-
Improvement
-
Resolution: Fixed
-
Normal
-
None
-
None
-
None
-
None
After working on https://github.com/metabrainz/acousticbrainz-server/pull/311 I had some thoughts about statistics.
By keeping stats every hour we're using a lot of space for what I see as not much value.
We spend a lot of time calculating stats, storing them, transferring them (which takes a lot of bandwidth), and then we throw them away and only show daily values on the stats graph anyway.
I think we could work on a few cleanups to store this data better:
- Use json fields to store all numbers for a stats time period together. This will make it easier to write (no possibility of missing a value) and easier to read (just read and return)
- Consider only storing daily data. If the graph by default shows only daily data then we don't need hourly data
- If we do store only daily data and we want to keep hourly updates on the front page we could store hourly data for the last day in redis.