Reducing probabilistic timed petri nets for asynchronous architectural analysis

Abstract

This paper introduces structural reductions of probabilistic timed Petri nets that preserve a large class of performance measurements. In particular, the paper proposes a class of reductions that preserve efficiently computable bounds of statistics of time-separation of events (TSEs). It identifies two specific reductions within this class. It demonstrates the utility of these reductions by reducing a detailed Petri net describing the four-phase protocol of a well-known asynchronous pipeline template into a simpler two-phase architectural-level Petri net model. The benefit of this reduced model is that the run-time of subsequent TSE analysis can be greatly improved.