From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dario Faggioli Subject: [PATCH 0 of 2 RFC] xl: move domeins among NUMA nodes Date: Fri, 06 Jul 2012 11:54:03 +0200 Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: xen-devel Cc: George Dunlap , "Zhang, Yang Z" , Ian Jackson , Ian Campbell , Andre Przywara List-Id: xen-devel@lists.xenproject.org Hello Everyone, FIRST OF ALL, this is an RFC and, more important, it implements a new feature (although it's a small one), so it is not meant for 4.2. Of course, comments are more than welcome, if you happen to find the time to look at it (after all, it is why I'm releasing it)... Just don't rush if you're busy with some pressing release-related stuff! :-P The above being stated, this series tries to tackle the problem of moving an already existent domain from a (some) NUMA node(s) to a (some) other(s). In fact, changing vcpu-affinity affects the node affinity of the domain itself, but this is only effective wrt the memory that will be allocated in the future, while all the existent pages are untouched. This of course makes a lot of sense, but there might be the need to _fully_ move a domain, both vcpus and memory wise, and that's what this series tries to achieve. Just let me add that the whole process looks a lot like a suspend/resume (or either to a local migration) to me, and that's why I implemented it like that for now. I went for the suspend resume approach, but I think I can make it more migration-oriented solution if wanted, as it'd reduce the downtime (although, in that case, we'd need to have twice the memory of the domain available during the process). Of course, I can also add the proper logic and interface to the hypervisor, but I'm not sure it buys that much, as there will still be something very similar to suspension or migration, just called and used a little bit differently (but of course I'm open to change my mind :-)). Thanks and Regards, Dario -- <> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://retis.sssup.it/people/faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)