TO PREDICTION OF PERFORMANCE IN PARALLEL ALGORITHMS

With the availability of powerful PC and networking devices, the recenttrend in parallel computing is to connect a number of individual workstations to solvecomputation intensive tasks in parallel way on connected class of workstations.Current trends in high performance computing are to use networks of workstations asa cheaper alternative to traditionally used massively parallel multiprocessors orsupercomputers. The individual workstations as parallel computers based on modernsymmetric multicore implemented within workstation. To exploit the parallelprocessing capability of such cluster, the application program must be paralleled. Onapplication ex. we demonstrate the various influences in process of performanceprediction modelling and the consequences for their parallel implementations.