The number 1 most requested feature for Microsoft Azure is disaster recovery. The keynote speech mentioned some details. Also breakout session DCIM-B322 revealed some details.
At TechEd 2014 in Houston Microsoft announced a new service running on Microsoft Azure. Azure Site Recovery enables failover of Hyper-V virtual machines running on-premise using Microsoft Azure as a target.
Azure Site Recovery will be the new name for the current Azure service named ‘Hyper-V Recovery Manager’. Recovery Manager is able to orchestrate a failover and failback of System Center Virtual Machine Manager managed virtual machines. Required however is having two sites managed by the customer. Azure cannot be used to run virtual machines in a failover scenario. Azure just has meta data and runs the orchestration logic.
Azure Site Recovery will go in preview in June 2014. It will allow data replication set at a schedule of every 30 seconds, 5 or 15 minutes. It has a maximum of 15 recovery points and supports VSS snapshots for application consistency. Customers can select to encrypt data at rest.
Customers will only be charged for the consumption of storage. During normal replication no virtual machines will be running. Only during failover or failover test compute resources will be charged.
Azure does not provide tailor made virtual machine sizes. This means customers has to select from a number of fixed configured virtual machines (small, large, A6, A8 etc). Azure Site Recovery will automatically adjust the Azure VM to have a best match with the failover VM which run on-premises.
Virtual disk format VHDX is supported but Generation 2 virtual machines are not supported.
Like for Hyper-v Recovery Manager, Virtual Machine Manager is required for Site Recovery
The service allows to map on-premise networks to Azure networks. It also has recovery plans. This enables administrators to control the sequence in which virtual machines are started in case of a failover to Microsoft Azure.
A failover can be tested while the production site running on-premises is operational and not disturbed.
The Azure management console which is used to manage Site Recovery supports any HTML5 compliant browser. This means failover can be performed from any device and from any location.
Site Recovery does not support seeding. Seeding is shipping a complete virtual disk on removeable media or a NAS to a datacenter to prevent tranfering data of an initial replication over the WAN. Which is remarkable because Azure does have such an import feature for virtual machines. I guess the challenge with Site Recovery is that no virtual machine is created. Only at a failover (test) a virtual machine is provisioned using replicated data. ExpressRoute (a private WAN connection between an on-premise datacenter and Microsoft Azure datacenter) can be used for replication traffic.
Site Recovery supports a failback in which virtual machines are replicated back to the on-premise datacenter. This is done by only sending the delta or change data. So only the data which has been changed during the failover is sent back to the original VM.
The session on Site Recovery is available for viewing here.
Site Recovery will be available in 6 different regions at Preview. Two in the US, two in the UK (this is what one of Microsoft staff told in an interview) and two in Asia. The UK is Dublin and Amsterdam by the way, both not located in the UK.
Microsoft did not publish the pricing of this new service.
Channel 9 did a live interview by Symon Perriman with Vishal Mehrotra and Manoj K Jain titled ‘Enterprise-Scale Disaster Recovery’ It was at May 14 from 11:30AM to 12:15PM and was recorded. It is now shown on https://channel9.msdn.com/ but later I will add the direct URL of the recording
Some nice new information was revealed.
Ofcourse this new feature will be described in my to be released book on Microsoft Cloud OS.