From owner-freebsd-smp@FreeBSD.ORG Sun Oct 2 23:57:09 2005 Return-Path: X-Original-To: freebsd-smp@freebsd.org Delivered-To: freebsd-smp@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 6EB9516A41F for ; Sun, 2 Oct 2005 23:57:09 +0000 (GMT) (envelope-from ohartman@mail.uni-mainz.de) Received: from mailgate1.zdv.Uni-Mainz.DE (mailgate1.zdv.Uni-Mainz.DE [134.93.178.129]) by mx1.FreeBSD.org (Postfix) with ESMTP id 3B8A543D55 for ; Sun, 2 Oct 2005 23:57:06 +0000 (GMT) (envelope-from ohartman@mail.uni-mainz.de) Received: from [84.245.182.115] (84-245-182-115.ipool.celox.de [84.245.182.115]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mailgate1.zdv.Uni-Mainz.DE (Postfix) with ESMTP id 851EB3000750; Mon, 3 Oct 2005 01:57:03 +0200 (CEST) Message-ID: <434073C8.2000608@mail.uni-mainz.de> Date: Mon, 03 Oct 2005 01:56:56 +0200 From: "O. Hartmann" Organization: Institut =?ISO-8859-1?Q?f=FCr_Geophysik?= User-Agent: Mozilla Thunderbird 1.0.7 (X11/20050930) X-Accept-Language: en-us, en MIME-Version: 1.0 To: mchartzell@getdts.com References: <200510022209.38594.robert@orcafat.com> <4340636D.3090408@getdts.com> In-Reply-To: <4340636D.3090408@getdts.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at uni-mainz.de Cc: freebsd-smp@freebsd.org Subject: Re: FreeBSD 5.4 network performance with SMP X-BeenThere: freebsd-smp@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: FreeBSD SMP implementation group List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 02 Oct 2005 23:57:09 -0000 Matt Hartzell wrote: > In my experience I had to disable the USB in the bios. > > Do a quick vmstat -i and note how many interrupts are on the uhci > device - with no devices even connected to the usb bus. > > Also, what sysctl settings are you using? > > If your network supports it, how about using jumbo frames? > > Robert Lindgren wrote: > >> Hi All, >> >> I have a strange little problem on Dell 1800's. If HT,SMP is enabled >> in the kernel (5.4p7), performance on the integrated em0 is not >> especially good, half the speed of the current 100Mbit connection. >> This is true from the server and a little better to the server. On a >> pci rl0 card in the same machine performance is even worse. About >> 15Mbit from the server! >> >> Then I read a lot of people having poor performance from the rl0 >> cards and a lot of FreeBSD people bashing it for being crap, even in >> the source code for the module it's stated that these cards aren't so >> good. I guess that is right to some extent but not 15Mbit/sec. >> >> So when I rebuilt the kernel without HT and SMP I get full wirespeed >> from the em0 and the rl0 (100Mbit/sec). Is there anything I've been >> doing wrong to get hit by this poor performance with the SMP kernel. >> I read the disabling USB could improve performance but tried that >> with the HT kernel without any luck. >> >> The server has 1 3Ghz cpu, and I would prefer too run a HT enabled >> kernel, but not with that kind of performance, so please advice. The >> same machines (Dell 1800) run linux with HT enabled without any >> network performance degradation. >> >> Best regards >> Robert Lindgren >> >> _______________________________________________ >> freebsd-smp@freebsd.org mailing list >> http://lists.freebsd.org/mailman/listinfo/freebsd-smp >> To unsubscribe, send any mail to "freebsd-smp-unsubscribe@freebsd.org" >> >> >> > > _______________________________________________ > freebsd-smp@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-smp > To unsubscribe, send any mail to "freebsd-smp-unsubscribe@freebsd.org" I see something similar in FreeBSD 6.0-BETA5, no SMP, but with both sk0 and nve0 NICs. This is what my vmstat -i shows up. I do not have attached any USB device, but there seems to be a lot of interrupt requests. nve0 seems to 'hang' sometimes. As I can see, nve0 and the sound device ocm0 (snd_ich) shares an IRQ. Can this be changed anyway? interrupt total rate irq1: atkbd0 15130 3 irq4: sio0 6178 1 irq6: fdc0 27 0 irq12: psm0 42158 10 irq14: ata0 13047 3 irq16: fwohci0 1 0 irq17: skc0 4 0 irq22: ehci0+ 11525 2 irq23: pcm0 nve0 27 0 lapic0: timer 8350298 1999 Total 8438395 2020 Oliver