From owner-freebsd-stable Tue Jan 2 22:51:43 2001 From owner-freebsd-stable@FreeBSD.ORG Tue Jan 2 22:51:39 2001 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from wantadilla.lemis.com (wantadilla.lemis.com [192.109.197.80]) by hub.freebsd.org (Postfix) with ESMTP id 978BB37B400 for ; Tue, 2 Jan 2001 22:51:36 -0800 (PST) Received: by wantadilla.lemis.com (Postfix, from userid 1004) id 021F36A911; Wed, 3 Jan 2001 17:21:32 +1030 (CST) Date: Wed, 3 Jan 2001 17:21:32 +1030 From: Greg Lehey To: Zero Sum Cc: Jim King , Alfred Perlstein , Thomas Seck , freebsd-stable@FreeBSD.ORG Subject: RAID costs (was: Vinum safe to use for raid 0?) Message-ID: <20010103172132.T4336@wantadilla.lemis.com> References: <20010102230107.A559@basildon.homerun> <01010313305000.03936@shalimar.net.au> <00e301c0752e$7ba65e60$04e48486@marble> <01010315274900.04373@shalimar.net.au> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <01010315274900.04373@shalimar.net.au>; from count@shalimar.net.au on Wed, Jan 03, 2001 at 03:27:49PM +1100 Organization: LEMIS, PO Box 460, Echunga SA 5153, Australia Phone: +61-8-8388-8286 Fax: +61-8-8388-8725 Mobile: +61-418-838-708 WWW-Home-Page: http://www.lemis.com/~grog X-PGP-Fingerprint: 6B 7B C3 8C 61 CD 54 AF 13 24 52 F8 6D A4 95 EF Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Wednesday, 3 January 2001 at 15:27:49 +1100, Zero Sum wrote: > On Wed, 3 Jan 2001, I (Zero Sum) wrote: > On Wednesday 03 January 2001 13:49, Tom wrote in reponse: > >> But you don't use RAID 5 for performance. RAID is about reliablity. If >> you want reliability and performance, RAID10 wins. If you are operating >> on a budget use RAID5, or if you are putting toghether a lot of storage, >> RAID50. > > Jim and Tom mention the factor of expense. In Australia, I think we pay > about five times the price that Americans seem to (this is not a > price/currency conversion) Either I'm misunderstanding you, or you're paying too much. > but I still can't take that factor seriously. Disk space is > *cheap*. As Tom says, for reliability and performance RAID1/0 wins. > Using RAID 5 is "spoiling the ship for a pennyworth of tar", or so > it seems to me. For a certain definition of "cheap", sure. Let's assume you want a 10 TB web server. Never mind the cost of enclosures, just using relatively cheap disks like the IBM DTLA-307075, 75 GB a throw and available for about AUD 800, you're talking 130 drives, or a little over AUD 100K. And that with no redundancy. Add RAID-5, and you may pay $20K more. Use RAID-1 instead and your costs have gone up by $100K instead of $20K. What do you have to show for it? Increased power consumption, increased likelihood of a drive failure. But hardly any difference in performance. > Hardware RAID is expensive. If you make that investment, the price > of more disk should not be important. By comparison, hardware RAID is cheap. Software RAID is cheaper, of course, and what we've seen indicates that the performance is significantly better. > Software RAID gets slower the more complex it is. RAID 5 is > considerably more complex than 0/1. Wouldn't you be better off > spending money on a less capable processor and more disk? I haven't > done any measurements here. I am only going from (limited) > experience. You're also being rather vague. There's a more detailed explanation at http://www.vinumvm.org/vinum/intro.html. > But I can't see 'expense' as a justifying factor here. Not in a > professional arena. Precisely there. $80K for no particular purpose is not commercially justifiable. > Greg mentioned a situation (web server) where there are many reads > (99%). I think that for RAID 5 to help matters much, you would have > to have a very high hit rate on an immense number of different > static pages. The read performance of 5 vs 1/0 is there, but it is > marginal. Whether it gives you any improvement or not (how much, if > any) is very dependant on what is going on in bufferspace. No, in real life you can assume that you're not going to have very much cached. And RAID-5 is no faster than RAID-1 on reads, but it's also not measurably slower. 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-stable" in the body of the message