On bounds for token probabilities in a class of generalized stochastic Petri nets

Methods are presented for computing tight bounds for steady-state token probabilities of a class of generalized stochastic net (GSPN) models. Such bounds also give a better estimate of the error produced when decompositions and aggregations are used to compute the various performance measures. First a method is described to compute the best lower and upper bounds for conditional token probabilities of a class of GSPN when only the subsequent is considered. The authors show that such bounds can be improved if additional information about other subnets is available. They extend the technique and outline an algorithm to compute the bounds for error due to aggregation and decomposition at the GSPN level. An example is presented to illustrate the technique and algorithm