From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: limited network bandwidth with 3.2.x kernels Date: Thu, 16 Feb 2012 20:44:52 +0100 Message-ID: <1329421492.2645.23.camel@edumazet-laptop> References: <1329399616.9395.9.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> <1329400265.3048.178.camel@deadeye> <1329404975.10004.3.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> <20120216.120149.437368615973561397.davem@davemloft.net> <1329416523.2645.7.camel@edumazet-laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , bhutchings@solarflare.com, netdev@vger.kernel.org, alekcejk@googlemail.com To: Neal Cardwell Return-path: Received: from mail-bk0-f46.google.com ([209.85.214.46]:37157 "EHLO mail-bk0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751800Ab2BPTo6 (ORCPT ); Thu, 16 Feb 2012 14:44:58 -0500 Received: by bkcjm19 with SMTP id jm19so2355633bkc.19 for ; Thu, 16 Feb 2012 11:44:57 -0800 (PST) In-Reply-To: <1329416523.2645.7.camel@edumazet-laptop> Sender: netdev-owner@vger.kernel.org List-ID: Le jeudi 16 f=C3=A9vrier 2012 =C3=A0 19:22 +0100, Eric Dumazet a =C3=A9= crit : > On sender side, there are also issues, because if I have TSO on or of= f > on my 'netem delay 50ms' machine, performance is completely different= =2E Well, my sender is a 32bit kernel, so tcp_wmem[2] was less than 1Mbytes.. $ cat /proc/sys/net/ipv4/tcp_wmem 4096 16384 903780 Socket cwnd is then capped to ~400 with TSO=3Doff, and ~1500 with TSO=3D= on TSO off, a 1500 bytes frame now has truesize=3D2048+192, while in previ= ous kernel it had truesize=3D1500+192 before linux 3.2 (903780/2240 =3D 403) (TSO permits a better truesize/len ratio, close to 1) With 50ms RTT, 400 packets in flight means no more than 90Mbit/s