Prof. Dr. Joost Pieter-Katoen

"Next Generation Stochastic Petri Nets"

Generalised Stochastic Petri Nets (GSPNs) originated in the mid 1980s and have been widely used in performance and dependability analysis. Mature tool support for the analysis of GSPNs exists. Extensions of UML, business process modeling languages such as BEPL, have been all defined in terms of GSPNs.

There is however an important caveat: the analysis and semantics of GSPNs is restricted to so-called "well-defined" nets -- nets that lack non-determinism. To overcome this limitation, several GSPN extensions and checks have been defined in the past. They all have their limitations though.

In this talk, we cover every GSPN. This includes nets that are not "well-defined". We present a semantics for every GSPN and show that for "well-defined" nets our approach is equivalent to the existing one. Thus all analysis results, all tools, and all algorithms for "well-defined" nets still apply. As our approach covers a larger class of nets, the question arises how to analyse the non "well-defined" nets. We will present algorithms, inspired by stochastic game theory, and give experimental results on their usage.