Near-complete decomposability of queueing networks with clusters of strongly interacting servers

Publisher Website
Google Scholar
Abstract
The near-complete decomposability of queueing network models of computer systems is generally supported by very large differences in the service rates of the servers. In this paper we show how such models may still be nearly completely decomposable if on the one hand these large differences can no longer be realistically assumed (as is the case, for example, in computer networks) but if on the other hand clusters of strongly interacting servers exist. Our results may be viewed as a bridge between the approaches to the approximate analysis of queueing networks advanced by Courtois and by Chandy, Herzog and Woo, since we show circumstances under which the former approach leads to exactly the same method of analysis as the latter. In contrast to the Chandy, Herzog and Woo theorem, however, the theory of near-complete decomposability does not rely on the beneficent properties of queueing networks exhibiting product form solutions. Thus our results may point the way towards the theoretically sound application of simple and intuitively appealing approximate analysis techniques to non - product form networks.

This publication has 2 references indexed in Scilit: