Approximate Mean Value Analysis of Process Algebra Models

Studying the existence of product forms of performance models described with compositional techniques is of central importance since this may lead to particularly efficient solution methods. This paper considers a class of models in the stochastic process algebra PEPA which do not enjoy the exact product form solutions available in the literature. However, they can be interpreted as queueing networks with service vacations and multiple resource possession, which have been shown to admit accurate analytical approximations based on mean value analysis. Special attention is devoted to situations where the use of the competing approximate method based on ordinary differential equations may be questionable due to the presence of components with few replicas.