From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [PATCH] rps: make distributing packets fairly among all the online CPUs default Date: Tue, 23 Mar 2010 15:05:31 +0100 Message-ID: <874ok7dtg4.fsf@basil.nowhere.org> References: <4BA85EA2.1090304@gmail.com> <1269329250.3043.19.camel@edumazet-laptop> <412e6f7f1003230303l69fc142ey1581fff4c3e749be@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Eric Dumazet , "David S. Miller" , Tom Herbert , netdev@vger.kernel.org, Ben Hutchings To: Changli Gao Return-path: Received: from one.firstfloor.org ([213.235.205.2]:49822 "EHLO one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753568Ab0CWOFd (ORCPT ); Tue, 23 Mar 2010 10:05:33 -0400 In-Reply-To: <412e6f7f1003230303l69fc142ey1581fff4c3e749be@mail.gmail.com> (Changli Gao's message of "Tue, 23 Mar 2010 18:03:22 +0800") Sender: netdev-owner@vger.kernel.org List-ID: Changli Gao writes: > > I do think distributing packets fairly among all the online CPUs will > helps most of users. I remember the smp_affinity of IRQ is the > online_cpu_mask be default. The established wisdom is that most apps prefer locality. The cost of data transfer between CPUs is worse than the overhead of processing the packet header in the stack. That's why no system does interrupt RR by default anymore. Besides if you do such a change you would need extensive benchmarking to verify that it's a good idea. -Andi -- ak@linux.intel.com -- Speaking for myself only.