Possibilities of Workstation Clusters for Parallel Finite–Element Analysis

Since the computing speed and cost performance of engineering workstations are being improved, parallel finite–element analysis is expected using workstation clusters. Parallel performance of the gaussian elimination method, the domain decomposition method, and the domain partitioned CG method are estimated and compared for various workstation clusters with different computing and communication performances. The accurate estimation is based on basic performance data and models to predict the execution time. Consideration is given to the possibilities of the workstation clusters and to the performance required for the workstation cluster system.