From owner-freebsd-performance@FreeBSD.ORG Fri Nov 19 03:16:56 2004 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 4D71416A4CE for ; Fri, 19 Nov 2004 03:16:56 +0000 (GMT) Received: from note.orchestra.cse.unsw.EDU.AU (note.orchestra.cse.unsw.EDU.AU [129.94.242.24]) by mx1.FreeBSD.org (Postfix) with ESMTP id B501743D45 for ; Fri, 19 Nov 2004 03:16:55 +0000 (GMT) (envelope-from lukem@cse.unsw.edu.au) Received: From wagner With LocalMail ; Fri, 19 Nov 2004 14:16:51 +1100 From: Luke To: Aaron Glenn Date: Fri, 19 Nov 2004 14:16:51 +1100 (EST) In-Reply-To: <18f6019404111817533b93cbba@mail.gmail.com> Message-ID: References: <20041118101026.55888.qmail@web14121.mail.yahoo.com> <20041118105543.10295.qmail@web41208.mail.yahoo.com> <18f6019404111809224fb97c06@mail.gmail.com> <18f6019404111817533b93cbba@mail.gmail.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Mailman-Approved-At: Fri, 19 Nov 2004 13:18:23 +0000 cc: freebsd-performance@freebsd.org cc: "lukem.freebsd@cse.unsw.edu.au" Subject: Re: I've ran out of ideas X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 19 Nov 2004 03:16:56 -0000 On Thu, 18 Nov 2004, Aaron Glenn wrote: >> I would theorise that your web server is probably only issuing fairly >> small disk reads, so your network performance is being bound by the disk. > > That was my first hunch; hence creating a separate partition with > obscene blocksizes (-: Increasing the block size of the disk need not imply that the application is issuing larger reads. >> To test this theory, you could: >> >> * create an mfs partition and serve off it to see what kind of performance >> you can get >> * see if you can tweak the web server to use larger reads and writes. >> >> In your original message you said that increasing the disk block size >> improved performance, which would also indicate that you are being limited >> by disk performance. > > I doubt increasing it even more would help. Law of diminishing returns > and all that; then again I'm no expert. Thoughts? I have not suggested that you increase the blocksize further. Please read the above points again. As Sean pointed out, if your web server supports sendfile then that would be a good option. Your recent post which showed poor performance from /dev/zero does bring the theory that your disk is the bottleneck into question, however you may still find it is worth trying the above suggestions. -- Luke