From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:37976) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1SdzZ5-0003FT-MF for qemu-devel@nongnu.org; Mon, 11 Jun 2012 04:01:41 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1SdzYx-0006xK-Mm for qemu-devel@nongnu.org; Mon, 11 Jun 2012 04:01:31 -0400 Received: from mx1.redhat.com ([209.132.183.28]:36797) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1SdzYx-0006x4-Eo for qemu-devel@nongnu.org; Mon, 11 Jun 2012 04:01:23 -0400 Message-ID: <4FD5A5CB.8020609@redhat.com> Date: Mon, 11 Jun 2012 10:01:15 +0200 From: Paolo Bonzini MIME-Version: 1.0 References: <4FD0B759.8030002@redhat.com> <4FD1FBF1.10305@redhat.com> <4FD20D11.1080603@redhat.com> <4FD22422.1080301@redhat.com> <4FD23A73.8080502@redhat.com> <4FD23D0D.6040602@redhat.com> <4FD4C6FE.8010501@redhat.com> <4FD5A56E.1020701@redhat.com> In-Reply-To: <4FD5A56E.1020701@redhat.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] CoW image commit+shrink(= make_empty) support List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Kevin Wolf Cc: Stefan Hajnoczi , jcody@redhat.com, Zhi Hui Li , Taisuke Yamada , qemu-devel@nongnu.org Il 11/06/2012 09:59, Kevin Wolf ha scritto: >>>>> >>>> I agree, doing it like mirroring for new writes on the top layer makes >>>>> >>>> sense, as long as you are willing to violate the (optional) speed >>>>> >>>> parameter. I wouldn't think violating the speed parameter in that case >>>>> >>>> would be an issue, as long as it is a documented affect of performing a >>>>> >>>> live commit on the active (top) layer. >>> >> The question is what the speed parameter really means for a live commit >>> >> block job (or for an active mirror). I think it would make sense to >>> >> limit only the speed of I/O that doesn't come from the guest but is from >>> >> the background copying. >> > >> > Not necessarily, you can throttle the copying speed. If the I/O from >> > the guest has bursts, you will pick up the work later when the rate >> > calms down. > Then it's not an active mirror any more, but a passive one. Yes, I guess what I was saying is that the passive mirror code can be used also for live commit of the top image. Paolo