From owner-freebsd-isp Thu Sep 19 10:30:08 1996 Return-Path: owner-isp Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id KAA08792 for isp-outgoing; Thu, 19 Sep 1996 10:30:08 -0700 (PDT) Received: from ns.frihet.com (root@frihet.bayarea.net [205.219.92.1]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id KAA08732; Thu, 19 Sep 1996 10:29:59 -0700 (PDT) Received: from ns.frihet.com (tweten@localhost [127.0.0.1]) by ns.frihet.com (8.7.5/8.6.12) with ESMTP id KAA06675; Thu, 19 Sep 1996 10:29:20 -0700 (PDT) Message-Id: <199609191729.KAA06675@ns.frihet.com> X-Mailer: exmh version 1.6.7 5/3/96 Reply-To: "David E. Tweten" To: grefen@carpe.net cc: FREEBSD-SCSI-L , FREEBSD-CURRENT-L , FREEBSD-ISP-L Subject: Re: Streamlogic RAID array benchmarks Date: Thu, 19 Sep 1996 10:29:20 -0700 From: "David E. Tweten" Sender: owner-isp@freebsd.org X-Loop: FreeBSD.org Precedence: bulk grefen@carpe.net said: >I assume from your statements that they use the 3 disks as "Data >Data Parity" not as a 'true' RAID 5 "Data Data Data Data Parity" >(this would have other side effects too). If I read this right, it is an interesting, though incorrect, interpretation of the meaning of the various levels of RAID. Garth Gibson's definition, in the original RAID paper written while he was a Berkeley PhD. student, required that RAID 5 have no disk devoted exclusively either to data or to parity. Instead, the responsibility for parity rotated through all disks in the array as a function of sequential block on disk. That way the potential for concentrating parity block accesses on one disk could be avoided in the presence of small writes. Small writes can be handled by reading all the blocks in a parity group that aren't going to be written and combining all data blocks to get a new parity block. They can also be handled by reading (or remembering) the original data and reading the original parity, followed by combining old data, new data, and old parity to get new parity. The statistical effect of it all is to generate more disk traffic for "the parity disk". The "5" had nothing to do with the number of disks. The numbers, 1 through 5 had everything to do with various disk array architectures. If my memory can be trusted (a dubious proposition), it went something like this: RAID 1 mirroring (ie, write the same block on n disks) RAID 2 bitwise SECDED (eg., Thinking Machines Data Vault) RAID 3 multi-disk array with dedicated parity disk RAID 4 ? (something truly awful that I can't remember) RAID 5 multi-disk array with rotating parity assignment So, while designating three disks as "data, data, parity" is certainly not RAID 5, neither is designating five disks as "data, data, data, data, parity". They are both RAID 3 architecture. -- David E. Tweten | 2047-bit PGP Key fingerprint: | tweten@frihet.com 12141 Atrium Drive | E9 59 E7 5C 6B 88 B8 90 | tweten@and.com Saratoga, CA 95070-3162 | 65 30 2A A4 A0 BC 49 AE | (408) 446-4131 Those who make good products sell products; those who don't, sell solutions.