GSPNs Revisited: New Semantics and Analysis Algorithms

This talk shows how Markov Automata (MA) can be used to provide a truly simple semantics of Generalized Stochastic Petri Nets (GSPNs), a popular model in performance and dependability analysis that exists for more than 25 years. In fact, our approach works for all GSPNs. No restrictions are imposed on the concurrent/conflicting enabledness of immediate transitions. This contrasts with existing solutions for GSPNs. We complement the semantics by novel analysis algorithms for expected time and long-run average time objectives of MA, i.e., GSPNs. Two case studies indicate the feasibility of these algorithms and show that a classical reliability analysis for confused GSPNs may lead to significant over-estimations of the true probabilities. The key message is: nondeterminism is not a threat, treat it as is! This yields both a simple GSPN semantics and trustworthy analysis results.