From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans Henrik Happe Subject: Re: Netchannels: netchannel vs. socket. 2:0. Date: Fri, 9 Jun 2006 01:00:24 +0200 Message-ID: <200606090100.24827.hhh@imada.sdu.dk> References: <20060608171555.GA10273@2ka.mipt.ru> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from berlioz.imada.sdu.dk ([130.225.128.12]:16123 "EHLO berlioz.imada.sdu.dk") by vger.kernel.org with ESMTP id S965048AbWFHXA1 convert rfc822-to-8bit (ORCPT ); Thu, 8 Jun 2006 19:00:27 -0400 Received: from localhost (localhost [127.0.0.1]) by localhost.imada.sdu.dk (Postfix) with ESMTP id 77A5162706 for ; Fri, 9 Jun 2006 01:00:26 +0200 (CEST) Received: from berlioz.imada.sdu.dk ([127.0.0.1]) by localhost (berlioz.imada.sdu.dk [127.0.0.1]) (amavisd-new, port 10025) with ESMTP id 22937-02 for ; Thu, 8 Jun 2006 23:00:26 +0000 (UTC) Received: from [192.168.2.100] (unknown [85.218.152.97]) (using SSLv3 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by berlioz.imada.sdu.dk (Postfix) with ESMTP id 28BE962704 for ; Fri, 9 Jun 2006 01:00:26 +0200 (CEST) To: netdev@vger.kernel.org In-Reply-To: <20060608171555.GA10273@2ka.mipt.ru> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Thursday 08 June 2006 19:15, you wrote: > After some enhancements made for netchannel subsystem I'm pleased to > announce, that netchannel subsystem outperforms existing layered desi= gn > both in CPU usage and network speed. >=20 > Well, after such pretentious introduction I want to cool things down. > CPU usage is about 1-2% less for netchannels and network performance = is > about 1-2 MB higher and sometimes exceeds 84 MB/sec which, I think,=20 > is maximum for given network setup (e1000 receive, r8169 send, 1500 M= TU). I have followed your work closely and have wondered how it affects late= ncy? I have somewhat limited knowledge about TCP and how the kernel handles = it, but=20 I guess the path from NIC to userspace hasn't increased. What about sys= call=20 overhead caused by userspace TCP processing? H=B3 =20