| From: | Andres Freund <andres(at)anarazel(dot)de> |
|---|---|
| To: | Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> |
| Cc: | Julien Rouhaud <rjuju123(at)gmail(dot)com>, "Tharakan, Robins" <tharar(at)amazon(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: track_planning causing performance regression |
| Date: | 2020-06-29 23:10:15 |
| Message-ID: | 20200629231015.qlej5b3qpfe4uijo@alap3.anarazel.de |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On 2020-06-29 17:55:28 +0900, Fujii Masao wrote:
> One idea to reduce that lock contention is to separate per-query spinlock
> into two; one is for planning, and the other is for execution. pgss_store()
> determines which lock to use based on the given "kind" argument.
> To make this idea work, also every pgss counters like shared_blks_hit
> need to be separated into two, i.e., for planning and execution.
I suspect that the best thing would be to just turn the spinlock into an
lwlock. Spinlocks deal terribly with contention. I suspect it'd solve
the performance issue entirely. And it might even be possible, further
down the line, to just use a shared lock, and use atomics for the
counters.
Greetings,
Andres Freund
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Peter Eisentraut | 2020-06-29 23:14:31 | Re: pgstattuple: Have pgstattuple_approx accept TOAST tables |
| Previous Message | Andres Freund | 2020-06-29 23:00:19 | Re: track_planning causing performance regression |