A discrete time model for performance evaluation and correctness verification of real time systems

Abstract

In the development of reactive systems, correctness and performance requirements are often intertwined. An integral approach to both the aspects of the problem requires models and analysis techniques which jointly capture probabilistic and minimum/maximum characterization of time delays. A discrete time extension of Time Petri Nets is introduced, which extends minimum/maximum timing constraints with a discrete probability density. As a characterizing trait, the model is interpreted according to a semantics of true concurrency, representing dynamic behavior in terms of the probability that a set of transitions fires at the current tick of the clock. Behavioral and probabilistic rules in the semantics of the model are soundly captured into an enumerative technique which represents the state space of the model as a transition system labeled with delays and probabilities. This enables both Markovian analysis techniques and inspection algorithms supporting verification of sequencing and minimum/maximum durational properties.