From owner-freebsd-questions Wed Feb 3 03:01:01 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id DAA11849 for freebsd-questions-outgoing; Wed, 3 Feb 1999 03:01:01 -0800 (PST) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from implode.root.com (root.com [208.221.12.98]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id DAA11844 for ; Wed, 3 Feb 1999 03:01:00 -0800 (PST) (envelope-from root@implode.root.com) Received: from implode.root.com (localhost [127.0.0.1]) by implode.root.com (8.8.8/8.8.5) with ESMTP id CAA01058 for ; Wed, 3 Feb 1999 02:59:55 -0800 (PST) Message-Id: <199902031059.CAA01058@implode.root.com> To: FreeBSD Questions Subject: Re: Excessive collisions on Ethernet From: David Greenman Reply-To: dg@root.com Date: Wed, 03 Feb 1999 02:59:55 -0800 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > > Anything less than 50% collision rate is okay and doesn't reduce the > > throughput significantly. All of the numbers below are in the noise. > > You mean it's the noise that's causing the collisions? > > Seriously, it's clear that it's not affecting the performance, though > that surprises me; how long does a board hold off after a collision? > I haven't been able to find that info anywhere, but I once worked for > a nameless computer manufacturer whose net had about 50% collisions, > and the throughput was terrible. > > The real question was: why is this happening? I still suspect that > it's trying to tell me something, but I haven't a good feeling for > what. The collisions are caused by the packet acks in the opposite direction. Collisions are usually detected very early in the transmission and the backoff is usually very short. Modern hardware leaves no room for the acks to slip in without colliding, so you get a lot more collisions. A high collision rate with a small number of hosts involved is usually nothing to worry about. It's when you have a lot of machines all trying to talk that you get into trouble (...but then the collision rate in that case can easily be several hundred percent). In the nameless comp manufacturer case above, they probably had other problems with their net that caused CRC errors and dropped packets...very different from collisions which are retried very quickly and for up to 16 times before the packet is dropped. -DG David Greenman Co-founder/Principal Architect, The FreeBSD Project To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message