From owner-freebsd-questions Wed May 9 21:58:48 2001 Delivered-To: freebsd-questions@freebsd.org Received: from sandbox.csi.net (sandbox.csi.net [209.44.173.10]) by hub.freebsd.org (Postfix) with ESMTP id 3E05D37B422 for ; Wed, 9 May 2001 21:58:44 -0700 (PDT) (envelope-from smace@intt.org) Received: from metal.intt.org (internet.soes.com [209.44.173.253]) by sandbox.csi.net (INTT-s) with ESMTP id f4A4wd279977; Wed, 9 May 2001 23:58:39 -0500 (CDT) (envelope-from smace@intt.org) Received: from cruncher (cruncher.intt.org [10.112.102.68]) by metal.intt.org (INTT-s) with SMTP id f4A4wX348321; Wed, 9 May 2001 23:58:33 -0500 (CDT) (envelope-from smace@intt.org) From: "Scott Mace" To: "Greg Lehey" Cc: "Dan Phoenix" , "FreeBSD Questions" Subject: RE: vinum on 2 ide drives? Date: Wed, 9 May 2001 23:58:33 -0500 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2911.0) In-Reply-To: <20010510133425.K56501@wantadilla.lemis.com> Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Absolutely, this is not a direct comparison. My results were on an BX chipset with the normal intel ATA33 chip. The MB has an on-board promise ATA100 chip (ASUS CUBX-E) but the system was very flakey, random filesystem corruption with no kernel error messages (yeah). I tried several combinations to see if I could get better performance on the ATA33. These were with IBM DLTA-307030 drives. When I started I was getting about 10MB/sec write 20MB/sec read on a single drive. I enabled tagged queueing on the drive. This sent the performance to ~20MB/sec write 21MB/sec read. After intensive cvsup + cvs co + tars, I got kernel error messages about tag error and timeouts. I turned off tagged queueing and turned on write caching. I got ~20MB/sec write ~20MB/sec read, so I was happy. This was all on single drives (no vinum). I then decided to mirror a filesystem. write performance was almost exactly 50% slower. Without write-cache I got 3MB/sec write (yeah!!). I tried putting each drive on a seperate bus, no change. The current setup is one drive master the other slave on the same bus. I've used vinum for a while on wide scsi and u2wide scsi with none of these problems. As long as the bus isn't congested, a mirror isn't any slower than a single drive and a stripe+concat is typically much faster than a single drive, and you can always just add another bus to stop the congestion problems... Sequential I/O tests are not a real-world, but is does give you an idea of whether or not things are working right. Scott > -----Original Message----- > From: Greg Lehey [mailto:grog@lemis.com] > Sent: Wednesday, May 09, 2001 11:04 PM > To: Scott Mace > Cc: Dan Phoenix; FreeBSD Questions > Subject: Re: vinum on 2 ide drives? > > > > It's difficult to guess what's going on here. Looking at your > configuration below, I'd guess that you're not comparing things > directly. > > > drive vinumdrive0 device /dev/ad0s1e > > drive vinumdrive2 device /dev/ad0s1f > > drive vinumdrive4 device /dev/ad0s1g > > drive vinumdrive1 device /dev/ad1s1e > > drive vinumdrive3 device /dev/ad1s1f > > drive vinumdrive5 device /dev/ad1s1g > > These are four drives too many. You don't want more than one drive > per spindle. > > 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