Grid resource selection by application benchmarking for computational haemodynamics applications

Grid benchmarking for improved computational resource selection can shed a light for improving the performance of computationally intensive applications. In this paper we report on a number of experiments with a biomedical parallel application to investigate the levels of performance offered by hardware resources distributed across a pan-European computational Grid network. We provide a number of performance measurements based on the iteration time per processor and communication delay between processors, for a blood flow simulation benchmark based on the lattice Boltzmann method. We have found that the performance results obtained from real application benchmarking are much more useful for running our biomedical application on a highly distributed grid infrastructure than the regular resource information provided by standard Grid information services to resource brokers.