I have a use case in which a microserver (in Go) has read-only access to a potentially very large SQL database containing events. The microserver transforms event slices into various things that are suitable for analyses by a (Python) analytical server, including time series, vectors, (associative) arrays of various sorts and so on. The transformed data is not persisted, but is consumed by things like numpy and scipy in the analytical server. My question is, would using HDF5 as a transfer format from the data transformation server to the analytical server make any sense?
1 Like
Since you set such a low bar – does it make any sense – my answer is: yes, of course. 
It is even possible to access in-memory HDF5 file. What other data formats are you considering?
Aleksandar