Monte Carlo integration in a distributed heterogeneous environment
J.P. Dougherty · 2002
The potential of distributed processing using existing topologies and processing nodes is explored with a sample Monte Carlo application. It is demonstrated how Monte Carlo algorithms map well onto a heterogeneous topology of processing nodes, operating systems, and communication protocols. Using the Synergy system for distributed processing an application for performing Monte Carlo integration is developed to demonstrate how such techniques exhibit scalability, parallelize naturally, and realize significant speedup over sequential versions. Coordination is asynchronous via tuples, and distributed across a combination of LANs and MANs. Processors are heterogeneous and nondedicated. Results indicate that granularity affects performance and that a significant percentage of potential processing power can be captured.>