From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: Shaping TCP ACK packets Date: Tue, 06 Dec 2011 06:59:26 +0100 Message-ID: <1323151166.2467.11.camel@edumazet-laptop> References: <7cf7b0f6-b30f-465a-9e50-20bd2c31bbe9@jasiiieee> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev@vger.kernel.org To: "John A. Sullivan III" Return-path: Received: from mail-ee0-f46.google.com ([74.125.83.46]:41507 "EHLO mail-ee0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752567Ab1LFF7a (ORCPT ); Tue, 6 Dec 2011 00:59:30 -0500 Received: by eeaq14 with SMTP id q14so2545077eea.19 for ; Mon, 05 Dec 2011 21:59:29 -0800 (PST) In-Reply-To: <7cf7b0f6-b30f-465a-9e50-20bd2c31bbe9@jasiiieee> Sender: netdev-owner@vger.kernel.org List-ID: Le lundi 05 d=C3=A9cembre 2011 =C3=A0 23:18 -0500, John A. Sullivan III= a =C3=A9crit : > Hello, all. I've seen several sites recommend prioritizing TCP ACK > packets using tc. Is that a good idea? I have two reservations: >=20 > 1) If my line is congested, isn't delaying ACKs a good way to slow > down the conversation? Accelerating ACKs under congestion sounds a bi= t > like stuffing more food in my mouth when I'm already choking. >=20 It depends which way is congested. Sites recommends prioritizing ACKS o= n the up way, because most of the time ADSL links are assymetric. an uplink is congested. > 2) By messing with normal coordination of TCP, do I risk making a > mess. Specifically, can I create problems with out of order packet > delivery - send an ACK before the packet which should have preceded > it? I suppose in a well behaved protocol that wouldn't happen but I'm > not sure. Thanks - John > -- Well, if you prioritize true acks (packets with no payload), there is n= o reorder problem per se.