donutdb
donutdb copied to clipboard
idea: train zstd dictionary for the sectors
Since the sectors are pretty small, could probably achieve better compression ratios by training dictionary on bunch of raw sectors