Performance analysis & load balancing of linux clusters

Application with massive amount of computation requires a High-Speed Processor e.g. Super Computer to complete processing in reasonable time. The cost of Super Computer is not affordable for many levels of users/organizations. An alternate to this is use of Clustering Technology which is cost-effective in comparison to Super Computer. Cluster offer many advantages such as to make use of idle cycles of all nodes in it, overall low cost, High Availability, and are more scalable (up to Thousands of nodes) than any other parallel computer architecture. Nowadays many software systems exist for cluster parallel computing, like PVM (parallel Virtual Machine), MPI (Message Passing Interface), AFAPI (Aggregate Function API). When Parallel application is built with this for problems with large amount of computation & executed over small/medium size cluster, it is highly necessary that Initial Assignment of parallel job is done in such a way that it places an approximately even load on all nodes of clusters.