STOCHASTIC ACTIVITY NETWORKS: STRUCTURE, BEHAVIOR, AND APPLICATION.

Stochastic activity networks (SANs), a generalization of stochastic Petri nets, permit the representation of concurrency, timeliness, fault tolerance, and degradable performance in a single model. Two important subclasses of SANs are those whose state behavior may be characterized as a semi-Markov process and those whose state behavior may be characterized as a Markov process. Conditions under which stochastic activity networks exhibit these two properties are investigated through the introduction of a higher-level model, a stochastic activity system (SAS), which allows for systematic formulation and investigation of SAN behavior. The use of stochastic activity systems permits the authors to state conditions under which Markov and semi-Markov behavior is exhibited and specify the nature of this behavior in terms of the SAS realized by SAN. Application of these models to the construction of performability models is also addressed.