Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 5 Nov 2000 17:43:55 +1100
From:      Greg Lehey <grog@lemis.com>
To:        Gabor Kincses <gabor@acm.org>
Cc:        randy@psg.com, FreeBSD Questions <questions@FreeBSD.ORG>
Subject:   Re: CCD
Message-ID:  <20001105174355.V307@sydney.worldwide.lemis.com>
In-Reply-To: <3A038A39.D646F2F2@acm.org>; from gabor@acm.org on Fri, Nov 03, 2000 at 10:02:01PM -0600
References:  <3A038A39.D646F2F2@acm.org>

next in thread | previous in thread | raw e-mail | index | archive | help
[moved to -questions; this has nothing to do with SCSI]

On Friday,  3 November 2000 at 22:02:01 -0600, Gabor Kincses wrote:
> Hello,
>
> This is all I had to do:

Is this a reply to something?

> KERNEL config file:
> pseudo-device   ccd 4   # Concatenated disk driver
>
> gabor@steinbeck: <1160> cat /etc/ccd.conf
> ccd0            512     0 /dev/ad0s3f /dev/ad2s1f
>
> I have been using this since June.   Early on I measured performance and
> I got ~75% increased random read and sequential block read/write per
> bonnie.  bonnie suffers from file system caching but I ran it with 2x
> the size of my ram, which should eliminate caching artifacts.

No, it just reduces them.  Use rawio to eliminate them.

> BTW, the two disks are quite dissimilar (13 and 4 GB), I just made
> sure to create identical sized `f' partitions.  Since performance
> speaks for itself, I consider the disk geometry a non-issue.

If you don't mind losing 9 GB of storage space.  I'd call that an
issue.  You'd probably be better off concatenating.

> Granted I had to increase the interleave factor to much bigger than
> what the man page recommended to get optimum performance.

Yes, the man page is wrong.  What stripe size did you use?  You don't
want a power of 2 as an interleave factor; that will leave all
superblocks on one drive, at least for 2 disks.

> You may also want to strip swap like this (/etc/fstab):
>
> /dev/ad0s2b     none        swap    sw      0   0
> /dev/ad2s1b     none        swap    sw      0   0

I'm not sure what you're saying here.

> I may have had to fiddle with disklabel when installing, but it
> wasn't a big deal, I think I just followed ccdconfig(8).
>
> BTW, does anybody have numbers on IDE disk speeds to help decide if a
> third disk would max-out a 33MHz UDMA IDE controller?

You can't put a third disk on an IDE controller.  But even the second
one will give you serious performance hits, since the transfers are
serialized.

It's been a while since I've compared ccd and Vinum performance.
Vinum used to be much better, but some problems in ccd have been
fixed.  You get many other advantages from Vinum.

Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20001105174355.V307>