| Recent advances in metacomputing such as volunteer and desktop grid computing that aggregate loosely coupled resources have transformed the execution of certain computational workloads that, in the past, were reserved for processing on dedicated clusters. Parallel discrete event simulations have different requirements than programs that can readily exploit loosely coupled resources such as embarrassingly parallel codes. Consequently, parallel discrete event simulations are typically run on tightly coupled machines providing the best opportunity for maximum speedup. However, these facilities may not be readily available to many users. The focus of this thesis explores the merging of these distinct computational domains involving the execution of parallel discrete event simulation across loosely coupled resources. A master/worker architecture for parallel discrete event simulation is proposed providing robust executions under a dynamic set of services with system-level support for fault tolerance, semi-automated client-directed load balancing, portability across heterogeneous machines, and the ability to run codes on idle or time-sharing clients without significant interaction by users. Results indicate that a master/worker approach utilizing loosely coupled resources is a viable means for high throughput parallel discrete event simulation by enhancing existing computational capacity or providing alternate execution capability for less time-critical codes.;Research questions and challenges associated with issues and limitations with the work distribution paradigm, targeted computational domain, performance metrics, and the intended class of applications to be used in this context are analyzed and discussed. A portable web services approach to master/worker parallel discrete event simulation is proposed and evaluated. Optimizations to increase the efficiency of large-scale simulation execution through distributed master service design and intrinsic overhead reduction are proposed and evaluated. Finally, challenges for optimistic parallel discrete event simulation such as rollbacks and message unsending with an inherently different computation paradigm utilizing master services and time windows are addressed and evaluated. |
