The HDF Group is looking for ways to get better support for HDF5 on HPC platforms and I'd like to talk to developers who are using HDF5 to perform parallel I/O from their application code or benchmark, or those who would like to do so. We're looking for cases where the performance of using HDF5 isn't matching up with what the underlying system is capable of and improving HDF5's parallel I/O performance or adding new features to HDF5 would address the issue. If you've got a good example, please contact me directly.
Director of Software Development
The HDF Group