Hardware consolidation is going well with your new virtualization stack, so it should be a no-brainer to leverage that technology for more advanced purposes, like resource sharing, right?
Well, not so fast. You see, virtualization is a key component for the futuristic pooling of resources that will drive things like cloud computing and global load balancing. But for most of you, there is still quite a bit of work to be done.
First of all, as Cath Jennings points out in her most recent Computerweekly piece, that kind of functionality will require a substantial investment in networking technology. Virtualization is great for creating new operating environments within a single piece of hardware, but if you want to share it across the network, you'll need a coordinated strategy that links available servers and storage with available I/O. That means a possible revamping of everything from the server backplane to the storage controller.
And that's not even half the story. There's the application architecture (s) to consider. The fact is that there are some applications and workloads that lend themselves to virtualization and some that don't. The running line from most vendors is that they expect to see 100 or more VMs on a single server before too long, but as Boulder, Colo.'s Enterprise Management Associates found out in a recent study, the ceiling for most organizations is about 15 when they start loading up the more mission-critical or data-intensive applications like ERP and CRM.
This could be part of the reason Forrester estimates that less than 2 percent of enterprises have deployed working internal cloud architectures, even as the definition of such structures is very much in the air. The group's James Staten told Computerworld that while the desire is there, many organizations simply do not have the necessary experience with virtualization, automation or half-a-dozen other recent developments necessary to build and maintain such a futuristic environment.
But that's not to say that the experience level isn't growing. Indeed, it probably won't be long before a network platform or device that does not include some kind of pooling technology will be hard to find. Take, for example, the newest version of IBM's SAN Volume Controller, which now extends support to STEC SSDs. On the surface, it's a simple performance play, with the technology said to achieve upwards of 1 million IOPS. Once in place, however, it becomes clear that with SSDs in the mix and support for multi-vendor SANs, pooled storage is one of the main selling points.
The fact that shared-resource architectures are so complex makes them a daunting prospect. But at the same time, there's comfort in knowing that hardly anybody has taken serious steps in this direction, so it's not like you're behind the curve, yet.
And operationally, the pooling of resources is a very desirable tool, even if it's not as simple a matter as some people would have you believe.