In: Proceedings of the 2nd IMACS International Multiconference on Computational Engineering in Systems Applications (CESA'98). April 1998. Published on CD-ROM.
Abstract: Stochastic Petri nets is a well-known formalism adequate for the design, validation, and performance evaluation of discrete event and manufacturing systems. In this paper, we deal with steady-state throughput approximation of complex concurrent systems modelled with stochastic Petri nets. More precisely, we generalize to arbitrary stochastic P/T systems a response time approximation technique that was firstly proposed for special net subclasses. The presented technique is based on the divide and conquer principle and it is achieved in two steps. The first one, a net-driven decomposition of the model into several subsystems and the second one, an iterative solution algorithm that computes a throughput approximation of the original model transitions based on the solution of the embedded continuous time Markov chain of the subsystems. Experimental results on several examples generally have an error of less than 5%, and the state space is usually reduced by more than one order of magnitude.