This research will develop new analytical techniques for Stochastic Petri Nets which are both efficient and exact. The methodology will provide performance evaluation for realistic models of distributed and parallel computer systems which is not currently possible without simulation. The results of this research will provide algorithms for the efficient evaluation of Stochastic Petri Net models. These algorithms have a computational complexity similar to the product form techniques for queuing network models. Evaluation of the applicability of the theoretical framework will be accomplished using a prototype design system, SPAN, which is already under development. Justification and Recommendation A relatively young investigator is studying Stochastic Petri Net models. If successful the research could lead to improved modelling and understanding distributed and parallel computations. Support is recommended.
