Performance Evaluation of Parallel Applications Using Message Passing Interface in Network of Workstations of Different Computing Powers

A network created from so called commodity hardware can effectively and affordably be used for high performance computing. Major attraction from such networks is cost-effectiveness due to efficient utilization of existing hardware. By combining processing powers of many scattered workstations, a large computational problem can be solved within desirable time constraint, which is comparable to that of a proprietary or dedicated parallel computer. Popularly such systems are known as Cluster or Network of Workstations (NOWs). Among the fastest computers in the world, many of them have been developed by using the same technology. In this paper, we evaluate performance of parallel applications on cluster of nodes having different computing powers in terms of hardware attributes/parameters. We also investigates impact on execution time after running parallel application on selected favorable nodes, in terms of computing power and compare it with performance of application executed on random nodes.