Timed Petri net models of queueing systems

It is shown that for timed Petri nets with exponentially distributed firing times (M-timed Petri nets), the state space can be generated directly from net specifications, and then the stationary probabilities of states can be obtained by standard methods, developed for analysis of (continuous-time) Markov chains. Numerous performance measures can be derived from stationary probabilities of states. For unbounded nets (models of open network systems are usually unbounded), the state space is infinite. A transformation is thus needed that folds this infinite space into a finite representation, used for effective evaluation of probabilities. The author presents a short theoretical background for performance evaluation using timed Petri nets, followed by several examples of closed and open network models of simple computer systems.<
>