The word around the IT watering hole is that server virtualization is yesterday’s news. VM saturation at most enterprises is about as high as comfort levels allow, so any future expansion of the technology and the platforms that provide it will come from new business infrastructure, not by virtualizing existing ones.
But despite the lackluster economy, the “new business” side of the coin is growing rapidly and is still ripe for dramatic expansion. In a nutshell, server virtualization is still a vital component of the data ecosystem because that thing we call “the cloud” could not exist without it.
For cloud providers, the higher the virtual density, the greater the profit potential because each one of those VMs represents additional workload, and therefore revenue. In fact, the challenge for most cloud providers isn’t just to boost their virtual footprint, but to manage and deliver that virtual architecture in ways that are most useful to clients. The best way to accomplish this task is by improving the migration and integration process using the same virtual platforms that exist in the data center.
A case in point is Rackspace. The company recently launched a new vCenter management service, Dedicated VMware vCenter Server, which oversees single-tenant VMware environments. The system uses the same vCenter APIs that enterprises have deployed in their own virtual architecture, offering easier integration with the provider’s bare-metal servers that support both VMware virtualization and the OpenStack cloud protocol. The idea is to get to a point at which the Rackspace infrastructure can be viewed as an extension of the client data center, which would simplify overall data operations and make it easier for the enterprise to shift loads onto the Rackspace cloud (see above comment regarding profit motivation).
Of course, how much better would this be if the entire process was automated? RiverMeadow Software will be at VMworld next week showing its RiverMeadow Cloud Migration SaaS platform that handles the mundane task of migrating workloads between vCloud environments. The system acts without agents and does not interfere with normal server operations, but simply collects selected images and related metadata and ports them to a new target without shutting down and rebooting the source. Automating the migration process, by the way, helps the client as much as the cloud provider because it can be used to manage the volumes, and thus the costs, going to the cloud.
As I mentioned in a previous post, however, different virtual platforms behave differently in the cloud—or more to the point, each platform vendor views the cloud in a different way. VMware is working to set up a broad cloud universe made up of numerous providers, system vendors and platforms all running VMware virtualization. Microsoft strives for the same thing for Hyper-V, but with the fundamental difference that cloud interoperability would flow from the operating system, rather than from the virtual layer. For Microsoft, then, the challenge is to convince the enterprise that its approach offers improved functionality, greater efficiency or some other cost-benefit advantage that other, more open platforms can’t deliver.
Part of this strategy seems to be appealing to the very immediate needs that enterprises have for the cloud, namely backup and recovery. Microsoft has unveiled the Hyper-V Recovery Manager for the Azure cloud, an integrated DR solution that features asynchronous VM replication for Windows and Hyper-V Server 2012 environments. Under the platform, enterprises can centrally configure and manage a cloud-based DR infrastructure—even on non-Azure clouds as long as they support System Center 2012—while storing the recovery plan in Azure to ensure continuity should primary systems go down.
What we have here is a classic battle for the soul of enterprise computing. In a way, it’s a lot like the Microsoft vs. Apple tussle of the old days, except the new Apple is Microsoft, with its more integrated approach vs. VMware’s broader, vendor ecosystem strategy.
One thing is certain, though: None of this would be happening if plain, old server virtualization didn’t matter anymore.