A macroscopic model of an interactive computing system

Abstract

The development of effective performance evaluation techniques for interactive systems is a complex problem [1,2]. Numerous papers have presented evaluation models based upon queuing theory [3,4]. This paper presents an alternative model which is quite general since it is independent of the microscopic properties of any given interactive system. By supplying to the model from four to seven experimentally determined parameters for a given interactive system, we can compute a numerical throughput optimization factor, and also a demand optimization factor in the case of time-dependent demands from the terminals. Additionally, for systems meeting a small number of restrictions, the model predicts quantitative relationships between the throughput and the implementation of techniques such as that of dynamically varying the scheduling algorithm.