The MPI message passing system was tested for performance. Table 1 shows how the LAM implementation of MPI performed for blocking sends of 128K blocks between two nodes.
Table 1: MPI blocking send single node user-level I/O rates (MB/sec).
