Comparison of simulation results of dissimilar models

Abstract

Analytical and simulation models for the same idealised but realistic computer system are considered. The results for CPU utilisation and CPU queue length from the models compare very favourably, suggesting that the technique could be used to confirm the accuracy of a simulation model. Increasingly dissimilar models are also systematically investigated. It is found that the degree of dissimilarity of the models can be surprisingly large before the method of comparison can be considered as having broken down.