From owner-freebsd-performance@FreeBSD.ORG Sat Jan 13 20:03:20 2007 Return-Path: X-Original-To: freebsd-performance@freebsd.org Delivered-To: freebsd-performance@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 35B1116A492 for ; Sat, 13 Jan 2007 20:03:20 +0000 (UTC) (envelope-from michel@lucenet.com.br) Received: from msrv.matik.com.br (msrv.matik.com.br [200.152.83.14]) by mx1.freebsd.org (Postfix) with ESMTP id A057113C458 for ; Sat, 13 Jan 2007 20:03:19 +0000 (UTC) (envelope-from michel@lucenet.com.br) Received: from webmail.matik.com.br (localhost [127.0.0.1]) by msrv.matik.com.br (8.13.8/8.13.1) with ESMTP id l0DK3HEF052964; Sat, 13 Jan 2007 18:03:18 -0200 (BRST) (envelope-from michel@lucenet.com.br) Received: from 200.152.83.36 (SquirrelMail authenticated user luc.michel) by webmail.matik.com.br with HTTP; Sat, 13 Jan 2007 18:03:18 -0200 (BRST) Message-ID: <53057.200.152.83.36.1168718598.squirrel@webmail.matik.com.br> In-Reply-To: <20070113194154.GA65864@xor.obsecurity.org> References: <64656.200.152.83.36.1168651673.squirrel@webmail.matik.com.br> <45A87878.1050505@paradise.net.nz> <63758.200.152.83.36.1168689227.squirrel@webmail.matik.com.br> <20070113164232.GA34348@xor.obsecurity.org> <64857.200.152.83.36.1168710081.squirrel@webmail.matik.com.br> <20070113180036.GA64359@xor.obsecurity.org> <60639.200.152.83.36.1168714686.squirrel@webmail.matik.com.br> <20070113190447.GA65571@xor.obsecurity.org> <64716.200.152.83.36.1168716219.squirrel@webmail.matik.com.br> <20070113194154.GA65864@xor.obsecurity.org> Date: Sat, 13 Jan 2007 18:03:18 -0200 (BRST) From: "Michel Santos" To: "Kris Kennaway" User-Agent: SquirrelMail/1.4.9a MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) Importance: Normal Cc: freebsd-performance@freebsd.org, Mark Kirkwood Subject: Re: diskio low read performance X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 13 Jan 2007 20:03:20 -0000 Kris Kennaway disse na ultima mensagem: >> >> But also I checked the ULE/BSD against my particular problem and there >> is >> no difference at all. I get no acceptable disk read performance when >> comparing what I had with 4.11, wether with ULE or with 4BSD > > Be very careful, because I've personally measured severe disk I/O > penalties with ULE on SMP hardware. In fact in my testing ULE gives > worse SMP performance under load across the board compared to 4BSD > (it's only faster for the lightest of workloads). > seems suspicious and I will compile this night and recheck to be sure > If you're absolutely certain that ULE is not to blame (and want to > continue to take the risk of other performance and stability problems > down the line), that basically leaves something to do with the scsi > driver and/or its interaction with your hardware as the probable > cause. I don't know enough about this particular hardware to comment > further though. I will check 4BSD again, but I do not have stability problems. The system works really constant and without any odds but the diskio thing certainly the reason why I have a LSI card is that I was blaming the adaptec driver first, but even with the adaptec disabled and the LSI card on the Supermicro nothing changed. So I do not know for sure but unlikely both drivers have the same problem I guess. > > * adaptive mutexes are usually a win so it's a bit unusual that you > have disabled them, but I assume you have tested this. > I get better network performance when not using polling with it and seems not to harm polling in my case so I let it in > * Dunno about the AUTO_EOI_1 option, I don't think you even have this > hardware on your system (device atpic didn't probe in your dmesg). right, it is a old setting I ever forget taking it out > > * HZ=1000 is superfluous since it is the default but may or may not > help. In some workloads the increased overhead relative to the old > default of HZ=100 gives a performance loss. Maybe it helps with > polling though, I dunno. that I did not know but does no bad either. With lower HZ it does not work for me and I have the impression it works better with 2000. I have 5 NICs and that is the reason I guess. But it is not my priority at this moment and I believe it does not mess with diskio. Michel computador é como nem cavalo e mulher mais que montam neles, pior que ficam ... **************************************************** Datacenter Matik http://datacenter.matik.com.br E-Mail e Data Hosting Service para Profissionais. ****************************************************