Re: PG 8.3 and large shared buffer settings

At 12:36 AM -0400 9/25/09, Tom Lane wrote:
>Dan Sugalski <dan(at)sidhe(dot)org> writes:
>> Is there any practical limit to the number of shared buffers PG 8.3.7
>> can handle before more becomes counter-productive?
>
>Probably, but I've not heard any definitive measurements showing an
>upper limit. The traditional wisdom of limiting it to 1G or so dates
>from before the last rounds of revisions to the bufmgr logic.

Excellent.

> > My production DB's around 200G, and the box hosting it has 192G of
>> memory on it, running a 64 bit AIX build of 8.3.7.
>
>Yowza. You might be able to do measurements that no one has done
>before. Let us know what you find out.

:) It's a machine of non-trivial size, to be sure. I'll give the
buffer setting a good bump and see how it goes. I may be able to take
one of the slony replicas off-line the next holiday and run some
performance tests, but that won't be for a while.

>BTW, does AIX have any provision for locking shared memory into RAM?
>One of the gotchas for large shared memory has always been the risk
>that the kernel would decide to swap some of it out.

I'll have to go check, but I think it does. This box hasn't actually
hit swap since it started -- a good chunk of that RAM is used as
semi-permanent disk cache but unfortunately the regular day-to-day
use of this box (they won't let me have it as a dedicated DB-only
machine. Go figure :) doing other stuff the cache tends to turn over
pretty quickly.
--
Dan

--------------------------------------it's like this-------------------
Dan Sugalski even samurai
dan(at)sidhe(dot)org have teddy bears and even
teddy bears get drunk