How does shared-nothing live migration change things for data centers running Hyper-V?
Prior to Hyper-V 3.0, live migration was possible only within a clustered Hyper-V deployment. Furthermore, the cluster had to be built around a shared storage architecture and make use of Cluster Shared Volumes (CSV). This shared storage was connected to each host in the cluster by either an iSCSI or a Fibre Channel.
The shared storage requirement put Hyper-V clustering financially out of reach for many smaller shops. Shared storage tends to be expensive, and it has to be designed in a way that delivers adequate performance for all of the virtual machines. Sometimes this can be a tall order.
Shared-nothing live migration eliminates the requirement for CSV. Organizations that want a clustered Hyper-V deployment can build the cluster in a way that allows virtual machines to reside on local, direct-attached storage within each cluster node. This type of clustering allows virtual machines to be live migrated between clustered hosts without the hosts having to share storage -- hence the name shared-nothing live migration.
Organizations that have already invested heavily in storage need not fret. Even though Windows Server 2012 and Hyper-V 3.0 make it possible to live migrate virtual machines without CSV, the use of shared storage is still supported. In fact, Microsoft recommends that you use shared storage (as opposed to local, direct-attached storage) whenever possible.
One of the nice things about the way that Microsoft has designed the Windows Server 2012 failover clustering architecture is that the same basic clustering components are used regardless of whether or not CSV is being used. This means that an organization could build a Hyper-V cluster with no shared storage and then later add shared storage to the cluster if the need arises.
This was first published in October 2012