Re: Improving connection scalability: GetSnapshotData()

From: Andres Freund <andres(at)anarazel(dot)de>
To: Konstantin Knizhnik <k(dot)knizhnik(at)postgrespro(dot)ru>
Cc: pgsql-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: Improving connection scalability: GetSnapshotData()
Date: 2020-09-06 18:52:14
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers


On 2020-09-05 16:58:31 +0300, Konstantin Knizhnik wrote:
> On 04.09.2020 21:53, Andres Freund wrote:
> >
> > I also used huge_pages=on / configured them on the OS level. Otherwise
> > TLB misses will be a significant factor.
> As far as I understand there should not be no any TLB misses because size of
> the shared buffers (8Mb) as several order of magnitude smaler that available
> physical memory.

I assume you didn't mean 8MB but 8GB? If so, that's way large enough to
be bigger than the TLB, particularly across processes (IIRC there's no
optimization to keep shared mappings de-duplicated between processes
from the view of the TLB).

> Yes, there is also noticeable difference in my case
> | Idle Connections | Active Connections | TPS pre | TPS post |
> |-----------------:|-------------------:|--------:|---------:|
> | 5000 | 48 | 758914 | 1184085 |

Sounds like you're somehow hitting another bottleneck around 1.2M TPS


Andres Freund

In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Andres Freund 2020-09-06 18:56:19 Re: Improving connection scalability: GetSnapshotData()
Previous Message Andrey M. Borodin 2020-09-06 18:33:25 Re: Yet another fast GiST build (typo)