I just thought I would share some of my own (messy) experiments with benchmarking file uploading. (I currently really need to upload a bunch of data so I am motivated to improve this.)
A key issue is to try operations with different dask.get options (single-threaded, multithreaded, multiprocessing, distributed). They can behave very differently.
I just thought I would share some of my own (messy) experiments with benchmarking file uploading. (I currently really need to upload a bunch of data so I am motivated to improve this.)
https://gist.github.com/rabernat/ba26802071271f088f5b4f8f9f5db81d
A key issue is to try operations with different dask.get options (single-threaded, multithreaded, multiprocessing, distributed). They can behave very differently.