Conclusion




The obtained results are encouraging as we got similar computation times on the cluster and a parallel machine. For the tested cases the use of 2 processors brings a significant gain of computation time. The full adantage of a cluster appears with larger problems, it is therefore annoying that the 3D case could not be tested. However, the communication is clearly the limiting factor for the speed-up, and the network intalled (2 network cards for 20 stations) seems to be unsufficient for large clusters.