Date: Wed, 3 Feb 1999 18:22:58 +1100 (EST) From: "John Saunders" <john.saunders@nlc.net.au> To: Greg Lehey <grog@lemis.com> Cc: freebsd-questions@FreeBSD.ORG Subject: Re: advice on setting up a Vinum volume Message-ID: <Pine.LNX.3.95.990203181240.27450B-100000@nhj.nlc.net.au> In-Reply-To: <19990203171610.Z1179@freebie.lemis.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 3 Feb 1999, Greg Lehey wrote:
> > Sustained performance is much better. On my pair of Quantum 6.4GB drives
> > they get approx 7MB/s sustained (dd if=/dev/zero of=file bs=262144
> > count=1024). With a vinum strip of 256K in size I get sustained
> > 12.5MB/s. Running squid on this setup I can watch the load being
> > distributed by running (systat -iostat 1)
>
> Interesting. Which version of FreeBSD is this? Are you using the
> standard vinum, which includes slow debugging aids, or have you
> compiled them out?
This is done with RELENG_3 cvsupped about a week ago. I did a recent cvsup
and noticed a lot of vinum changes, so it was prior to the lastest
round of changes. Particularly the vinum_slices -> vinum_disks change
bit me (I use rc.conf.local for host config).
The machine is a PII-350MHz/100MHz bus with 128MB SDRAM so I think any CPU
intensive debug would be dealt with rather quickly.
On a slower P120MHz I get 4.8MB/s on a single drive and 7.2MB/s with a
vinum stripe. So CPU does play a factor in getting close to the
theoretical double in performance.
> > although vinum stats show that one drive seems to be used about 20%
> > more than the other.
>
> I've noticed this too. There's no way this can be a vinum problem (in
> other words, you should see exactly the same distribution when using
> ccd). I suspect that it's an incompatibility between ufs and the
> stripe size. I suspect that you may end up with all super blocks on
> one drive. It would be interesting to calculate what stripe size
> would balance the cylinder groups across the drives.
Is there anything wrong with a strip size that is not a power of 2?
Although it still would have to be a multiple of the block size. On Linux
the mke2fs command has an option to specify the stripe size and it does
some tweaks to avoid putting super blocks on the same drive. With the ufs
32MB cylinder groups I woudln't be surprised if if all super blocks were
on the same disk. Anyone for prime number stripe sizes?
Cheers.
-- +------------------------------------------------------------+
. | John Saunders - mailto:john@nlc.net.au (EMail) |
,--_|\ | - http://www.nlc.net.au/ (WWW) |
/ Oz \ | - 02-9489-4932 or 041-822-3814 (Phone) |
\_,--\_/ | NHJ NORTHLINK COMMUNICATIONS - Supplying a professional, |
v | and above all friendly, internet connection service. |
+------------------------------------------------------------+
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?Pine.LNX.3.95.990203181240.27450B-100000>