From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757562Ab3GLK20 (ORCPT ); Fri, 12 Jul 2013 06:28:26 -0400 Received: from cantor2.suse.de ([195.135.220.15]:54505 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757244Ab3GLK2Z (ORCPT ); Fri, 12 Jul 2013 06:28:25 -0400 Date: Fri, 12 Jul 2013 11:28:19 +0100 From: Mel Gorman To: Peter Zijlstra Cc: Srikar Dronamraju , Ingo Molnar , Andrea Arcangeli , Johannes Weiner , Linux-MM , LKML Subject: Re: [PATCH 08/16] sched: Reschedule task on preferred NUMA node once selected Message-ID: <20130712102819.GH2355@suse.de> References: <1373536020-2799-1-git-send-email-mgorman@suse.de> <1373536020-2799-9-git-send-email-mgorman@suse.de> <20130711123038.GH25631@dyad.programming.kicks-ass.net> <20130711130322.GC2355@suse.de> <20130711131158.GJ25631@dyad.programming.kicks-ass.net> <20130711140914.GE2355@suse.de> <20130712101408.GV25631@dyad.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20130712101408.GV25631@dyad.programming.kicks-ass.net> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Jul 12, 2013 at 12:14:08PM +0200, Peter Zijlstra wrote: > On Thu, Jul 11, 2013 at 03:09:14PM +0100, Mel Gorman wrote: > > That might be necessary when the machine is overloaded. As a > > starting point the following should retry the migrate a number of times > > until success. The retry is checked on every fault but should not fire > > more than once every 100ms. > > Yeah, something like that might work. But getting a working imbalance bound is > important. The current very weak direct migration is the only thing that keeps > the system from massively skewing load. I've corrected a flaw in that proposed patch, pulled in the check for dst load vs src load, and am adding another patch that will fallback to less preferred nodes if task_numa_find_cpu fails to find a suitable CPU. -- Mel Gorman SUSE Labs