Performance optimization of PVM based parallel applications using optimal number of slaves

Parallel computing operates on the principle that large problems can often be divided into smaller ones, which are then solved concurrently which results in saving time to solve larger problems and to provide concurrency using desktop PC's. The main aim is to form a cluster based parallel computing architecture for demonstrating PVM based parallel applications which works on the Master-Slave computing paradigm. The master will monitor the progress and be able to report the time taken to solve the problem, taking into account the time spent in breaking the problems into sub-tasks and combining the results along with the communication delay. The slaves are capable of accepting sub problems from the master and finding the solution and sending back to the master. We aim to evaluate these time statistics of parallel execution for solving matrix multiplication problem and find the relation between the number of cores and number of slaves utilized for computation. When the number of nodes required for the computation is fixed by the user, the computation time mainly depends on the number of slaves specified for computation. In our work, we find the optimal number of slaves required for PVM based parallel computation when the number of nodes is fixed by a user. The analysis is made for the computation of different sizes of matrices over the different number of nodes.