From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: Socket buffer sizes with autotuning Date: Fri, 25 Apr 2008 09:48:34 +0200 Message-ID: <48118CD2.6000706@firstfloor.org> References: <48118308.1090407@firstfloor.org> <20080425.002804.152322163.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: hkchu@google.com, johnwheffner@gmail.com, rick.jones2@hp.com, netdev@vger.kernel.org To: David Miller Return-path: Received: from one.firstfloor.org ([213.235.205.2]:53108 "EHLO one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760107AbYDYHsi (ORCPT ); Fri, 25 Apr 2008 03:48:38 -0400 In-Reply-To: <20080425.002804.152322163.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: David Miller wrote: > From: Andi Kleen > Date: Fri, 25 Apr 2008 09:06:48 +0200 > >> The only big problem then would be if there are more submitting threads >> than packets in the TX queue, but I would consider that unlikely for >> GB+ NICs at least (might be an issue for older designs with smaller queues) > > It's probably exactly what happens in those chat server benchmarks. In that case we might be better off just blocking some of these threads until the TX queue cleared? What good is it to send faster than the network allows? -Andi