From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932145AbWIGPaA (ORCPT ); Thu, 7 Sep 2006 11:30:00 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932146AbWIGP37 (ORCPT ); Thu, 7 Sep 2006 11:29:59 -0400 Received: from omx1-ext.sgi.com ([192.48.179.11]:4526 "EHLO omx1.americas.sgi.com") by vger.kernel.org with ESMTP id S932145AbWIGP36 (ORCPT ); Thu, 7 Sep 2006 11:29:58 -0400 Date: Thu, 7 Sep 2006 08:29:45 -0700 (PDT) From: Christoph Lameter To: Nick Piggin cc: akpm@osdl.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH] Fix longstanding load balancing bug in the scheduler. In-Reply-To: <20060907105801.GC3077@wotan.suse.de> Message-ID: References: <20060907105801.GC3077@wotan.suse.de> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 7 Sep 2006, Nick Piggin wrote: > So what I worry about with this approach is that it can really blow > out the latency of a balancing operation. Say you have N-1 CPUs with > lots of stuff locked on their runqueues. > > The solution I envisage is to do a "rotor" approach. For example > the last attempted CPU could be stored in the starving CPU's sd... > and it will subsequently try another one. > > I've been hot and cold on such an implementation for a while: on one > hand it is a real problem we have; OTOH I was hoping that the domain > balancing might be better generalised. But I increasingly don't > think we should let perfect stand in the way of good... ;) > > Would you be interested in testing a patch? Sure but I think we should move fast on this one. This has now been known for around a year or so.