Optimistic protocol analysis in a performance analyser and prediction tool

Abstract

Parallelising sequential discrete event simulation programs is often a tedious process, with no guarantee for speedup. This paper describes a performance analyser tool developed to predict the execution performance of parallel discrete event simulation programs, using either optimistic or conservative protocols. The tool aims to provide the user with the flexibility of choosing the simulation protocol desired, and to make useful comparisons between the results obtained. The novel feature of the analyser is that it predicts the parallel execution performance based on a sequential simulation run, without the parallel implementation developed. This paper describes the structure and operation of the analyser program focusing on the analysis of the optimistic simulation protocol. As the prediction of the optimistic protocol is very different from the conservative one, its uniqueness will be illustrated in this paper. The performance analyzer and prediction tool can be used to explore different protocols and parameters in order to provide faster decision making about the final implementation strategy.