Thread: Data and logs on different physical drives - advantage?
I learned for other database systems, that it is an advantage to have database on one physical drive and transaction logs on a different physical drive, as this should minimize seeks on the hard-disks.
Is this true for PostgreSQL (V 8.0 on ReadHat)? My tests do not show any speed gained when placing pg_clog and pg_xlog on a different drive.
"Weber, Johann (ISS Kassel)" <jweber@iss.net> writes:
> Is this true for PostgreSQL (V 8.0 on ReadHat)? My tests do not show any
> speed gained when placing pg_clog and pg_xlog on a different drive.
The conventional wisdom is that it's a win to have pg_xlog on a drive by
itself. The above is not that. pg_clog is more in the nature of data,
and in any case you lose the advantage as soon as the drive handling
pg_xlog has to move the head away from the current xlog file.
Whether any particular test case would show an advantage is another
question of course. In a heavy-write scenario I would think you could
probably measure a win.
regards, tom lane
Thanks, after playing around a little I found that placing the indices
on a different drive gave a real performance increase in my scenario.
Johann
-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Monday, October 10, 2005 4:35 PM
To: Weber, Johann (ISS Kassel)
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Data and logs on different physical drives -
advantage?
"Weber, Johann (ISS Kassel)" <jweber@iss.net> writes:
> Is this true for PostgreSQL (V 8.0 on ReadHat)? My tests do not show
> any speed gained when placing pg_clog and pg_xlog on a different
drive.
The conventional wisdom is that it's a win to have pg_xlog on a drive by
itself. The above is not that. pg_clog is more in the nature of data,
and in any case you lose the advantage as soon as the drive handling
pg_xlog has to move the head away from the current xlog file.
Whether any particular test case would show an advantage is another
question of course. In a heavy-write scenario I would think you could
probably measure a win.
regards, tom lane