| From: | Tomas Vondra <tomas(dot)vondra(at)enterprisedb(dot)com> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Andres Freund <andres(at)anarazel(dot)de> |
| Cc: | PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: postgres_fdw batching vs. (re)creating the tuple slots |
| Date: | 2021-06-04 11:48:13 |
| Message-ID: | a3233163-829d-94a0-7ca0-87bdfde92c90@enterprisedb.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
Here's two WIP patches that fixes the regression for me. The first part
is from [1], so make large batches work, 0002 just creates a copy of the
tupledesc to not cause issues in resource owner, 0003 ensures we only
initialize the slots once (not per batch).
With the patches applied, the timings look like this:
batch timing
----------------------
1 64194.942 ms
10 7233.785 ms
100 2244.255 ms
32k 1372.175 ms
which seems fine. I still need to get this properly tested etc. and make
sure nothing is left over.
regards
[1]
https://postgr.es/m/OS0PR01MB571603973C0AC2874AD6BF2594299%40OS0PR01MB5716.jpnprd01.prod.outlook.com
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tomas Vondra | 2021-06-04 11:52:28 | Re: postgres_fdw batching vs. (re)creating the tuple slots |
| Previous Message | Amit Langote | 2021-06-04 11:44:30 | Re: Skip partition tuple routing with constant partition key |