MPI-IO on a parallel file system for cluster of workstations

Since the MPI-IO definition, a standard interface for parallel IO, some implementations are available for clusters of workstations, but the performances are within the limits of the file system (typically NFS). New parallel file systems are now available on clusters of workstations and provide higher performances. A first idea is to interface such a parallel file system to a portable implementation of MPI-IO. We present first experimental results of a straightforward port of ROMIO, an MPI-IO implementation from Argonne National Laboratory, on PVFS, a parallel file system for cluster from Clemson University. We measure the performances for typical file accesses found in data-parallel applications, and we compare the results with the performances of ROMIO on NFS.