How should storage be provisioned if I plan to use the Hyper-V replication feature?
Hyper-V replication is a feature in which block-level changes to a virtual hard disk are asynchronously replicated to a secondary host server containing an offline, replica virtual machine. The Hyper-V replica feature was first introduced in Windows Server 2012, but when Microsoft created Windows Server 2012 R2, the company introduced the ability to create three-way replicas.
Many organizations have begun using three-way replicas as a way of placing a standby, duplicate virtual machine (VM) in their primary data center, and a second standby replica in a remote data center. Although remote replicas are fully supported, there are some important considerations that you should take into account with regard to the replication process.
One such consideration is the initial seeding process. When you create a replica, the full contents of any replicated hard disks have to be copied to the replica. This seeding process isn't usually a big deal if the primary and the replica VMs exist within a common data center, but it can be problematic for remote replicas -- especially if the virtual hard disk being replicated contains more than a few gigabytes of data (or if its contents change frequently).
In these types of situations, it is usually more effective to use removable media for the seeding process. Rather than writing the replicated data directly to the remote VM, the data is written to a removable hard disk. That removable hard disk is then manually connected to the off-site replica and the data is read into the VM.
Another consideration is the replication frequency. Windows Server 2012 R2 offers an adjustable replication frequency. The replication process can occur as frequently as every 30 seconds or as infrequently as every 15 minutes.
For remote replicas, it is usually best to use a 15-minute replication frequency. Hyper-V is resilient against failed replications, but if there are too many replication failures in a row, the replication process will stop until an administrator manually intervenes. Using a 15-minute frequency provides the best protection against a WAN link failure leading to a full-blown replication failure.
This was first published in April 2014