[Top] [All Lists]

Re: hardware RAID sunit and swidth values?

To: Robert Sander <gurubert-dated-1050097897.ckffphdm@xxxxxxxxxxxxxxx>
Subject: Re: hardware RAID sunit and swidth values?
From: "Martin K. Petersen" <mkp@xxxxxxx>
Date: 11 Apr 2003 18:12:46 -0400
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <news2mail-slrnb9eed1.qm2.gurubert-dated-1050097897.ckffphdm@raman.epigenomics.epi>
Organization: mkp.net
References: <news2mail-slrnb9cnsc.156.gurubert-dated-1050042067.onpbonhn@raman.epigenomi cs.epi> <1050072633.16906.1.camel@two.nks.net> <1050073584.23735.10.camel@stout.americas.sgi.com> <1050075759.16906.48.camel@two.nks.net> <20030411162601.GA14523@pc9391.physik.uni-regensburg.de> <1050081295.16906.70.camel@two.nks.net> <news2mail-slrnb9eed1.qm2.gurubert-dated-1050097897.ckffphdm@raman.epigenomics.epi>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
>>>>> "Robert" == Robert Sander <ml-linux-xfs@xxxxxxxxxxxxxxx> writes:

Robert> That may be the case with software RAID, where mkfs.xfs is
Robert> able to get the stripe size from the md driver. But with at
Robert> least our hardware RAID boxen sunit and swidth are both set to
Robert> 0 (zero).

Robert> My only remaining question is: Does this affect performance in
Robert> any dramatic way?

In general, no.

But it depends on the internal workings of your hardware RAID
controller.  Only way to find out is to experiment with different
values and your anticipated I/O load.

Martin K. Petersen      http://mkp.net/

<Prev in Thread] Current Thread [Next in Thread>