By Richard Scannell, GlassHouseNext StepsVirtualization is big news in the IT industry. There is a fundamental cost savings and increased IT agility—I call it fluidity—possible from being able to aggregate hundreds or thousands of highly-distributed servers into relatively few, more easily-managed, centralized servers. And some forward-looking companies are looking into the benefits of virtualization for enhanced disaster recovery planning, more flexible desktop provisioning through virtual desktop interfaces, and more cost-effective computing for remote offices using "infrastructure in a box." But, like every other "big" IT technology, virtualization is following the classic hype cycle first defined by Gartner in the mid-90s:
With virtualization, the disillusionment usually involves not what the technology can do—for the most part, it's delivering on its promises—but how well it lives up to the business case laid out for it. You can't assume the business case is solid just because it has the word "virtualization" in it. I'd like to share some observations about three areas in particular where over-enthusiasm about virtualization can lead you astray:
If you make these the core of your due-diligence, you can develop a solid business case, avoid surprises, and build a robust and "future-proof" virtual infrastructure. A Long-Term InvestmentROI is the rock on which a lot of early virtualization projects have floundered. Companies have gone out, spent a lot of money on big iron and hypervisor software (the virtualization platform that allows multiple operating systems to coexist on one computer), but haven't been able to deliver the efficiencies they sought and the return they promised. In some cases it was simply poor math—they got carried away by technical enthusiasm and simply didn't write a very good business case. But in other situations they ran up against the reality that virtualization often requires a significant investment in new hardware and software. Many companies are running a lot of server boxes that are three to four years old, and aren't really up to the demands of virtualization. The good news is, these boxes are already "off the books" in terms of amortization, and they need to be replaced. The bad news is, you can't really build a good virtualized infrastructure in dribs and drabs; or rather, if you do, you'll run smack into structural IT problems. Piecemeal implementation is fine for a proof of concept, but to actually launch virtualization can be a forklift effort and cost serious money. We recently worked with a Fortune 100 client who found they'd have to spend US$5-$6 million on hardware and software just to get started. So the ROI is a long-term play, and requires some real selling on the part of the CIO. The Virtualization CliffBut even if you get the ROI part of the equation right, there's another challenge waiting: management. If you can't manage an infrastructure with 2000 servers when you can go out and physically touch each one of them, what makes you think you're going to be able to manage 2000 virtual servers any better when you can't see them? How do you handle demand planning, provisioning, metrics, accounting, all the behind-the-scenes details and processes that IT tends to have a reputation for not doing well? The fact is, too many IT departments are likely to follow the same path with virtualization that they did with SANs a few years back, and IP networks before that: deploying a technically-elegant solution and finding that without the right management tools, they've just sent themselves over a cliff at 60 mph instead of 30 mph. The size of this challenge is illustrated by the experience of Credit Suisse. The bank jumped into virtualization aggressively in 2005 and found it had to develop a tool to manage it. Afterwards, the effort involved convinced them that the tool was commercially viable, and in January of this year they spun off a company, DynamicOps, to sell the tool to enterprise companies struggling with the same problem. One of the areas where this challenge causes the most difficulty is in accounting, or chargeback. When you were dealing with hardware it was fairly easy to pass the cost—or the savings—along to your internal customers. But how do you charge for a virtual server, especially if part of the promise of virtualization is commodity server cycles, where someone may only need a virtual server for three weeks, or a year? I know of one company where IT promised, and delivered, a $10 million cost reduction through virtualization. But imagine how their business users reacted when they still ended paying the same price for a virtual server as for a physical one because IT had no model for individual chargeback—all the savings accrued to the corporate bottom line. Organizational TransformationManagement tools are available, although they're going to get better as competition heats up between virtualization vendors and the industry gains experience. But even with good management, there's one more challenge to consider, and in some ways it can be the biggest. You can't realize the benefits the fluidity of a virtual infrastructure gives you if you shoehorn it into the same old IT organizational structure. Unless IT is reconfigured as a service provider to the business as a whole, you're never going to see the full return from a virtualization initiative. Which brings up my final point, something that CIOs will have to face eventually as part of IT's ongoing evolution from utility to strategic partner; do you really need to be in the business of owning servers? Isn't it your data and the processes you've developed that matter to the business? Who cares where the servers are? All those telecommunication and co-location providers out there certainly aren't going to overlook this opportunity for long, and eventually server capacity will be just as much of a commodity as bandwidth. So while you're reorganizing your IT department as a server-cycle provider to the business as a whole, give some thought to the day when you'll outsource even that. Richard Scannell is co-founder and SVP, GlassHouse Technologies, a Framingham, MA-based consultancy specializing in consolidation, virtualization and data infrastructure management |
 |
Guest