3 ms·
Looks like a solid approach without much downside, except for some added complexity. Without the high profile use case I can see how time and attention might be
by rch 10y ago
Looks like a solid approach without much downside, except for some added complexity. Without the high profile use case I can see how time and attention might be better spent elsewhere, but this seems like a good idea at present.
- gamache 10y agoI think it makes a bunch of sense. As described in the earlier Postgres discussion around Uber's problem, it's not all that weird a use case -- just a difficult one. Summary: a table with around 50,000 rows that gets joined to other tables a lot, with one index per column, and that table receives ~500 writes a second, causing a "write amplification" because all the indexes need updating constantly. It's not hard to imagine lots of other workloads involving patterns like that. Postgres has most simple use cases solved, so IMHO improving complex patterns like this is good for Postgres' future.