| From: | "Scott Marlowe" <scott(dot)marlowe(at)gmail(dot)com> |
|---|---|
| To: | "Josh Trutwin" <josh(at)trutwins(dot)homeip(dot)net> |
| Cc: | pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: Shared Buffer setting in postgresql.conf |
| Date: | 2007-10-10 20:58:54 |
| Message-ID: | dcc563d10710101358p55abb925pba894a693b21718f@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
On 10/10/07, Josh Trutwin <josh(at)trutwins(dot)homeip(dot)net> wrote:
> On Wed, 10 Oct 2007 10:20:02 -0500
> "Scott Marlowe" <scott(dot)marlowe(at)gmail(dot)com> wrote:
>
> > In 7.4, using 25% is often too high a setting for it to handle well,
> > and the practical useful maximum is usually under 10,000
> > shared_buffers, and often closer to 1,000 to 5,000
>
> Scott - interesting reply. Is this also true for 8.1? I currently
> have mine set to 16384 - server has 3.5 GB of total memory.
No, starting with 8.0, the code to manage the shared_buffers is much
more efficient with large numbers of shared buffers. With 8.0 and up
the primary considerations are that the shared_buffers be big enough
to hold your working set, but not so big as to run the system out of
memory for other things, sorts, kernel caching, etc...
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Kevin Grittner | 2007-10-10 21:02:59 | Re: hashjoin chosen over 1000x faster plan |
| Previous Message | Kevin Grittner | 2007-10-10 20:48:32 | Re: hashjoin chosen over 1000x faster plan |