Multiprocessor performance
Parallel discrete event simulation
Communications of the ACM - Special issue on simulation
Environment partitioned distributed simulation of queueing systems
WSC '91 Proceedings of the 23rd conference on Winter simulation
Concurrent simulation: an alternative to distributed simulation
WSC '86 Proceedings of the 18th conference on Winter simulation
Simulation modeling with event graphs
Communications of the ACM
Fundamentals of Performance Modeling
Fundamentals of Performance Modeling
| Hi-index | 0.00 |
One approach to speed up a discrete-event simulation program is to parallelize the execution of events in a pregenerated traditional simulation event list on a shared-memory multiprocessor. The input to the parallel-event simulator consists of a series of event lists, corresponding to different simulation runs, which are organized in the form of state-dependency graphs with random structures. Each graph is divided into a number of dependency levels which are executed in sequence by the multiprocessor.In this paper, we develop an analytical model for the performance of the parallel-event simulator. In particular, a closed form expression is given for a lower bound on the average response time of a simulation run. Performance is also estimated by simulating the parallel-event simulator. Performance results indicate that, for simulation problems whose event lists contain enough parallelism, the parallel-event simulator is capable of achieving good performance.