Parallel IO performance and scalability study on the PRACE CURIE supercomputer.

The results of two kinds of parallel IO performance measurements on the CURIE supercomputer are presented in this report. In a first series of tests, we use the open source IOR benchmark to make a comparative study of the parallel reading and writing performances on the CURIE Lustre filesystem using different IO paradigms (POSIX, MPI-IO, HDF5 and Parallel-netCDF). The impact of the parallel mode (collective or independent) and of the MPI-IO hints on the performance is also studied. In a second series of tests, we use a well known scientific code in the HPC astrophysics community: RAMSES, which a grid-based hydrodynamics solver with adaptive mesh refinement (AMR). IDRIS added support for the 3 following parallel IO approaches: MPI-IO, HDF5 and Parallel-netCDF. They are compared to the traditional one file per MPI process approach. Results from the two series of tests (synthetic with IOR and more realistic with RAMSES) are compared. This study could serve as a good starting point for helping other application developers in improving parallel IO performance.