Subject: Re: RaidFrame poor performance
To: Mihai CHELARU <email@example.com>
From: Daniel Carosone <firstname.lastname@example.org>
Date: 01/19/2005 20:36:52
Content-Type: text/plain; charset=us-ascii
On Wed, Jan 19, 2005 at 10:55:27AM +0200, Mihai CHELARU wrote:
> START disks
> START layout
> 32 1 1 5
A 4-unit raid-5 is almost always going to suck for filesystem
performance. There is virtually no layout that will easily map a
useful filesystem blocksize onto the underlying 3-unit stripe (plus
Therefore, for almost every write you do, you wind up doing a
read-modify-write cycle; that's what kills your performance.
If you have the chance, for the sake of demonstration, compare
performance with 3-way and 4-way raid-0.
To resolve this, if you can, add a fifth disk and choose your stripe
size and filesystem block size carefully so the filesystem blocks
exactly map to a full stripe write. Otherwise, you'd be much better
off with stripe+mirror over the four disks.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (NetBSD)
-----END PGP SIGNATURE-----