From owner-freebsd-net@FreeBSD.ORG Fri Dec 12 06:07:53 2003 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id E046C16A4CE for ; Fri, 12 Dec 2003 06:07:53 -0800 (PST) Received: from mizar.origin-it.net (mizar.origin-it.net [194.8.96.234]) by mx1.FreeBSD.org (Postfix) with ESMTP id A567A43D09 for ; Fri, 12 Dec 2003 06:07:51 -0800 (PST) (envelope-from helge.oldach@atosorigin.com) Received: from matar.hbg.de.int.atosorigin.com (dehsfw3e.origin-it.net [194.8.96.68])hBCE7Fm3027867 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 12 Dec 2003 15:07:15 +0100 (CET) (envelope-from helge.oldach@atosorigin.com) Received: from galaxy.hbg.de.ao-srv.com (galaxy.hbg.de.ao-srv.com [161.89.20.4])ESMTP id hBCE7FQ8099563; Fri, 12 Dec 2003 15:07:15 +0100 (CET) (envelope-from helge.oldach@atosorigin.com) Received: (from hmo@localhost) by galaxy.hbg.de.ao-srv.com (8.9.3p2/8.9.3/hmo30mar03) id PAA10760; Fri, 12 Dec 2003 15:07:10 +0100 (MET) Message-Id: <200312121407.PAA10760@galaxy.hbg.de.ao-srv.com> In-Reply-To: <200312121346.03744.jrh@it.uc3m.es> from Juan Rodriguez Hervella at "Dec 12, 2003 1:46: 2 pm" To: jrh@it.uc3m.es (Juan Rodriguez Hervella) Date: Fri, 12 Dec 2003 15:07:09 +0100 (MET) From: Helge Oldach X-Address: Atos Origin GmbH, Friesenstraße 13, D-20097 Hamburg, Germany X-Phone: +49 40 7886 7464, Fax: +49 40 7886 9464, Mobile: +49 160 4782517 MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit cc: freebsd-net@freebsd.org cc: hali@ttsg.com cc: julian@elischer.org Subject: Re: grouping 2 or more interfaces as 1 X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 12 Dec 2003 14:07:54 -0000 Juan Rodriguez Hervella: >On Thursday 11 December 2003 23:14, Michael Sierchio wrote: >> Julian Elischer wrote: >> >>>more likely he wants something like ng_fec or ng_one2many >> >> >> >>Unless performance is the reason for bonding the ether channels... >> >> >> >>Can't we steal the Linux code? ;-) >> > >> > is the netgraph version particularly slow? >> >> Not slower than a single ether channel, no ;-) Considerably >> slower than link layer bonding. The netgraph version provides >> a really useful functionality, and I suppose that 2GB and 10GB >> fiber interfaces will do away with any pressure to give us >> bonding in the kernel. >> >For example, if we aggregate 4 ethernet cards into one >virtual interface (fec), do this mean that the throughput is >4 times the capacity of one ethernet card ?. In theory, yes. In practice, throughput is pretty often limited by PC architectural issues. Consider, for example, PCI bus speed... Also consider the overhead of actually distributing traffic between the physical interfaces... My personal experience tells me that channelling more than two FE interfaces tends to be a slightly pointless exercise. On the other hand, FECs are often implemented not for performance reasons but for resilience reasons. If you just need throughput, Gigabit is probably a better choice. Channeling of gigabit interfaces IMHO doesn't make sense, given the hardware choices that support FreeBSD. >Also, if the pyshical interfaces are connected to different LANs, They are not. A FEC (Fast Ether Channel) is a point-to-point link, commonly between a terminal device (computer) and a network device (switch). Both sides must have a common and identical understanding of the remote end, and both ends necessarily belong to the same single (V)LAN or 802.1Q trunk. Usually this also involves protocol support such as PAgP. Helge