Implementing the Mean Value Algorithm for the Solution of Queueing Network Models

The Mean Value Analysis (MVA) algorithm of Reiser and Lavenberg is a very simple technique for the solution of fairly complex queueing network models: This paper describes a computer program which uses the MVA algorithm to solve almost all known versions of closed queueing networks which have 'product form' solutions. Introduction Queueing network models of computer systems are emerging as both a tool useful in analyzing and predicting system performance and as a topic of research [BCMP75]. Until 1978, the most popular technique for obtaining (exact) solutions of these networks was the convolution technique [ReKa75. BaBS77, BrBaBO]. In 1978. Reiser and Lavenberg [ReLa7B] introduced a new technique, based on a. Mean Value Analysis, which could be used to solve these models. This MYA technique is simpler to describe and implement than the convolution technique. Also, the convolution technique usually required a step to obtain performance parameters w~ch resulted in a numerically unstable algorithm. The MVA technique does not have such a step and is thought to be more stable [CbS.BO]. LC)