A generalized stochastic high-level Petri net model for performance analysis

Abstract

We present a generalized stochastic high-level Petri net (GSHLPN) model for evaluating the performance of parallel/distributed systems. In our model, we have classified the transitions of the predicate/transition nets (PrT-nets) into two types: timed predicate transitions with exponentially distributed firing times and immediate predicate transitions with zero firing times. Also, the PrT-nets are extended by including inhibitor predicate arcs. The motivation for introducing these new constructs in the GSHLPN model is discussed. The model is developed to be a hybrid of the PrT-net and generalized stochastic Petri net models. We define the GSHLPN model and show how performance estimates are obtained from the GSHLPN model. To manage the complexity of large systems, we have reduced the size of the reachable state space by combining the GSHLPN model with the compound marking technique (CMT). We show how the GSHLPN model with the CMT can be useful for the performance analysis of complex systems, such as hypercube multicomputer systems. Finally, we investigate another approach to further reduce the size of the state space of this model by applying two different types of aggregation techniques on the GSHLPN model: time scale decomposition technique and CMT. This aggregation concept is illustrated through the example of a degradable Star multicomputer system.