From owner-freebsd-scsi Sun Mar 14 20:32:21 1999 Delivered-To: freebsd-scsi@freebsd.org Received: from duey.wolves.k12.mo.us (duey.wolves.k12.mo.us [207.160.214.9]) by hub.freebsd.org (Postfix) with ESMTP id 651CE1506C for ; Sun, 14 Mar 1999 20:32:18 -0800 (PST) (envelope-from cdillon@wolves.k12.mo.us) Received: from duey.wolves.k12.mo.us (cdillon@duey.wolves.k12.mo.us [207.160.214.9]) by duey.wolves.k12.mo.us (8.8.8/8.8.8) with ESMTP id WAA04819; Sun, 14 Mar 1999 22:31:55 -0600 (CST) (envelope-from cdillon@wolves.k12.mo.us) Date: Sun, 14 Mar 1999 22:31:55 -0600 (CST) From: Chris Dillon To: Joerg Wunsch Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: cvs commit: src/sys/cam cam_xpt.c In-Reply-To: <19990314201946.00024@uriah.heep.sax.de> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-scsi@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org On Sun, 14 Mar 1999, J Wunsch wrote: > As Jean-Marc Zucconi wrote: > > > Yes this is very poor for a wide drive. On my narrow DCAS-34330: > > Writing the 250 Megabyte file, 'iozone.tmp'...80.351562 seconds > > Reading the file...38.601562 seconds > > > > IOZONE performance measurements: > > 3262463 bytes/second for writing the file > > 6791020 bytes/second for reading the file > > I get about 4.4 MB/s writing and 5.8 MB/s reading, with tagged queuing > still enabled. I also get better bonnie figures than those mentioned > in PR kern/10398, about 4 MB/s `Per Char' and 5 MB/s `Block'. > > However, as Ken already mentioned, see the audit-trail from PR 10398 > for a counterexample. I just did some tests with tagging totally disabled, enabled, and with only 2 tags. This is long. Sorry in advance for the messy wrapping. Hopefully everybody can parse it... da1 at ncr0 bus 0 target 1 lun 0 da1: Fixed Direct Access SCSI-2 device da1: 40.000MB/s transfers (20.000MHz, offset 15, 16bit) da1: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) da0 at ncr0 bus 0 target 0 lun 0 da0: Fixed Direct Access SCSI-2 device da0: 40.000MB/s transfers (20.000MHz, offset 15, 16bit) da0: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) Without tagged queueing - single drive: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4742 96.0 7154 47.3 2540 20.1 5739 95.7 7422 30.2 91.7 4.4 IOZONE performance measurements: 6816542 bytes/second for writing the file 7593647 bytes/second for reading the file Without tagged queueing - 64 block striped ccd partition: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4499 95.5 9556 86.6 3383 30.5 5537 93.8 12330 65.7 133.9 6.6 IOZONE performance measurements: 6629672 bytes/second for writing the file 11590477 bytes/second for reading the file da0 at ncr0 bus 0 target 0 lun 0 da0: Fixed Direct Access SCSI-2 device da0: 40.000MB/s transfers (20.000MHz, offset 15, 16bit), Tagged Queueing Enabled da0: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) da1 at ncr0 bus 0 target 1 lun 0 da1: Fixed Direct Access SCSI-2 device da1: 40.000MB/s transfers (20.000MHz, offset 15, 16bit), Tagged Queueing Enabled da1: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) With tagged queueing - single drive: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4883 97.0 4103 25.8 2263 23.6 5789 95.8 7420 32.4 99.6 4.8 IOZONE performance measurements: 6839119 bytes/second for writing the file 7425600 bytes/second for reading the file With tagged queueing - 64 block striped ccd partition: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4570 96.2 6725 55.1 2534 23.0 5575 94.0 11776 65.2 148.3 7.5 IOZONE performance measurements: 6687480 bytes/second for writing the file 11481413 bytes/second for reading the file da0 at ncr0 bus 0 target 0 lun 0 da0: Fixed Direct Access SCSI-2 device da0: 40.000MB/s transfers (20.000MHz, offset 15, 16bit), Tagged Queueing Enabled da0: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) da1 at ncr0 bus 0 target 1 lun 0 da1: Fixed Direct Access SCSI-2 device da1: 40.000MB/s transfers (20.000MHz, offset 15, 16bit), Tagged Queueing Enabled da1: 4134MB (8467200 512 byte sectors: 255H 63S/T 527C) With tagged queueing (min/max 2 tags) - single drive: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4882 97.1 6961 46.4 2897 24.1 5761 95.6 7427 32.7 100.9 4.6 IOZONE performance measurements: 6759895 bytes/second for writing the file 7600097 bytes/second for reading the file With tagged queueing (min/max 2 tags) - 64 block striped ccd partition: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 200 4562 96.5 9323 86.9 3275 30.4 5663 95.1 12382 68.4 143.2 7.1 IOZONE performance measurements: 6603578 bytes/second for writing the file 11226911 bytes/second for reading the file By these measurements, completely turning off tagging helped certain write scenarios and hurt others. You'll notice, however, 2 tags gained almost as much performance increase as disabling them totally, and didn't hurt the other scenarios as much. 2 tags may not be be the optimal selection... its just a starting point that I used, after noticing some Seagate drives had similar quirks. -- Chris Dillon - cdillon@wolves.k12.mo.us - cdillon@inter-linc.net /* FreeBSD: The fastest and most stable server OS on the planet. For Intel x86 and compatibles (SPARC and Alpha under development) ( http://www.freebsd.org ) */ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message