2013
06.25

Hyper-V Replica was added in WS2012 Hyper-V to give us built-in, no special licensing required, asynchronous, per-VM/virtual hard disk, replication from one host/cluster to another, with compression and historical copy retention.

Note: That boils down about 1700 words of explanation into 1 sentence Smile I should have just left it at that in TechEd Speaker Idol!

There were always a few questions when I explained HVR to people:

  • Can I change the replication interval from every 5 minutes?
  • Can I replicate from Site A to Site B, and then on to Site C?

The answers were no.  That changes in WS2012 R2 Hyper-V.

Finer Grained Control of Replication

The default interval for replication of the Hyper-V Replica logs (the change log of each replicated virtual ahrd disk) is every 5 minutes.  You can change this to every 30 seconds, assuming you have the bandwidth/latency to push through the changes in this short window.  You can also increase it to every 15 minutes if you need more time to push through spikes in activity over latent/lower bandwidth links.

Before you ask: the options are restricted to 30 seconds, 5 minutes, and 15 minutes.

Extended Replication

You now can, if you want to, do the following:

  • Replicate VM1 from Site A to Site B
  • Then replicate the offline replica of VM1 from Site B to Site C

You cannot replicate as follows:

  • Replicate VM1 from Site A to Site B
  • Replicate VM1 from Site A to Site C

This is purely an A->B-> scenario.

Combining Extended Replication With Finer Grained Control

Here’s an example that might be useful:

  • A company replicates from their computer room in Office A to another computer room across a campus network in Office B.  This is all contained in Site 1.  There is a nice campus network so replication is performed every 30 seconds.  The RPO of a local BCP invocation (a fire damages Office A) is maybe 30 seconds.
  • The company replicates from Office B in Site 1 to Site 2 over a WAN link.  This is a slower link so replication is performed every 15 minutes.  The RPO of a long-distance BCP invocation (a disaster hits the city where Site 1 is located) is maybe under 16 minutes.

One variation on this scenario is that Site 2 is a hosting company that is selling a virtual DR site service (DR-as-a-Service aka DRaaS).

BTW, it would be pretty pointless to replicate from Site B to Site C more frequently (every 30 seconds) than from Site A to Site B (every 15 minutes).  That’s because there will only be changes to replicate to Site C every 15 minutes.

No Comment.

Add Your Comment

Get Adobe Flash player