From owner-freebsd-questions Thu Nov 30 10:40:20 2000 Delivered-To: freebsd-questions@freebsd.org Received: from 2711.dynacom.net (2711.dynacom.net [206.107.213.3]) by hub.freebsd.org (Postfix) with ESMTP id A8BA337B400 for ; Thu, 30 Nov 2000 10:40:13 -0800 (PST) Received: from urx.com (dsl1-160.dynacom.net [206.159.132.160]) by 2711.dynacom.net (Build 101 8.9.3/NT-8.9.3) with ESMTP id KAA04232; Thu, 30 Nov 2000 10:40:04 -0800 Message-ID: <3A269F04.3DEF4C32@urx.com> Date: Thu, 30 Nov 2000 10:40:04 -0800 From: Kent Stewart Reply-To: kstewart@urx.com Organization: Dynacom X-Mailer: Mozilla 4.76 [en] (Windows NT 5.0; U) X-Accept-Language: en MIME-Version: 1.0 To: David Kelly Cc: bob@sfcei.com, freebsd-questions@FreeBSD.ORG Subject: Re: ftp transfer rates on my LAN References: <20001130103627.B22943@HiWAAY.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG David Kelly wrote: > > On Thu, Nov 30, 2000 at 09:33:24AM -0500, bob@sfcei.com wrote: > > Just my .02 here. IIRC, Windows seems to max out around 1472 for the MTU, > > otherwise there are lots of collisions on the net. I set the MTU on my FBSD > > box to 1472 and no collisions. This may affect your throughput, if only > > marginally. > > Ethernet collisions are not bad. Don't sweat 'em until/if they reach > 150% to 200%. A 1500 octet packat takes a while to send. But a"collision" > happens in the first 64, takes very little wire time. Many NICs do not > report these collisions at all. Don't believe I've seen them on 3com > NICs I have used. Rather those NICs report *late* collisions which are > bad, very bad, indicating a protocol implementation error, hardware > failure, or a network which is too long. Yup, the network can get so > big the speed of light is no longer fast enough to meet the ethernet > timing specs. > > An analysis I no longer can find the URL for showed a 200% collision > rate on 10 Mbps ethernet resulted in an 8% reduction in network > capacity. So don't sweat the collisions. > > In this thread the user has two machines connected point-to-point with > a crossed cable. No way for collisions to occur. There might be some > advantage to turning on full duplex, which I've never seen auto-negotiated > when connected that way, only when connected to a switch which does > something to suggest to the machine that full duplex is available. I haven't measured what the affect is but Window's has a registry key called TCPWindowSize on W2K that you can set to 16k and improve throughput on files being sent to the Windows machine. I was told that their receive windows size is setup to ack/nack records over dialup and is really to small for real networks. I know where the key is on W2K but not off of my head for the 9x variety. Kent > > -- > David Kelly N4HHE, dkelly@hiwaay.net (hm) > ====================================================================== > The human mind ordinarily operates at only ten percent of its > capacity -- the rest is overhead for the operating system. > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-questions" in the body of the message -- Kent Stewart Richland, WA mailto:kbstew99@hotmail.com http://kstewart.urx.com/kstewart/index.html FreeBSD News http://daily.daemonnews.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message