From owner-freebsd-hackers Mon Nov 1 8:48:32 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from yana.lemis.com (yana.lemis.com [192.109.197.140]) by hub.freebsd.org (Postfix) with ESMTP id 2BB65152C6 for ; Mon, 1 Nov 1999 08:48:23 -0800 (PST) (envelope-from grog@lemis.com) Received: from mojave.sitaranetworks.com ([199.103.141.157]) by yana.lemis.com (8.8.8/8.8.8) with ESMTP id DAA29763; Tue, 2 Nov 1999 03:18:07 +1030 (CST) (envelope-from grog@lemis.com) Message-ID: <19991101104100.56909@mojave.sitaranetworks.com> Date: Mon, 1 Nov 1999 10:41:00 -0500 From: Greg Lehey To: Tor.Egge@fast.no, wes@softweyr.com Cc: sjr@home.net, freebsd-hackers@FreeBSD.ORG Subject: Re: CCD questions Reply-To: Greg Lehey References: <3819E1EA.83DD04B7@softweyr.com> <199910300414.GAA22428@midten.fast.no> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <199910300414.GAA22428@midten.fast.no>; from Tor.Egge@fast.no on Sat, Oct 30, 1999 at 06:14:32AM +0200 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Saturday, 30 October 1999 at 6:14:32 +0200, Tor.Egge@fast.no wrote: >> Because Vinum is being maintained, and because Vinum will allow you to >> stripe your disks instead of simple concatenate them, which will probably >> result in better I/O rates. > > Some simple measurements shows ccd to be slightly faster when striping > 11 disks, then running 130 threads reading 1MB blocks from random > sector-aligned positions in a 10 GB file. > > with Vinum: 70 MB/s (2 of the 11 subdisks shown in iostat output) > > > tty da5 da6 cpu > tin tout KB/t tps MB/s KB/t tps MB/s us ni sy in id > 0 94 171.82 36 5.96 174.53 38 6.48 0 0 16 3 80 > 0 35 173.90 36 6.13 174.79 36 6.10 0 0 9 3 88 > 1 47 175.33 35 6.01 173.96 38 6.44 0 0 9 3 88 > 1 378 175.69 38 6.53 170.85 38 6.28 0 0 9 3 88 > 1 37 176.47 37 6.42 173.21 35 5.87 0 0 9 3 88 > 0 36 176.05 39 6.72 172.27 35 5.87 0 0 9 3 87 > 0 36 168.95 40 6.60 172.89 36 6.14 0 0 9 4 87 > 0 36 169.44 39 6.42 171.73 38 6.30 0 0 8 4 88 > 0 36 172.64 37 6.24 174.79 38 6.44 0 0 9 3 88 > > With ccd: 80 MB/s > > tty ccd1 cpu > tin tout KB/t tps MB/s us ni sy in id > 0 39 204.46 408 81.56 0 0 5 1 94 > 0 39 205.28 413 82.75 0 0 5 1 94 > 0 39 204.91 438 87.72 0 0 7 1 92 > 0 39 204.96 429 85.96 0 0 6 1 93 > 0 39 205.13 439 87.85 0 0 6 1 93 > 0 39 204.80 448 89.61 0 0 5 1 94 > 0 39 204.77 428 85.66 0 0 5 1 94 > 0 39 204.66 440 87.95 0 0 6 1 93 > 0 39 205.58 421 84.61 0 0 5 1 94 What program did you use to get this information? I'm surprised by the size of the transfers, which is larger than MAXPHYS (128 kB, the largest physical transfer allowed). There's also a discrepancy between the transfer size for Vinum and ccd which almost exactly corresponds to the measured performance. It would be interesting to see the output of rawio (in the Ports Collection) on these two volumes. > ccd is also capable of striping disks with different sizes. Correct. That's one aspect I didn't consider worth doing in Vinum. 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-hackers" in the body of the message