Performance evaluation using unbounded timed Petri nets

Abstract

Unbounded timed Petri nets are place-unbounded free-choice place/transition nets with exponentially distributed firing times associated with transitions of a net. In such nets the finite state space is generated by a finite set of linear equations. The regularity of this linear description can be used for a 'projection' (or 'folding') of the infinite state space into an equivalent finite representation that can be described by a finite set of nonlinear equilibrium equations. The solution of these equations determines the stationary probabilities of the states. Many performance measures can be obtained directly from this stationary solution. Such unbounded nets can eliminate the state explosion problem of some models by using unbounded but simple approximations to bounded but complex models.