Return-path: Received: from mail-pw0-f46.google.com ([209.85.160.46]:35931 "EHLO mail-pw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750814Ab0CHFUG convert rfc822-to-8bit (ORCPT ); Mon, 8 Mar 2010 00:20:06 -0500 Received: by pwj8 with SMTP id 8so3557398pwj.19 for ; Sun, 07 Mar 2010 21:20:05 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <4B93F42F.7030306@openwrt.org> References: <4B93F42F.7030306@openwrt.org> Date: Sun, 7 Mar 2010 21:20:05 -0800 Message-ID: <70f6a1ed1003072120w37fa176ya2c3a8af742dfd97@mail.gmail.com> Subject: Re: [RFC v3] minstrel_ht: new rate control module for 802.11n From: Duy Nguyen To: Felix Fietkau Cc: linux-wireless , Derek Smithies , Benoit PAPILLAULT , "Luis R. Rodriguez" , Christian Lamparter , =?ISO-8859-1?Q?Bj=F6rn_Smedman?= , Bob Copeland Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-wireless-owner@vger.kernel.org List-ID: Hi Felix, thanks for the awesome patch and great progress on the 802.11n! I just tried out your patch version 3, and got the following results. The results are an average of 10 different iperf runs, Throughputs are in Mbits/sec AP: dir-615 (settings: 802.11n mode only, Auto 20/40 Mhz). This AP does not seem to allow me to set 40Mhz only Node interfaces: D-Link DWA-556 Xtreme N PC scenario 1: node 1 to node 2 node1-----wireless_link-----AP-----wireless_link-----node2 802.11n Felix's Minstrel Avg Transfer= 38.6 Avg Bandwidth=32.3 802.11n Atheros Default Avg Transfer= 36.6 Avg Bandwidth=30.6 scenario 2: node 1 to node 2 node1-----wireless_link-----AP-----wired_link-----node2 802.11n Felix's Minstrel Avg Transfer= 83.4 Avg Bandwidth=69.8 802.11n Atheros Default Avg Transfer= 95.2 Avg Bandwidth=79.8 Overall, the results look pretty good. I am much more interested in results in scenario 1 but I am going to take a look at scenario 2 again. Let me know if I can help you test any other scenarios. Duy On Sun, Mar 7, 2010 at 10:45 AM, Felix Fietkau wrote: > Here's v3 of minstrel_ht. > > Changelog: > > v3: > - Add two fixes by Christian Lamparter (division by zero, debugfs mcs > ?rate display) > - Improve throughput a bit more by spreading out slow rate sampling > ?attempts over multiple sampling intervals > > v2: > - Use accurate A-MPDU statistics from tx feedback > - Rewrite the sampling algorithm to optimize for good A-MPDU lengths > ?and faster throughput recovery > - Implement .rate_update to handle changes in HT capabilities or > ?channel bandwidth > - Remove the private tx flags hack