From owner-freebsd-net@FreeBSD.ORG Mon Sep 23 10:02:51 2013 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTP id 340303A0 for ; Mon, 23 Sep 2013 10:02:51 +0000 (UTC) (envelope-from lists@rewt.org.uk) Received: from hosted.mx.as41113.net (abby.lhr1.as41113.net [91.208.177.20]) by mx1.freebsd.org (Postfix) with ESMTP id EE9E02D29 for ; Mon, 23 Sep 2013 10:02:48 +0000 (UTC) Received: from [192.168.1.218] (unknown [212.9.98.193]) (using TLSv1 with cipher ECDHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) (Authenticated sender: lists@rewt.org.uk) by hosted.mx.as41113.net (Postfix) with ESMTPSA id 3ck1Ch0czxzCY for ; Mon, 23 Sep 2013 10:57:04 +0100 (BST) Message-ID: <5240106F.6010105@rewt.org.uk> Date: Mon, 23 Sep 2013 10:57:03 +0100 From: Joe Holden User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/20130801 Thunderbird/17.0.8 MIME-Version: 1.0 To: freebsd-net@freebsd.org Subject: Re: Network stack changes References: <521E41CB.30700@yandex-team.ru> <523F4F14.9090404@yandex-team.ru> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 23 Sep 2013 10:02:51 -0000 On 23/09/2013 06:34, Adrian Chadd wrote: > On 22 September 2013 21:52, Luigi Rizzo wrote: > > >> locking affects scalability; but dxr and similar algorithms have much fewer >> memory lookups, not to mention the huge memory footprint of >> the freebsd radix tree code. >> >> Anyways i'd really encourage you to read the dxr paper, it is short >> and hopefully can give you a better idea of the details (and with data >> supporting them) than these short notes. >> >> > I read the paper. :-) > > I believe it! It's not the first paper that I've read that packed a FIB > into a sensibly cacheable structure. I'm just as interested however in > making sure that we actually give people the tools to inspect this stuff > for themselves, rather than all of us hacking up something from scratch > every time we want to profile this kind of thing. > > The other side of this coin is locking, and the paper didn't go into that. > Eliminating the radix tree overhead is great; now we just have to avoid > grabbing all those locks all the damned time for each frame.. > > > > -adrian The paper is insane, awesome work - actually faster and more efficient than many popular ASICs.... is there any chance of a patch for HEAD/9 being made? I had a go at forward porting it but there are too many changes that I don't understand :( Cheers, Joe