From: Christiaan Willemsen <cwillemsen@technocon.com>
To: pgsql-performance@postgresql.org
Subject: How to setup disk spindles for best performance
Date: Thu, 21 Aug 2008 00:25:59 +0200
Message-ID: <972BDF1A-B4ED-46EC-B416-0CDF15A0439A@technocon.com> (raw)
I'm currently trying to find out what the best configuration is for
our new database server. It will server a database of about 80 GB and
growing fast. The new machine has plenty of memory (64GB) and 16 SAS
disks, of wich two are already in use as a mirror for the OS.
The rest can be used for PostgreSQL. So that makes a total of 14 15k.5
SAS diks. There is obviously a lot to interesting reading to be found,
most of them stating that the transaction log should be put onto a
separate disk spindle. You can also do this with the indexes. Since
they will be updated a lot, I guess that might be a good idea. But
what no-one states, is what performance these spindle should have in
comparison to the data spindle? If I create a raid 10 of 6 disks for
the data, 4 disk raid 10 for the log, and 4 disk raid 10 for the
indexes, will this yield best performance? Or is it sufficient to just
have a simple mirror for the log and/or indexes...? I have not found
any information about these figures, and I guess it should be possible
to give some pointers on how these different setup might affect
performance?
So I hope someone has already tested this and can give some tips...
Kind regards,
Christiaan
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-performance@postgresql.org
Cc: cwillemsen@technocon.com
Subject: Re: How to setup disk spindles for best performance
In-Reply-To: <972BDF1A-B4ED-46EC-B416-0CDF15A0439A@technocon.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox