New primitives for interlaced memory policies in Markov regenerative Stochastic Petri Nets

The non-Markovian Stochastic Petri Net (SPN) models appeared so far in the literature, are based on the assumption that the underlying marking process can be specified by assigning an individual memory policy to each timed transition. The present paper proposes to introduce interlaced, or state dependent, memory policies, where the memory of a transition can be modified by the occurrence of some condition on the net. Adhering to the spirit of the graphical language of the PN, we introduce new primitives in the form of suitable arcs connecting places to transitions, and whose effect is to modify the memory policy of the transition. Through a number of simple examples, we show how the new primitives increase the modeling power of non-Markovian SPN by allowing firing mechanisms which were not possible in the traditional models. Numerical results on the steady state behavior of a dependable processor system with two kinds of interruptions are also presented.