From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.taht.net (mail.taht.net [IPv6:2a01:7e00::f03c:91ff:feae:7028]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by lists.bufferbloat.net (Postfix) with ESMTPS id 12F353CB99; Sat, 6 Feb 2016 13:01:00 -0500 (EST) Received: from dair-1120.local (c-73-252-201-217.hsd1.ca.comcast.net [73.252.201.217]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by mail.taht.net (Postfix) with ESMTPSA id A1D1221322; Sat, 6 Feb 2016 18:00:58 +0000 (UTC) To: Trevor Paskett , make-wifi-fast@lists.bufferbloat.net, cerowrt-devel@lists.bufferbloat.net References: <56AA90C9.6050504@taht.net> <56B0FE22.3050704@taht.net> <73AA134D-76DC-4D37-94F0-CC3C4FEAF5C6@gmail.com> <56B12DE1.8040009@taht.net> <79F77DD2-2BFF-4697-A118-25D17FD06CBB@gmail.com> <56B15105.7040405@taht.net> <0BAB0F1C-9F93-4579-9426-648CBE1EF5D5@gmail.com> <56B63105.5060909@taht.net> From: =?UTF-8?Q?Dave_T=c3=a4ht?= X-Enigmail-Draft-Status: N1110 Message-ID: <56B635A1.8080603@taht.net> Date: Sat, 6 Feb 2016 10:04:17 -0800 User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:38.0) Gecko/20100101 Thunderbird/38.5.1 MIME-Version: 1.0 In-Reply-To: <56B63105.5060909@taht.net> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Subject: Re: [Make-wifi-fast] AREDN and Make Wifi Fast X-BeenThere: make-wifi-fast@lists.bufferbloat.net X-Mailman-Version: 2.1.20 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 06 Feb 2016 18:01:01 -0000 The simplest thing you can do today to improve latency on your ath9k devices is to disable wmm (802.11e), and to reduce the ath9k driver queues to smaller sizes. I typically run the yurtlab network with a qlen_be of 12 (slow links) to 24 rather than the default. On links with significant speed issues, congestion or interference, I typically increase the fq_codel target to 20ms. We have fiddled with using large quantums with per-dst FQ on ptp links with some success on top of that. In all cases it helps to have a set of flent runs for each change. Throughput will drop, fairness and latency will improve. The yurtlab userbase is generally oblivious to a speed problem, it's having low latency for gaming, dns, and basic web traffic that works for them. None of the above are "optimal" fixes in any way, but perhaps you will get some use from them.