Rework S3 bucket storage hierarchy / strategy?
After discussing the Pleroma S3 usage at work, we have some criticisms for long term archival of these objects especially around backup/restore. We have experience with S3 buckets that have millions of objects and many operations are painfully slow at that size.
Example of a good hierarchy:
"idx": "t/t/w/ttwi3eda/poster/4.jpg",
"idx": "t/t/z/ttz6qsiu/small.jpg",
"idx": "t/u/3/tu3zixdp/large.jpg",
"idx": "t/u/6/tu68dlz3/large.jpg",
"idx": "t/u/7/tu7j1i32/large.jpg",
"idx": "t/u/c/tucz706h/medium.jpg",
"idx": "t/u/j/tuj95zf9/thumb.jpg",
"idx": "t/u/m/tumujmfo/poster/5.jpg",
Also suggested condensing our filenames from sha hashes to something denser. "base58 instead of base16" is a recommendation as well.
Edited by feld