With the coming changes in Windows Server 2012 (WS2012), I thought I’d take a few minutes to explain the difference between HA and Live Migration.
High Availability (HA)
HA is when we make a virtual machine highly available by placing it on a cluster of hosts, e.g. a group of Hyper-V hosts in a Windows Server Failover Cluster, or a group of ESXi hosts in a vSphere cluster. If a host has a failure VM can be failed over to a different host. This means that an automated system provides better levels of availability than you get with non-clustered hosts. Typically, the failed host has crashed or powered down, and the VM is booted up on a different host.
Live Migration (aka VMotion)
Live Migration is a system where a running virtual machine can be moved from host A to host B with zero downtime. It is different from HA because it is more proactive than a failover operation. For example, Live Migration will be used by an administrator to move a VM from one host, that is going to have some maintenance operation, to another host with spare capacity.
Windows Server 2012
I have brought up this topic because in Hyper-V, Live Migration and HA went hand in hand, because they both required a Hyper-V cluster. That’s changing.
In Windows Server 2012 the tie between Live Migration and Failover Clustering has been broken. All you will need to do Live Migration is a 1 Gbps network, meaning you can live migrate a VM between two non-clustered hosts, between clusters, and of course, between clustered hosts.
As you now know, Live Migration is not high availability. I’m sure the tech media is going to confuse things – in fact, I believe they already are If you want your virtualised workloads to be independent of hardware/host fault then you must still install a Hyper-V cluster … and don’t forget that Windows guests need appropriate clustering (you can move a MSFT license from one hardware to another once every 90 days so you really should license hosts). If you work in a small/medium business, then don’t fret – Windows Server 2012 doesn’t require that expensive SAN any more – have a look at my previous posts on Storage Pools where you can now store HA VMs on active/active clustered file shares using cheaper JBOD.
Don’t get me wrong; Live Migration without clustering is an excellent addition. Many choose not to do HA because they cost of downtime is less than the cost of HA for some applications. They will like the agility and mobility they’ll now get with zero downtime.