|From:||Stas Kelvich <s(dot)kelvich(at)postgrespro(dot)ru>|
|To:||Aleksander Alekseev <a(dot)alekseev(at)postgrespro(dot)ru>|
|Cc:||Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Robert Haas <robertmhaas(at)gmail(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>|
|Subject:||Re: Patch: ResourceOwner optimization for tables with many partitions|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
I have applied this patch and can confirm ~10% speedup by this patch in presence of big amount of inherited tables.
Test case was as suggested by Aleksander: create 1000 inherited tables, no constraints, insert a row in each one, and issue single row queries over this table.
Xeon-based server 12C/24T, 50 connections, 30-min average:
TPS, no patch: 393 tps
TPS, with patch: 441 tps
The same setup but with single table with 1000 rows give performance about 188_000 tps, so overall speed of requests over a inherited table in this scenario is quite pathological (probably this is expected because database just execute 1000 selects to each inherited table). I've also tried to set range constraints for all inherited tables, so only one table was affected by query, but planning time increased a lot and total tps was again about 500 tps.
Also attaching two flame graphs measured during tests. It’s clearly visible that PortalDrop takes x4 less time after patch.
|Next Message||Kyotaro HORIGUCHI||2016-01-21 09:26:49||Re: [PoC] Asynchronous execution again (which is not parallel)|
|Previous Message||Artur Zakirov||2016-01-21 09:15:45||Re: Fuzzy substring searching with the pg_trgm extension|