I’d love to have pytables and pandas integrated with HDF Server, but that would take some work. A lot of effort went into h5pyd and it’s working pretty well as a h5py drop in.
So, I think the best approach would be to first refactor pytables to use h5py. This is what Andy Scopatz proposed earlier: https://www.hdfgroup.org/2015/09/python-hdf5-a-vision/. Unfortunately, not much progress has been made in this effort.
Anyway, once we have pytables on h5py, it would be easy to have pytables use h5py or h5pyd depending on if the file is a local posix file, or a HDF Server-based content. And from there, pandas support should come for free as well.
For the question about production environments – the project is relatively new, so please exercise reasonable precautions – Backup any data you put into h5serv, follow best practices for security, don’t use for nuclear power plant control systems ( , etc.