From owner-freebsd-arch@FreeBSD.ORG Wed Sep 1 00:05:13 2004 Return-Path: Delivered-To: freebsd-arch@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 5FACE16A4CE; Wed, 1 Sep 2004 00:05:13 +0000 (GMT) Received: from pooker.samsco.org (pooker.samsco.org [168.103.85.57]) by mx1.FreeBSD.org (Postfix) with ESMTP id 196A343D3F; Wed, 1 Sep 2004 00:05:08 +0000 (GMT) (envelope-from scottl@samsco.org) Received: from [192.168.0.201] ([192.168.0.201]) (authenticated bits=0) by pooker.samsco.org (8.12.11/8.12.10) with ESMTP id i8104c2L073499; Tue, 31 Aug 2004 18:04:38 -0600 (MDT) (envelope-from scottl@samsco.org) Message-ID: <4135118A.5030807@samsco.org> Date: Tue, 31 Aug 2004 18:02:18 -0600 From: Scott Long User-Agent: Mozilla/5.0 (X11; U; FreeBSD i386; en-US; rv:1.7.2) Gecko/20040831 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Julian Elischer References: <4134DF35.7070605@freebsd.org> <20040831203929.GB25134@odin.ac.hmc.edu> <4134E4B6.2030409@elischer.org> <4134FCAE.7374599A@freebsd.org> <4134FF74.4010105@freebsd.org> <4135051E.2070007@elischer.org> In-Reply-To: <4135051E.2070007@elischer.org> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Spam-Status: No, hits=0.0 required=3.8 tests=none autolearn=no version=2.63 X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on pooker.samsco.org cc: Andre Oppermann cc: Sam cc: Scott Long cc: freebsd-arch@freebsd.org Subject: Re: option directive and turning on AOE X-BeenThere: freebsd-arch@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Discussion related to FreeBSD architecture List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 01 Sep 2004 00:05:13 -0000 Julian Elischer wrote: > > > Scott Long wrote: > >> Andre Oppermann wrote: >> >>> >> >> Having a single common interface is definitely attractive, but there are >> performance and locking issues with the Netgraph framework that should >> probably be resolved first. > > > both of these issues are in fact not major.. > netgraph itself has no locking issuess.. (Netgraph is the framework), > but some of teh node types have issues. Specifically, node types that > can be caled > from outside the netgraph framework, such as nodes that tie netgraph to > other subsystems > need to be worked on so that control enterring netgraph code from those > subsystems > gets an appropriate lock. This is not always needed, but every node > needs to be > examined with this in mind now that we have locking in teh rest of the > system > more worked out. > > performace.. well it can be fast. it depends on a lot of issues however.. > in particular how many locks get contentions. > >> >> >> Scott > > > My employer has done extensive profiling of packet delivery through netgraph. While the locking of the netgraph framework is definitely correct, it's not terribly efficient and leads to a good deal of latency. We are looking at various proposals on how to address this. This isn't a criticism of you or Netgraph, just a set 'real-life' observations under very high load (bridging and packet inspection on 4 GigE links simultaneously qualifies as high load =-) Scott