Abstract
Stochastic models of contention for shared resources in an experimental multiprocessor prototype are presented and are validated with simulation and measurement results. Three modeling techniques are used (stochastic Petri nets, Markov chains, and queueing networks) that represent the system operations as Markovian stochastic processes. Each technique is best suited to a specific stage of the analysis. An integrated use of these techniques represents a very powerful tool for the performance analysis of multiprocessor systems and provides ways of investigating several extensions of the prototype architecture. Simulation results and measurements performed on the hardware prototype validate the analysis and show that the accuracy of the analytical results is excellent.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call