From owner-freebsd-net Tue May 21 4:17:15 2002 Delivered-To: freebsd-net@freebsd.org Received: from scribble.fsn.hu (scribble.fsn.hu [193.224.40.95]) by hub.freebsd.org (Postfix) with SMTP id 4A4B337B40F for ; Tue, 21 May 2002 04:17:11 -0700 (PDT) Received: (qmail 29490 invoked by uid 1000); 21 May 2002 11:18:52 -0000 Received: from localhost (sendmail-bs@127.0.0.1) by localhost with SMTP; 21 May 2002 11:18:52 -0000 Date: Tue, 21 May 2002 13:18:52 +0200 (CEST) From: Attila Nagy To: Terry Lambert Cc: freebsd-net@freebsd.org Subject: Re: HEADS UP: ALTQ integration developer preview In-Reply-To: <3CE6D976.3264DE53@mindspring.com> Message-ID: References: <3CE55A9B.73EA3DE4@mindspring.com> <3CE61675.BCE2A9E1@mindspring.com> <3CE6D976.3264DE53@mindspring.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org Hello, > > When using 32768 bytes MTU I can get around 190 Mbps out from a PIII 450. > > (and only 190 Mbps because the two frontends have fast ethernet cards) > > So why this is so bad? If the other end can keep up, it will increase > > throughput. > And you could get even better by getting rid of the request/response > turnaround stall by using TCP instead of UDP. Forgot to tell that these results are with TCP, not UDP! But as far as I can remember the original problem is still that with the gx driver I was unable to use a "standard" UDP NFS mount, because of the fragments (it worked with the em driver) and if I remember correctly it had problems with TCP too. My letter was about this: a warning that if people notice problems with the gx driver, they should try the em. It is often hard to find a driver which is not even in LINT... > Then don't add the fragment reassmbly code to the code path for packets > you send to the server. That way you'll have less overhead. I am not a big expert on this area, but if I get 200 Mbps instead of 15, I think increasing the packet size is good for me :) And going over the MTU with UDP also gives similar results. (the above is for TCP) > I run all my NFS over TCP. If I avoid intentionally triggering > fragmentation, it works out to a little over 100 machine instructions in > the fast path. Done any cycle counting on your use of UDP yet? I use TCP. I just noted that I *could* not use UDP mounts, with packet size bigger than the MTU with the gx driver. Which works with the em one, so either me, or the driver is buggy :) --------[ Free Software ISOs - ftp://ftp.fsn.hu/pub/CDROM-Images/ ]------- Attila Nagy e-mail: Attila.Nagy@fsn.hu Free Software Network (FSN.HU) phone @work: +361 210 1415 (194) cell.: +3630 306 6758 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message