Estimating the benefit of the parallelisation of discrete event simulation

Abstract

This paper presents a technique which attempts to aid the simulationist in the decision as to whether or not simulation should be implemented on a multiprocessing computer. The proposed technique has been used to estimate the performance of parallel discrete event simulations. This employs critical path analysis to determine the lower bound of the execution time of a parallelised simulation and has been used by other authors to study the effect that process scheduling and causality maintenance protocols have on performance. The contribution of this paper is the extension of this technique to a class of common simulation models which demand resource sharing. This forces a parallel implementation to use some kind of information exchange protocol. A case study is presented which illustrates the potential usefulness of this technique. This employs a performance analysis tool developed for this purpose which has been implemented using the simulation package SES/Workbench.