous instances where performance models gave good results when they were applied in situations where a large number of the assumptions made in developing the models were demonstrably false. These results have motivated him to study the current literature on performance modeling to find out why such models work as well as they do. This has led to the results discussed here. A large portion of the paper is a survey of some of the most interesting techniques for approximately modeling system performance that have been developed recently, along with some theoretical developments which indicate why highly simplified models often work surprisingly well. Since it is rare to find situations where the assumptions made in developing models are really valid, simplified models are normally as accurate as more elaborate models. A few examples of successful use of approximate models are discussed next. The paper concludes with a summary of the most important principles developed and an indication of some of the approaches to approximate modeling which the author feels merit further research.