From mboxrd@z Thu Jan 1 00:00:00 1970 From: raz ben yehuda Subject: Re: Subject: PATCH : offline scheduler ( some refer to it as a CPU isolation ) Date: Sun, 19 Apr 2009 13:09:33 +0300 Message-ID: <1240135773.3395.18.camel@raz> References: <1240081024.3534.3.camel@raz> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7BIT Cc: raz ben yehuda , linux-rt-users@vger.kernel.org To: Leon Woestenberg Return-path: Received: from mxout12.netvision.net.il ([194.90.6.37]:40013 "EHLO mxout12.netvision.net.il" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752057AbZDSHJf (ORCPT ); Sun, 19 Apr 2009 03:09:35 -0400 Received: from [192.168.2.102] ([217.132.154.72]) by mxout12.netvision.net.il (Sun Java System Messaging Server 6.2-8.04 (built Feb 28 2007)) with ESMTPA id <0KIC0031A6KFIJB0@mxout12.netvision.net.il> for linux-rt-users@vger.kernel.org; Sun, 19 Apr 2009 10:09:51 +0300 (IDT) In-reply-to: Sender: linux-rt-users-owner@vger.kernel.org List-ID: On Sun, 2009-04-19 at 03:23 +0200, Leon Woestenberg wrote: > Raz, > > On Sat, Apr 18, 2009 at 8:57 PM, raz ben yehuda wrote: > > offsched is a platform aimed to assign a service to an off-lined processor. > > Motivation is explained in: > > http://sos-linux.svn.sourceforge.net/viewvc/sos-linux/offsched/trunk/Documentation/OFFSCHED.pdf > > I quickly glanced over but could not find a comparison with other methods? > > > Suppose I run /bin/init with a certain CPU affinity (not covering all > CPUs) and then run my own specific service on the remaining CPUs, how > would this differ from your approach? First , not all services are children of init. there are workqueues, kernel daemons, softirqs and so on. offlet provides hard real time,in the sense that one can pre-determine how long an operation cost, it is NMI mode after all. Also, this is why i chose to implement a 1us timer, because i can do it. > >From what I quickly gathered, you are running parts of the kernel > space on the second CPU set, or is that too simple a statement? I am running **no kernel** on the offloaded processors, but offsched code, and the rest of the processors run the "kernel". > > Looks like a very fresh approach, nice project! Thank you. i appreciate it. > Regards,