From owner-freebsd-hackers Thu Nov 21 17:08:44 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id RAA02265 for hackers-outgoing; Thu, 21 Nov 1996 17:08:44 -0800 (PST) Received: from asstdc.scgt.oz.au (root@asstdc.scgt.oz.au [202.14.234.65]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id RAA02260 for ; Thu, 21 Nov 1996 17:08:34 -0800 (PST) Received: (from imb@localhost) by asstdc.scgt.oz.au (8.7.6/BSD4.4) id MAA26336 Fri, 22 Nov 1996 12:08:09 +1100 (EST) From: michael butler Message-Id: <199611220108.MAA26336@asstdc.scgt.oz.au> Subject: Re: Disk Striping In-Reply-To: from Jaye Mathisen at "Nov 21, 96 12:15:12 pm" To: mrcpu@cdsnet.net (Jaye Mathisen) Date: Fri, 22 Nov 1996 12:08:09 +1100 (EST) Cc: dwoods@netgazer.com, hackers@freebsd.org X-Mailer: ELM [version 2.4ME+ PL28 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Jaye Mathisen writes: > Uh, by definition, isn't striping spreading the load across spindles? > I realize it's a tad more complicated than that, but the essence is the > same. This is optimally effective when there is only one locality of reference since striping then serves to reduce the number of seeks (divide number of tracks transferred by N drives) and introduce an element of parallelism into the disk transactions (each drive will detach from the SCSI bus whilst busy). > > There are four activities which consume disk resources: > > > > i) maintenance of the active and history files > > ii) maintenance of the overview hierarchy > > iii) writing out the articles themselves > > iv) scribbling to /var/log/news These activities are usually not operating on adjacent locales when implemented on a single drive (or one logical constructed out of a few physical ones). Thus the striped drive will be seeking all over the place and losing much of the advantage of striping in the first place. You do not want to design something that inherently seeks a lot, you need to minimise seeking for both performance, long term reliability and drive life. It is, however, quite valid to stripe any one activity, such as scribbling out the articles across multiple drives. You could, for example, use two striped drives for the history stuff, one for the overview (only if you have any readers) and two striped for the article spool to achieve what you're after. Even better, split the two striped arrangements onto two separate (SCSI) controllers, michael