Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!nntp.coast.net!howland.erols.net!newsfeed.internetmci.com!metro.atlanta.com!news.pcslink.com!ryan
From: ryan@pcslink.com (Ryan Mooney)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: FreeBSD 2.1.6 Striping Performance
Date: Tue, 04 Feb 97 20:41:36 GMT
Organization: PCSLink
Lines: 32
Message-ID: <5d87kv$sin@news.pcslink.com>
Reply-To: ryan@pcslink.com
NNTP-Posting-Host: ryan.pcslink.com
X-Newsreader: News Xpress 2.0 Beta #2
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:35139
I was just playing with striping under 2.1.6 and I got fairly poor performance
results... I was wondering if anyone had any suggestions as to how to improve
it or what I'm doing wrong that could explain it.
Hardware:
Pentium 75 (32M RAM, PCI bus)
Adaptec 3940 Dual SCSI controller
2 Seagate 2G Fast SCSI 2 Barracudas (one per SCSI channel)
(500M for /, /usr, /var, etc... on the first drive)
(128M for swap on each drive - overkill I know, but I'm
pre-setting up for a much beefier machine)
Striping: 1 500M partition on each barracuda, Interleave=16
tests: bonnie, and dd (with a 10M file, bs=512)
I ran the tests on one HD (stand alone) and then on the striped set. The
numbers on the striped set were consistently just a little less than what I
was getting on a single drive. I would suspect that my interleave is wrong
(to low), but I'm not sure... I was wondering if anyone had any perls of
wisdom on how to scrape the most performance out of striped setts... The
end application will be striping on 4G Hawks (one set) and older 1G Connors
(another set), so they will be slower drives... does that affect what I should
use for an interleave?
Is the 3940 a "problem" device under 2.1.6? I know that 2.2 is delayed for
the ahc drivers, am I just seeing an earlier version of whats delaying 2.2?
Will I get better striping performance under 2.2?
OK enough questions :)
Thanks