4 ms·
Postgres storage engines are not mature yet to allow higher through put for insert heavy scenarios (That are not necessarily analytics focused). They also don't
by tehlike 2y ago
Postgres storage engines are not mature yet to allow higher through put for insert heavy scenarios (That are not necessarily analytics focused). They also don't have a way to offload storage to cheaper storage.
Orioledb, pg_analytics, hydra are getting there, but nothing out of the box postgres yet.
Compression also nearly non-existent except for TOAST.
- philippemnoel 2y agoThis is very much a work-in-progress in the ecosystem, though. We (ParadeDB) also have `pg_lakehouse`, which is a version of `pg_analytics` designed to work over cloud object stores (AWS S3, GCS, etc.) for "cheaper storage". I don't know if columnar engines will make it to core Postgres (I doubt it), but I do see the work of OrioleDB making it to core someday. We've joined an "alliance", alongside the core OrioleDB folks and some people from Timescale and Yandex, to push for it.
- tehlike 2y agoI have been following your work with much joy - so kudos to you. I have also some dream that one day we will see something like rocksdb/lsm backing postgres.
- digitology 2y ago[flagged]
- jamesgresql 2y agoJust out of interest what would you say a 'higher throughput' was?
- tehlike 2y agoIt is very application dependent but this email thread explains one of my problems https://www.postgresql.org/message-id/315b7ce8-9d62-3817-0a92-4b20519d0c51%40postgrespro.ru https://www.postgresql.org/message-id/315b7ce8-9d62-3817-0a9...
- justanotheratom 2y agoNoticed recently that Oriole joined Supabase: https://supabase.com/blog/supabase-aquires-oriole https://supabase.com/blog/supabase-aquires-oriole
- tehlike 2y agoThat's really cool - didn't know this was the case.