Performance Prediction Of Parallel Applications On Networks Of Workstations.
Daniel A. Menascé, Amar Rao · 1996
Networks of workstations (NOWs) are an alternative to supercomputers when it comes to running parallel applications. In most environments, users of a workstation use less than 20% of the available CPU cycles. The remaining capacity can beused to support tasks of aparallel application. This paper presents a hybrid analytic and simulation model that can beusedtopredict the performanceofparallel applications specified by a task graph on a network of workstations. The model was validatedthrough measurements and takes into account the specific workstations used, the scheduling discipline, and the network characteristics. Moreover, it considers the effect of theowner' interactive workload on the execution time of the parallel application. The paper briefly discusses a tool developed by the authors to implement the model.