Distributed IO

First targeting flat files and HDF5.

For flat files, mimic the numpy save / load interface:

http://docs.scipy.org/doc/numpy/reference/generated/numpy.save.html http://docs.scipy.org/doc/numpy/reference/generated/numpy.load.html

Notes on the .npy format, and why pickle wasn't sufficient:

https://github.com/numpy/numpy/blob/master/doc/neps/npy-format.txt

Distributed IO strategies:

https://fs.hlrs.de/projects/craydoc/docs/books/S-2490-40/html-S-2490-40/chapter-ht5w8mp0-oswald-iostrategies.html