From mboxrd@z Thu Jan 1 00:00:00 1970 From: Lucas Nussbaum Subject: Re: [PATCH 0/6] TCP CUBIC and Hystart Date: Fri, 11 Mar 2011 17:08:09 +0100 Message-ID: <20110311160809.GA30082@xanadu.blop.info> References: <20110310165119.224046957@vyatta.com> <20110311102809.GB15941@xanadu.blop.info> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Stephen Hemminger , davem@davemloft.net, rhee@ncsu.edu, netdev@vger.kernel.org To: Sangtae Ha Return-path: Received: from xanadu.blop.info ([178.79.145.134]:60686 "EHLO xanadu.blop.info" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754482Ab1CKQ2T (ORCPT ); Fri, 11 Mar 2011 11:28:19 -0500 Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On 11/03/11 at 10:58 -0500, Sangtae Ha wrote: > Hi Lucas, >=20 > From your setup, ca->delay_min is 90 and this means the one-way delay > is 90 >> 4 (5ms). > And our gap detection threshold is 2ms, which means that if the gap i= s > loosely spread over 5ms with delayed ACKs, it can early terminate the > slow start. But, given the optimal cwnd is 941 in your setup, exiting > slow start one RTT before the loss (half of the optimal cwnd) is what > hystart does. >=20 > Since the resolution is now ms, can you change the gap detection to > 1ms and run it again? > Also, the following change you did doesn't hurt (1ms more train to > detect the ACK train). Hi, Changing it to 1ms only improves the situation marginally. [25271.861481] found ACK TRAIN: cwnd=3D835 now=3D2779244747 ca->last_ac= k=3D2779244747 ca->round_start=3D2779244741 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D261 [25291.507340] found ACK TRAIN: cwnd=3D585 now=3D2779264393 ca->last_ac= k=3D2779264393 ca->round_start=3D2779264387 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D221 [25327.585396] found ACK TRAIN: cwnd=3D1034 now=3D2779300471 ca->last_a= ck=3D2779300471 ca->round_start=3D2779300465 ca->delay_min=3D90 delay_m= in>>4=3D5 nbacks=3D245 [25347.300351] found ACK TRAIN: cwnd=3D1463 now=3D2779320186 ca->last_a= ck=3D2779320186 ca->round_start=3D2779320180 ca->delay_min=3D90 delay_m= in>>4=3D5 nbacks=3D427 [25390.702328] found ACK TRAIN: cwnd=3D587 now=3D2779363588 ca->last_ac= k=3D2779363588 ca->round_start=3D2779363582 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D211 [25394.775396] found ACK TRAIN: cwnd=3D588 now=3D2779367661 ca->last_ac= k=3D2779367661 ca->round_start=3D2779367655 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D242 [25402.061328] found ACK TRAIN: cwnd=3D1282 now=3D2779374947 ca->last_a= ck=3D2779374947 ca->round_start=3D2779374941 ca->delay_min=3D90 delay_m= in>>4=3D5 nbacks=3D335 [25404.894336] found ACK TRAIN: cwnd=3D585 now=3D2779377780 ca->last_ac= k=3D2779377780 ca->round_start=3D2779377774 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D205 [25408.584337] found ACK TRAIN: cwnd=3D587 now=3D2779381470 ca->last_ac= k=3D2779381470 ca->round_start=3D2779381464 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D209 [25421.699331] found ACK TRAIN: cwnd=3D856 now=3D2779394585 ca->last_ac= k=3D2779394585 ca->round_start=3D2779394579 ca->delay_min=3D90 delay_mi= n>>4=3D5 nbacks=3D239 There are still some cases when ack trains are detected too early. --=20 | Lucas Nussbaum MCF Universit=E9 Nancy 2 | | lucas.nussbaum@loria.fr LORIA / AlGorille | | http://www.loria.fr/~lnussbau/ +33 3 54 95 86 19 |