From mboxrd@z Thu Jan 1 00:00:00 1970 From: "=?utf-8?q?R=C3=A9mi?= Denis-Courmont" Subject: Re: TCP rx window autotuning harmful at LAN context Date: Wed, 11 Mar 2009 11:02:31 +0200 Message-ID: <200903111102.31599.remi.denis-courmont@nokia.com> References: <20090309112521.GB37984@bts.sk> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: "netdev@vger.kernel.org" To: ext Marian =?utf-8?q?=C4=8Eurkovi=C4=8D?= Return-path: Received: from smtp.nokia.com ([192.100.122.233]:39704 "EHLO mgw-mx06.nokia.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751627AbZCKJCe convert rfc822-to-8bit (ORCPT ); Wed, 11 Mar 2009 05:02:34 -0400 In-Reply-To: <20090309112521.GB37984@bts.sk> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: On Monday 09 March 2009 13:25:21 ext Marian =C4=8Eurkovi=C4=8D wrote: > The behaviour could be descibed as "spiraling death" syndrome. Whil= e > TCP with constant and decently sized rx window natively reduces > transmission rate when RTT increases, autotuning performs exactly the > opposite - as a response to increased RTT it increases the rx window = size > (which in turn again increases RTT...) As this happens again and agai= n, the > result is complete waste of all available buffers at sending host or = at the > bottleneck point, resulting in upto 267 msec (!) latency in LAN conte= xt > (with 100 Mbps ethernet connection, default txqueuelen=3D1000, MTU=3D= 1500 and > sky2 driver). Needles to say that this means the LAN is almost unusab= le. This is very likely a stupid question, but anyway... Is this with all applications, or only some pathological ones (one of w= hich we=20 both wrote code for, alright) with abnormally large send buffers? --=20 R=C3=A9mi Denis-Courmont Maemo Software, Nokia Devices R&D