From: Chris Browne <cbbrowne@acm.org>
Subject: Re: Mount database on RAM disk?
Date: Thu, 07 Jul 2005 23:30:09 -0400
Organization: cbbrowne Computing Inc
Lines: 30
Message-ID: <60zmsykm9q.fsf@dba2.int.libertyrms.com>
References: <42CDD7B8.6050109@stuartbishop.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux)
Cancel-Lock: sha1:WI0k/qpXJjfTMWU2RHxH3a6FG8A=
To: pgsql-performance@postgresql.org

stuart@stuartbishop.net (Stuart Bishop) writes:
> I'm putting together a road map on how our systems can scale as our
> load increases. As part of this, I need to look into setting up some
> fast read only mirrors of our database. We should have more than
> enough RAM to fit everything into memory. I would like to find out
> if I could expect better performance by mounting the database from a
> RAM disk, or if I would be better off keeping that RAM free and
> increasing the effective_cache_size appropriately.

If you were willing to take on a not-inconsiderable risk, I'd think
that storing WAL files on a RAMDISK would be likely to be the fastest
improvement imaginable.

If I could get and deploy some SSD (Solid State Disk) devices that
would make this sort of thing *actually safe,* I'd expect that to be a
pretty fabulous improvement, at least for write-heavy database
activity.

> I'd also be interested in knowing if this is dependant on whether I
> am running 7.4, 8.0 or 8.1.

Behaviour of all three could be somewhat different, as management of
the shared cache has been in flux...
-- 
(format nil "~S@~S" "cbbrowne" "acm.org")
http://www.ntlug.org/~cbbrowne/sap.html
Rules of the Evil Overlord #78.  "I will not tell my Legions of Terror
"And he must  be taken alive!" The command will be:  ``And try to take
him alive if it is reasonably practical.''"
<http://www.eviloverlord.com/>