From owner-freebsd-questions Thu Jan 17 6:34:45 2002 Delivered-To: freebsd-questions@freebsd.org Received: from lv.raad.tartu.ee (lv.raad.tartu.ee [194.126.106.110]) by hub.freebsd.org (Postfix) with ESMTP id C5A1B37B404 for ; Thu, 17 Jan 2002 06:34:38 -0800 (PST) Received: Message by Barricade lv.raad.tartu.ee with ESMTP id g0HEY9P26471; Thu, 17 Jan 2002 16:34:09 +0200 Message-Id: <200201171434.g0HEY9P26471@lv.raad.tartu.ee> Received: from INFO/SpoolDir by raad.tartu.ee (Mercury 1.48); 17 Jan 02 16:33:50 +0200 Received: from SpoolDir by INFO (Mercury 1.48); 17 Jan 02 16:33:47 +0200 From: "Toomas Aas" Organization: Tartu City Government To: Cliff Sarginson , freebsd-questions@freebsd.org Date: Thu, 17 Jan 2002 16:33:37 +0200 MIME-Version: 1.0 Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Subject: Re: Is this ping normal? In-reply-to: <20020117135038.GB2371@raggedclown.net> References: <200201170836.g0H8aLN20447@lv.raad.tartu.ee> Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Hi Cliff! On 17 Jan 02 at 14:50 you wrote: > On Thu, Jan 17, 2002 at 10:35:55AM +0200, Toomas Aas wrote: > > Full Duplex Is Evil. > I think you need to qualify such a statement. I thought someone will say that :-) This issue has been discussed to death on a Novell (anyone remember Netware?) mailinglist that I belong to. Also, I have had personal experience which supports this point of view. The question we need to consider is - what happens when network load approaches maximum? In half duplex environment, this will cause a lot of collisions at data link layer (OSI layer 2), which the sending party can detect and retry sending the frame. This is a matter of milliseconds. In full duplex environment, by definition, there are no collisions. This means that once the network load reaches the point where receiving end is unable to process all the incoming packets, some of the incoming packets will be silently dropped on the floor and sender has no indication whatsoever that something went wrong. The problems can only be detected on higher layers of OSI model (in case of TCP/IP it's layer 4, TCP) and have to be rectified there. This can often take *seconds*. The conclusion is that full duplex behaves very badly under heavy load, which is exactly when you would actually need the advertised "increased bandwith" of full duplex. > My small network runs full duplex 100mbps with no > problems whatsoever. I don't doubt at all that there are many installations where full duplex works just fine. But after having been bitten myself and reading what wiser people have said about the issue, I try to avoid full duplex whenever possible. -- Toomas Aas | toomas.aas@raad.tartu.ee | http://www.raad.tartu.ee/~toomas/ * If you try to fail, and succeed, which have you done? To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message