A comparison between different parallelization methods on workstation clusters to Solve CFD-problems

Abstract

The efficient parallel solution of flow problems on parallel computers requires highly efficient numerical methods as well as highly efficient parallelization methods. Parallelization is mostly done using domain decomposition methods. With the introduction of the combination method a big jump in computational speed is possible. Here, a numerical solution is computed on a sparse grid. The algorithmic concept is based on the independent solution of many problems with reduced size and their linear combination. So, this method is perfectly suitable for parallelization. In this paper we compare both very efficient methods. We focus on the parallel solution of typical problems arising in the field of flow simulations. A comparison is done using the overall runtimes of both methods as well as the parallel speedup. The implementations are done on a workstation cluster of 16 HP machines.