Symbolic techniques for the performance analysis of generalised stochastic petri nets

Abstract

Binary Decision Diagrams (BDDs) have been successfully used in sequential circuit theory, VLSI, and model checking. They form a highly memory efficient canonical representation of a Boolean function. In this dissertation, following on the success of BDDs in other fields, we investiage the applicability of symbolic techniques in the performance analysis of timed transition systems, particularly those of Generalised Stochastic Petri Nets (GSPNs). We make use of symbolic methods, where states are represented implicitly rather than explicitly, primarily to conserve memory during the state space exploration process - a necessary step in the performance analysis pipeline. We have investigated the use of BDDs in two different ways. The first, our own novel technique, allows the user to effectively place an upper bound on the amount of memory to use during state space exploration. The second makes use of transition to find the successor states at each level of the state graph. Both of these techniques rely on a novel and efficient GSPN to BDD encoding function that we have derived.