Re: Six PostgreSQL questions from a pokerplayer

From: Craig Ringer <craig(at)postnewspapers(dot)com(dot)au>
To: Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
Cc: Stephen Frost <sfrost(at)snowman(dot)net>, Patvs <patvs(at)chello(dot)nl>, pgsql-performance(at)postgresql(dot)org
Subject: Re: Six PostgreSQL questions from a pokerplayer
Date: 2009-07-07 04:51:14
Message-ID: 1246942274.26636.318.camel@tillium.localnet
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Mon, 2009-07-06 at 15:27 -0400, Mark Mielke wrote:

> Even if you only have 4 GB of RAM, the 32-bit kernel needs to fight
> with "low memory" vs "high memory", whereas 64-bit has a clean address
> space.

That's a good point. The cutoff is probably closer to 2G or at most 3G.
Certainly it's madness to use hacks like PAE to gain access to the RAM
behind the PCI address space rather than just going 64-bit ... unless
you have a really pressing reason, at least.

It's also nice that on a 64 bit machine, there's no 2G/2G or 3G/1G
userspace/kernelspace address mapping split to limit your app's memory
use. I seem to recall that Windows uses 2G/2G which can be painfully
limiting for memory-hungry applications.

Personally, I'd probably go 64-bit on any reasonably modern machine that
could be expected to have more than 2 or 3 GB of RAM. Then again, I
can't imagine willingly building a production database server for any
non-trivial (ie > a couple of gigs) database with less than 8GB of RAM
with RAM prices so absurdly low. Skip-lunch-to-afford-more-RAM low.

--
Craig Ringer

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Saurabh Dave 2009-07-07 05:11:14 Re: Bundling postgreSQL with my Java application
Previous Message justin 2009-07-06 23:16:21 Re: Bundling postgreSQL with my Java application