From owner-freebsd-hackers Tue Jan 27 13:21:43 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id NAA14097 for hackers-outgoing; Tue, 27 Jan 1998 13:21:43 -0800 (PST) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from friley585.res.iastate.edu (friley585.res.iastate.edu [129.186.167.85]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id NAA14053 for ; Tue, 27 Jan 1998 13:21:31 -0800 (PST) (envelope-from ccsanady@friley585.res.iastate.edu) Received: from friley585.res.iastate.edu (loopback [127.0.0.1]) by friley585.res.iastate.edu (8.8.8/8.8.5) with ESMTP id PAA06499; Tue, 27 Jan 1998 15:21:10 -0600 (CST) Message-Id: <199801272121.PAA06499@friley585.res.iastate.edu> X-Mailer: exmh version 2.0.1 12/23/97 To: Amancio Hasty cc: freebsd-hackers@FreeBSD.ORG Subject: Re: Gigabit ethernet cards for FreeBSD? In-reply-to: Your message of "Tue, 27 Jan 1998 11:26:41 PST." <199801271926.LAA11444@rah.star-gate.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Tue, 27 Jan 1998 15:21:10 -0600 From: Chris Csanady Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk >Just wondering if there are any for FreeBSD , how they perform and how >much they cost? Packet Engines makes one, as well as a full duplex repeater. The NIC's are somewhere in the $1000-1500 range I believe. As for performance, it's not quite gigabit yet. With 2 PPro200's, the TCP performance for larger write sizes is still limited to around 20MB/s. With 533lx alphas, its around 30MB/s or so (the PCI implementation on these boards is completely broken though.) Sadly, NT seems to outperform Linux on the alphas. I have written a FreeBSD driver for these cards. The driver still needs some work, but it at least functions most of the time. :) See www.packetengines.com for more info. Although this hardware does not have an onboard MCP, the problem remains that the BSD TCP/IP stack is simply not suited for these speeds. There really should (at least) be support a combined checksum/copy. Some of the newer boards even have the capability to do the TCP checksum on the board itself. The overhead associated with mbuf chains, and various other things have a significant overhead. Is there anyone interested in doing some work on this? I have been planning on doing it for a while, but simply have not had enough time. The general idea was to implement something along the lines of what Van Jacobson describes in his talks. It would be nice if someone could get source from him, but so far I have had little luck. The viewgraphs are at http://www-nrg.ee.lbl.gov/nrg-talks.html "Design Changes to the Kernel Network Architecture for 4.4BSD" and "Some Design Issues for High-speed Networks" Hope this helps.. Chris Csanady > Tnks, > Amancio > >