Virtualization is all about flexibility. This has wide applications, one of which is as a disaster recovery system.
I first learned about the power of virtualization to recover from disaster when I was in college and using a VMware virtualized desktop system. When my production machine croaked (I was a magazine designer at the time), I was able to quickly recover because my environment was run through a virtual machine. It was simple enough to install the virtualization software on another computer I had in my room and spin back up the virtualized desktop of the machine that no longer worked. I lost maybe half a day.
Now, leveraging virtualization for enterprise disaster recovery is a lot more complex than what I did long ago in college. But, thankfully, the tools that are now available for virtualization are also far more robust.
To virtualize production storage for disaster recovery on the enterprise level, a product must first attach natively to the production storage area network (SAN) as a client node. The connectivity approach used on the production system needs to exist on the hardware used to virtualize the storage array as a result.
The volumes to be replicated must also be seen. The easiest way to replicate is using block-level replication, but this isn’t possible when using storage virtualization as a proxy.
As tech commentator Keith Townsend recently noted in a blog post, one solution is to use the storage virtualization product as a production controller.
“You can provision all available physical storage to the virtualization platform and then provision the production storage from the virtual storage platform,” Townsend wrote. “This works when you want to create a single virtual SAN from a pool of disks on different arrays.”
But, however, in the disaster recovery scenario it is important to take an existing production setup and replicate it without it impinging on production.
The most popular solution to this part of the problem, according to Townsend, is to leverage a product that can mount the operating system volumes as read-only targets.
“This means your production SAN needs to support zoning the LUNs or volumes to multiple hosts,” Townsend wrote. “In this case, you'd better be sure that it's read-only and that your underlying application will not be adversely affected by the dual reads from production and replication. This is where native Windows and Linux approaches fall short for most environments.”
The real trick in using virtualization is overcoming the limitations of the underlying storage subsystem; any commercial storage virtualization product will deliver the performance needed if this subsystem can be overcome.
Products such as VMWare’s vSAN, FalconStor (News - Alert) and HP’s StoreVirtual can help with this.
While this setup is admittedly more complex than preparing a single desktop for a disaster, it is worth the effort. Like insurance, disaster recovery planning feels like a waste of energy right up until the moment it is needed. Then it becomes a lifesaver.
The flexibility of virtualization makes it a great tool for such disaster preparations, because it truly can allow computing systems to be easily moved. That’s the promise of virtualization, and it is put to good use in the case of disaster recovery.