From owner-freebsd-questions Fri Aug 27 13:22: 0 1999 Delivered-To: freebsd-questions@freebsd.org Received: from mail.SKINNYHIPPO.COM (mail.SKINNYHIPPO.COM [216.25.13.94]) by hub.freebsd.org (Postfix) with ESMTP id D76F9155A8 for ; Fri, 27 Aug 1999 13:21:48 -0700 (PDT) (envelope-from panda@skinnyhippo.com) Received: from egg [202.188.111.43] by mail.SKINNYHIPPO.COM (SMTPD32-5.05) id A2EE3C300C8; Fri, 27 Aug 1999 16:19:58 -0400 Message-Id: <3.0.5.32.19990828045614.00a07d80@mail.skinnyhippo.com> X-Sender: panda@mail.skinnyhippo.com X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.5 (32) Date: Sat, 28 Aug 1999 04:56:14 +0900 To: freebsd-questions@freebsd.org From: chas Subject: how can you tell when disk i/o is limiting performance ? Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Is there any method to measure disk I/O, in particular to work out if it is limiting performance ? Context : 1 x 18 GB SCSI HD (10,000 RPM) Each day : 200,000 x CGI processes, each reading 2 or 3 flat files 100,000 x CGI processes, each reading/writing to mysql database 1,000,000+ x flat files (images, .html etc) plus 400 MB+ of apache log files are created each day. There seems to be ample RAM and CPU (and we're not even going into swap) but the website is really crawling. I realise that this could be due to the poor bandwidth in China (where the server is hosted) but would like to also monitor the disk i/o if it's possible since the CGIs open a lot of files and the apache log files are being written nonstop. Hopefully then I can plan/budget for more disks and/or a second server as traffic reaches a threshold. So : a) can this be measured ? (i.e how do you know when your disk i/o is the limiting factor ? ) b) would it make more sense to have 3 separate (physical) disks for : - apache log files - mysql database - operating system, applications, website files/cgi The reasoning behind having mysql database on a dedicated disk is that i remember working with Cold Fusion and SQL server on NT - it was better to have 2 small machines (cold fusion and website sat on one, sql server sat on another) than one huge beast of a machine with both application and server and database sat on it. So I guess that still hold true for hard disks on the same box, though I'm not sure 1 + 1 = 2 in terms of performance. Thanking you in advance, chas To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message