Mean Value Analysis of Stochastic Petri Nets

In this paper we present a Mean Value Analysis (MVA) algorithm for the computation of performance measures in Product Form Solution Stochastic Petri Nets. Basic to the derivation of this algorithm are recursive expressions for the performance indices that are a non trivial generalization of those derived for the Mean Value Analysis of multiple class product form queueing networks. In this paper we give a formal derivation of several recursive relations as well as details on their implementation. The advantage that MVA provides for the development of approximation techniques is also investigated and a first method that gives satisfactory results is presented. A few basic examples are evaluated with the techniques discussed in this paper to show the advantages of this approach.