Performance prediction for parallel scientific application

In this paper we studied performance predictions for parallel scientific applications on a homogeneous cluster of workstations. Performance prediction is important for analyses of the scalability of parallel applications and the estimation of the processing time for the application in a loaded/unloaded environment. We developed Automatic Parallel Application Prediction System (APAPS) for cluster computing environments. Here we are reporting the accuracy of the APAPS using two scientific applications. The measured result shows that APAPS has high prediction accuracy