Re: random slow query - Mailing list pgsql-performance

From Scott Marlowe
Subject Re: random slow query
Date
Msg-id dcc563d10906301106n5335d5e6r8a2faf7467c942b8@mail.gmail.com
Whole thread Raw
In response to Re: random slow query  (Mike Ivanov <mikei@activestate.com>)
List pgsql-performance
On Tue, Jun 30, 2009 at 12:01 PM, Mike Ivanov<mikei@activestate.com> wrote:
> Scott Marlowe wrote:
>>>
>>> The postgres shared cache is at 4G, is that too big?
>>>
>>
>> Not for a machine with 32Gig of ram.
>>
>>
>
> He could even add some more.

Definitely.  Really depends on how big his data set is, and how well
pgsql is at caching it versus the kernel.  I've found that with a
really big dataset, like 250G to 1T range, the kernel is almost always
better at caching a lot of it, and if you're operating on a few
hundred meg at a time anyway, then smaller shared_buffers helps.

OTOH, if you're working on a 5G data set, it's often helpful to turn
up shared_buffers enough to cover that.

OTOH, if you're running a busy transaction oriented db (lots of small
updates) larger shared_buffers will slow you down quite a bit.

pgsql-performance by date:

Previous
From: Mike Ivanov
Date:
Subject: Re: random slow query
Next
From: Mike Ivanov
Date:
Subject: Re: random slow query