I saw this and thought I would share it.  Given the rumours around what can and cannot be done with Virtual Server and the technologies coming with Longhorn Windows Server 2008.  I did not realise that we offered "Quick" migrations - a few seconds downtime - when moving from machine to machine.

With recent announcements there has been some confusion about what Windows Server Virtualisation, can do:

Virtual Server 2005 allows VMs to be moved between hosts and has a scripting API to enable this. VMs can move to Windows Server virtualisation, and you will move virtual machines between WSv servers. Our research shows scheduled maintenance - outside the working day - accounts for the overwhelming majority of moves (moving workloads for load balancing is very infrequent); workloads are moved to keep them running while their normal server is taken down to service the hardware or patch the parent/host operating system - something which will be required less with the new server core option.

Live Migration should be invisible to users. It copies memory contents from the source host to the target in the background before taking it down. Only the last few bytes are copied with the source server down so the switch happens in less than a second. Some competing products can do this today, but it has been postponed to a later release of Windows Server virtualisation. We will still offer Quick Migration functionality. 'Quick' varies between a couple of seconds and a couple of minutes depending the speed of memory and disk - because with quick migration we save the state of the virtual machine, reattach the VHD and saved state to a new host and resume. Depending on the situation users may not be aware anything has happened, or they may see a brief loss of service.

Both types use Windows Clustering to move the storage connectivity between hosts - we don't copy the VHD (that would be "'slow migration'). And this gives us a high availability failover solution for unplanned downtime - the virtual machines are restarted on other cluster nodes after a failure - and users may see a brief loss of service while this happens. Obviously live migration doesn't help unplanned downtime, because when the decision is made to move there isn't a live server any more! Other vendors support restarting a VM on another node, although they may charge for the feature: it's part of Windows Server 2008 Enterprise and DataCenter editions.

There's more information about Virtualisation on my blog.

James O'Neill

ttfn

David