To acclerate the compression of large datasets given that h5py doesn't support parallel compression, I use zlib to compress data parallelly and h5py to write it serially, and everything is good. To ...