Horton

Re: heavily contended lwlocks with long wait queues scale badly

Alexander Korotkov <aekorotkov@gmail.com>

View thread View on postgr.es
From: Alexander Korotkov <aekorotkov@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, pgsql-hackers@postgresql.org, Thomas Munro <thomas.munro@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>, Pavel Borisov <pashkin.elfe@gmail.com>
Date: 2022-10-31T12:38:46Z
Lists: pgsql-hackers 
Hi Andres,

Thank you for your patch. The results are impressive.

On Mon, Oct 31, 2022 at 2:10 PM Pavel Borisov <pashkin.elfe@gmail.com> wrote:
>
> I was working on optimizing the LWLock queue in a little different way
> and I also did a benchmarking of Andres' original patch from this
> thread. [1]
> The results are quite impressive, indeed. Please feel free to see the
> results and join the discussion in [1] if you want.
>
> Best regards,
> Pavel
>
> [1] https://www.postgresql.org/message-id/flat/CALT9ZEEz%2B%3DNepc5eti6x531q64Z6%2BDxtP3h-h_8O5HDdtkJcPw%40mail.gmail.com

Pavel posted a patch implementing a lock-less queue for LWLock. The
results are interesting indeed, but slightly lower than your current
patch have. The current Pavel's patch probably doesn't utilize the
full potential of lock-less idea. I wonder what do you think about
this direction? We would be grateful for your guidance. Thank you.

------
Regards,
Alexander Korotkov

Commits

  1. lwlock: Fix quadratic behavior with very long wait lists