From owner-freebsd-questions Fri Dec 4 17:38:29 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id RAA18882 for freebsd-questions-outgoing; Fri, 4 Dec 1998 17:38:29 -0800 (PST) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from n4hhe.ampr.org (tnt4-138.HiWAAY.net [208.166.127.138]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id RAA18873 for ; Fri, 4 Dec 1998 17:38:26 -0800 (PST) (envelope-from dkelly@n4hhe.ampr.org) Received: from n4hhe.ampr.org (localhost.ampr.org [127.0.0.1]) by n4hhe.ampr.org (8.9.1/8.9.1) with ESMTP id SAA19055; Fri, 4 Dec 1998 18:58:35 -0600 (CST) (envelope-from dkelly@n4hhe.ampr.org) Message-Id: <199812050058.SAA19055@n4hhe.ampr.org> X-Mailer: exmh version 2.0.2 2/24/98 To: Reinier Bezuidenhout cc: freebsd-questions@FreeBSD.ORG From: David Kelly Subject: Re: TCP and collisions In-reply-to: Message from Reinier Bezuidenhout of "Fri, 04 Dec 1998 16:06:30 +0100." <199812041506.QAA00427@borg.kryptokom.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Fri, 04 Dec 1998 18:58:34 -0600 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Reinier Bezuidenhout writes: > Hi ... > > I've been doing some throughput test etc. ans saw that wen > 'n transfer large amounts of data (ttcp or ftp) 'n get > round about 50% collision rate (according to netstat -i > and some mathematics :) ) > > Is this normal ??? Yes, there is nothing wrong with a modest 50% collision rate. Start worrying when it consistantly gets over 150% or 200%. > I used a x-over cable and a hub between the two machines, both > running 2.2.7 ... one 233 PII and the other 300 PII Why did you use a crossed up cable when connecting thru a hub? If there are only two machines on the net, use a crossed cable, no hub, and look into enabling full duplex. Full duplex will eliminate collisions. > Although there was such a high rate of collisions the throughput > was still +/- 880 to 990 kbytes/sec That's because they are "early collisions" which happen in the first 64 octets of the ethernet packet. So not much time was lost. When it happens both senders back off a random time and try again. Guess I could figure out how to calculate it, but the numbers I've heard claim 100% or 200% collision rate might result in a 15% thruput degradation. A late collision happens after the 64th octet. Indicates deficient hardware or software. Or that your network is too long, too many hubs, or something to cause a host to be too far away. Simplified, that host is so far away the speed of light (thru wire and hubs) is longer than the time it takes to send 64 octets. Not all ethernet hardware bother to report early collisions. -- David Kelly N4HHE, dkelly@nospam.hiwaay.net ===================================================================== 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