From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rick Jones Subject: Re: [PATCH] Re: Bad network performance over 2Gbps Date: Mon, 21 Apr 2008 09:42:00 -0700 Message-ID: <480CC3D8.3040700@hp.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Cc: Kok@vger.kernel.org, Auke , Ingo Molnar , Thomas Gleixner , "H. Peter Anvin" , Linux Kernel Mailing List , Anton Titov , Chris Snook , "H. Willstrand" , netdev@vger.kernel.org, Jesse Brandeburg , Linus Torvalds , Andrew Morton To: 7eggert@gmx.de Return-path: Received: from g5t0007.atlanta.hp.com ([15.192.0.44]:31131 "EHLO g5t0007.atlanta.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754811AbYDUQmH (ORCPT ); Mon, 21 Apr 2008 12:42:07 -0400 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: Bodo Eggert wrote: > Kok, Auke wrote: > > >>[X86] IRQBALANCE: Mark as BROKEN and disable by default >> >>The IRQBALANCE option causes interrupts to bounce all around on SMP systems >>quickly burying the CPU in migration cost and cache misses. Mainly affected >>are network interrupts and this results in one CPU pegged in softirqd >>completely. > > > If this is the problem, maybe it would help to only balance the IRQs each > e.g. ten seconds? Unfortunately I have no SMP system to try it out. Be it kernel or user space, for consistent benchmark results it needs to be able to be turned-off without turning the code. That leaves me in agreement with Stephen that if it must exist, the user space one would be preferable. It can be easily terminated with extreme prejudice. rick jones