This project met most of the intended goals, though not all. The primary goal, to assemble a working cluster computer capable of parallel execution of MPI code, succeeded completely. Some of the desired benchmarks did not get performed, partially due to time constraints, and partially due to irregular results from those that did occur. Even when averaged over thousands of iterations, average and minimum time results could vary from one set of iterations to another by a factor of 10. This does not seem acceptable, but no adequate cause of this anomaly has revealed itself yet.

In general, however, all the benchmarks taken together (see the graphs page) point toward an overall level of performance comparable to that of a single processor Pentium 200MMX system. This falls short of the original predicted performance by about 50%, but partially explained by the cluster only having seven nodes rather than the originally specified eight.