From owner-freebsd-hackers Sat Jul 13 14:00:37 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id OAA12018 for hackers-outgoing; Sat, 13 Jul 1996 14:00:37 -0700 (PDT) Received: from MindBender.HeadCandy.com (root@mindbender.headcandy.com [199.238.225.168]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id OAA12013 for ; Sat, 13 Jul 1996 14:00:32 -0700 (PDT) Received: from localhost.HeadCandy.com (michaelv@localhost.HeadCandy.com [127.0.0.1]) by MindBender.HeadCandy.com (8.7.5/8.7.3) with SMTP id NAA06514; Sat, 13 Jul 1996 13:58:46 -0700 (PDT) Message-Id: <199607132058.NAA06514@MindBender.HeadCandy.com> X-Authentication-Warning: MindBender.HeadCandy.com: Host michaelv@localhost.HeadCandy.com [127.0.0.1] didn't use HELO protocol To: Joe Greco cc: pst@shockwave.com (Paul Traina), hackers@freebsd.org Subject: Re: using ccd for striping? In-reply-to: Your message of Sat, 13 Jul 96 14:27:54 -0500. <199607131927.OAA00314@brasil.moneng.mei.com> Date: Sat, 13 Jul 1996 13:58:43 -0700 From: "Michael L. VanLoon -- HeadCandy.com" Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk >> Is anyone using the ccd driver in striping mode? I'd like to hear about >> other people's good/bad experiences before trying it out myself. > >I've seen it act a little funny with way large stripe sizes (65536), several >different times I have seen it develop "non-writable" and "non-accessible" >zones near the very end of the disk. I presume you're talking about the interleave factor? Since you say large, and 65536, I assume you're saying 64MB stripes? (64K disk blocks, 32K bytes, wouldn't be "large".) Is there a good reason for doing that? I would think you'd get a much better performance boost by going with interleaved stripes somewhere between the size of a filesystem cluster to a unit the size of the smallest drive's cache. (When I say "cluster", I am referring to the size of a filesystem block -- 8 fragments -- not a disk block -- 512 bytes.) I have been going with a 128 block (64K) interleave, since that's half the size of my drives' caches, and is the size of four filesystem clusters (whichh I have allocated as 16K). Any bigger than that and you lose all the benefits of interleaved access. In fact, I have considered reducing it to 16K so it would interleave on single-cluster sizes. On the other hand, if you get too fine of an interleave, you can get inefficient since the ccd driver will be doing a lot of work reading in lots of small pieces and assembling them. Still, I can't think of any reason you'd ever want 64MB interleaved stripes. I don't see any benefit in such an arrangement. ----------------------------------------------------------------------------- Michael L. VanLoon michaelv@HeadCandy.com --< Free your mind and your machine -- NetBSD free un*x >-- NetBSD working ports: 386+PC, Mac 68k, Amiga, Atari 68k, HP300, Sun3, Sun4/4c/4m, DEC MIPS, DEC Alpha, PC532, VAX, MVME68k, arm32... NetBSD ports in progress: PICA, others... Roll your own Internet access -- Seattle People's Internet cooperative. If you're in the Seattle area, ask me how. -----------------------------------------------------------------------------