Re: Survey: Max TPS you've ever seen - Mailing list pgsql-performance

From Mark Kirkwood
Subject Re: Survey: Max TPS you've ever seen
Date
Msg-id 54DD32F6.6010100@catalyst.net.nz
Whole thread Raw
In response to Re: Survey: Max TPS you've ever seen  ("Gudmundsson Martin (mg)" <martin.mg.gudmundsson@volvo.com>)
List pgsql-performance
On 13/02/15 00:20, Gudmundsson Martin (mg) wrote:
> Hi all!
>
>> - checkpoint_segments 1000
>> - checkpoint_completion_target 0.9
>> - wal_buffers  256MB
>> - shared_buffers 31 gb
>> - max_connections 500
>
> I see that some of you are using wal_buffers = 256MB.
> I was under the impression that Postgres will not benefit from higher value than the segment size, i.e. 16MB. More
thanthat will not do/help anything. 
>
> What's the reasoning behind setting it to higher than 16MB? Do I have old information?
>
> Best regards, Martin
>

There was some discussion a while ago in which 32MB and 8MB both
demonstrated better performance than 16MB (probably related to the fact
the the default wal file size is 16MB). We just experimented further
with bigger values, and saw some improvement.

Cheers

Mark


pgsql-performance by date:

Previous
From: Kevin Grittner
Date:
Subject: Re: Configuration tips for very large database
Next
From: "Mathis, Jason"
Date:
Subject: Re: Configuration tips for very large database