I just wrote a story about cloud computing, which I opened by citing Nicholas Carr's well-known analogy that it makes about as much sense for most companies to operate a dedicated data center as it would for them to run a dedicated power utility -- that is to say, no sense at all. I contrasted Carr's opinion with that of McKinsey & Co. analyst William Forrest, who claims that while the cloud is a logical choice for small companies, cost savings won't hold up for large companies that have already made considerable infrastructure investments.
Forrest's analysis compares the costs incurred by an unnamed McKinsey & Co. client to purchase and manage the processing cores that run the company's Linux and Windows applications to the price of running the same cores via Amazon's Elastic Cloud infrastructure service. His conclusion: The company will spend nearly 150 percent more in the cloud computing model.
Plenty of people had problems with Forrest's numbers, including Bernard Golden, CEO of consulting company HyperStratus, author of "Virtualization for Dummies" and one of several folks I interviewed for my story. In an opinion piece for CIO.com, Golden lists four problems he has with Forrest's report:
- The example isn't representative of all or even most internal computing environments. Yet some people (perhaps CIOs eager to affirm their skepticism of the cloud) may use it to jump to inaccurate conclusions, implies Golden.
- The conclusion that the company would reduce headcount by only about 170 with a switch from internal to external servers doesn't seem accurate.
- Forrest doesn't factor in the cost of ongoing maintenance of internal servers or associated assets like networking gear and storage arrays
- Forrest focuses too narrowly on server utilization rates rather than cost per unit of computing capacity.
I agree with Golden's points, especially No. 3. However, I'm someone who can generally see all sides of any given issue, making me either open-minded or wishy-washy, depending on how you look at it. So I can understand Forrest's recommendation that large companies with sunk infrastructure investments may save more money now by focusing on internal server virtualization and consolidation while the cloud computing model continues to coalesce, as it most certainly will.
Carr alludes to this in a post from his Rough Type blog, writing:
Like any utility system, cloud computing becomes more attractive as cloud providers gain scale and experience and are able to push down their prices while improving their services. As each year passes, the economic advantages of an expanding utility system become real for another set of companies. But that process takes time. Don't expect to see the biggest companies closing down their data centers in the next few years. Besides, the cloud in the end will be more interesting for the new models of computing it opens up rather than for its ability to accommodate the old ones.
There's a tendency to see the cloud as an either/or proposition, which it simply isn't for most companies, at least not yet. The sources I interviewed for my story agreed that most companies will start by migrating commodity-type applications like e-mail (costly to manage, offering no competitive advantage) to the cloud first. Some may then opt to take more strategic apps to the cloud, although.security and control questions will dictate how far up the "stack" those apps will reside. And certain applications, such as those demanding specialized hardware or very specific configurations, may never be a good candidate for the cloud.