From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bastian de Groot Subject: large overhead with blktap2/vhd Date: Wed, 23 Jun 2010 17:26:05 +0200 Message-ID: <201006231726.06620.degroot@univention.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org Hi everybody, I am searching for a copy-on-write solution for Xen and tested blktap2 with the vhd backend. My Test: I've created a master-image with "dd" and installed WindowsXP on the image. Then I used vhd-util to create a copy-on-write-image for the master-image. After booting the VM and logging in the, the cow-image had a size of 260 MB. That's a little bit weird, because I didn't really changend any data. Afterwards I downloaded eclipse in the VM, to see how big the overhead is. The eclipse zip-file is 190.64 MB big. But the cow-image had a size of 1468.29 MB after downloading eclipse. That's an overhead of 670 %. Then I unpacked eclipse. So that the actual change in the VM had 425.37 MB. At that moment the cow-image was 1979.79 MB big. The result is 365 % Overhead, that's better then the result before, but still not very good. My questions are: Have I done anything wrong at measuring the overhead? Is there a bug in blktap2/vhd? Or is this a usual result? Here the configuration: The host is a Xen 4.0 (I compiled it myself) on an 2.6.32-xen-amd64 debian kernel (also self compiled to include some mudules statically). The guest is a WindowsXP 32 bit SP2. I would appreciate your help on that very much. Bastian