From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:59906) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UBRIT-0005g2-OE for qemu-devel@nongnu.org; Fri, 01 Mar 2013 09:50:56 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UBRIP-0006Us-AN for qemu-devel@nongnu.org; Fri, 01 Mar 2013 09:50:53 -0500 Received: from ssl.dlhnet.de ([91.198.192.8]:56207 helo=ssl.dlh.net) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UBRIP-0006UU-4b for qemu-devel@nongnu.org; Fri, 01 Mar 2013 09:50:49 -0500 Message-ID: <5130C04F.9090803@dlhnet.de> Date: Fri, 01 Mar 2013 15:50:55 +0100 From: Peter Lieven MIME-Version: 1.0 References: <5130ADC0.9070402@dlhnet.de> <5130B29C.3060301@redhat.com> <5130B5D8.7030308@dlhnet.de> <5130B677.1060604@redhat.com> <5130B789.7050709@dlhnet.de> <5130B9D7.2050108@redhat.com> In-Reply-To: <5130B9D7.2050108@redhat.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] migration: use XBZRLE only after bulk stage List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Eric Blake Cc: Orit Wasserman , "qemu-devel@nongnu.org" , Paolo Bonzini On 01.03.2013 15:23, Eric Blake wrote: > On 03/01/2013 07:13 AM, Peter Lieven wrote: >> On 01.03.2013 15:08, Eric Blake wrote: >>> On 03/01/2013 07:06 AM, Peter Lieven wrote: >>>>> Do you have any benchmark numbers? At any rate, the explanation seems >>>>> sound, so a benchmark should show this. >>>> >>>> Do you have a particular test pattern in mind? If there is nothing >>>> going on >>>> in the VM XBZRLE will not be better than normal copy at all. >>>> >>>> Otherwise you will have N xbzrle misses and 0 xbzrle pages without the >>>> patch >>>> and 0 xbzrle misses and 0 xbzrle pages with the patch. >>> >>> How about a migration of a guest running the synthetic r/w load >>> generator in docs/xbzrle.txt? >>> >> Good idea. I will leave max downtime and bandwidth at default values. >> >> Would you be happy with 1GB vRAM and 256MB page cache? > > Sure - just any run that you can do that shows before and after numbers, > and that is described well enough to be a reproducible test. Final > statistics on the migration (pages transferred, cache hits and misses, > etc) and time spent on the migration will hopefully show an improvement, > but most important is that they do not show a regression. > just a quick test on my desktop: ~/git/qemu$ x86_64-softmmu/qemu-system-x86_64 -k de -enable-kvm -m 1024 -drive if=virtio,file=iscsi://172.21.200.31/iqn.2001-05.com.equallogic:0-8a0906-16470e107-713001aa6de511e0-001-test/0 -vnc :1 -boot dc -monitor stdio using ubuntu 12.04.1 desktop and the example from docs/xbzrle.txt a) with the patch (qemu) info migrate capabilities: xbzrle: on Migration status: completed total time: 22185 milliseconds downtime: 29 milliseconds transferred ram: 706034 kbytes remaining ram: 0 kbytes total ram: 1057216 kbytes duplicate: 108556 pages normal: 175146 pages normal bytes: 700584 kbytes cache size: 67108864 bytes xbzrle transferred: 3127 kbytes xbzrle pages: 117811 pages xbzrle cache miss: 18750 xbzrle overflow : 0 b) without the patch (qemu) info migrate capabilities: xbzrle: on Migration status: completed total time: 22410 milliseconds downtime: 21 milliseconds transferred ram: 721318 kbytes remaining ram: 0 kbytes total ram: 1057216 kbytes duplicate: 105553 pages normal: 179589 pages normal bytes: 718356 kbytes cache size: 67108864 bytes xbzrle transferred: 630 kbytes xbzrle pages: 21527 pages xbzrle cache miss: 179589 xbzrle overflow : 0