14 ways CIOs under estimate the costs of cloud computing
By David Carr
For CIOs who distrust most technological promises (having heard too many of them), cloud computing sets off alarm bells. Yet those CIOs finding success in the cloud say their colleagues should be equally skeptical of IT managers who claim they can deliver better and cheaper results internally. (For expert advice about cloud-vendor contracts, see “How the Cloud Can Turn Toxic.”)
DeVry University CIO Eric Dirst is careful to keep his team honest about the real costs of internal IT. “You have to think it through. We try to calculate 10-year [total cost of ownership] when we make a big purchase. We don’t want to have to replace things in two or three years,” he says.
An honest comparison must take into account the cost of provisioning servers and replacing them about every three years, plus the overhead of system administration, security patches and disaster recovery. Those are built into the monthly fee for a cloud computing service, Dirst says. So far, DeVry uses software as a service (SaaS) applications for CRM, HR and email. It has also has deployed custom applications on Saleforce.com’s Force.com and custom utilities on Amazon Web Services (AWS).
It’s difficult to make direct comparisons between cloud computing and internal IT, notes Larry Bolick, CIO of the creative services staffing firm Aquent. “It’s not like when you upgrade a server, and the clock speed is twice as fast or you’re getting four times the memory. That kind of comparison doesn’t exist necessarily in the cloud.”
Over the past two years, Aquent has converted more than 30 offices in North America to a cloud-based phone system and moved its core business system to the cloud. In the process of changing the company’s homegrown ERP system to a Web-based model, Aquent switched it from traditional data center colocation to hosting on AWS. The software now runs in three AWS clouds—one in the United States, one in Ireland and one in Singapore—to deliver global coverage. “With that particular application, we get some disaster-recovery capability in the bargain because we’re replicating between the three different instances,” Bolick says.
James Staten, a Forrester Research analyst, agrees that companies tend to underestimate the cost of internal IT. “Often, organizations only count the capital expenditure, not the operational costs, so they come to the incorrect conclusion that it would be cheaper to do it on premise,” Staten asserts.
The Capital Expenditure Monster
Today, cloud computing is most popular with small-to-midsize businesses (SMBs). Large enterprises tend to be more conservative, often citing security concerns. That will change as the servers that those large enterprises own come to the end of their useful life and CIOs look for money to replace them, predicts Michael Hugos, a consultant, former CIO and current CIO.com blogger.
At that point, says Hugos, “I think it becomes a CFO discussion about capital expenditure.” CFOs may increasingly force the issue of replacing investments in IT equipment with the variable costs of cloud computing. In addition to the direct costs, a CFO is likely to consider the opportunity cost of tying up capital in fixed assets such as servers, he says.
And precisely because customers are crying out for reassurances about the security of cloud services, service providers will invest in making them secure—at the same time many enterprises are cutting back on information-security investments. “The idea that data sitting on a server in my server room is more secure [than in the cloud] is disingenuous,” says Hugos. Underlying the argument, he says, are IT professionals trying “to convince the suits not to outsource my job.”
Because many technology managers have a vested interest in fending off this wave of change, CIOs need to turn a gimlet eye on their staffs (and their own) cost projections, Hugos says. “I would give my estimate to a bunch of hard-nosed accounting and finance guys and see what they say about it.”
Joe Weinman, a Hewlett-Packard executive who is one of HP’s chief spokesmen on cloud computing, predicts a more gradual transition. “I do believe most IT will move into the cloud for consumers and SMBs,” he says. On the other hand, large enterprises have enough systems-management discipline “that the idea that a cloud provider would achieve better, meaningful economies of scale is difficult for me to believe.”
For large enterprises, using a public cloud infrastructure will make sense in special cases, such as when they could profit from renting a large number of computers for a short period to run an intensive calculation quickly, Weinman suggests.
A big company also might use cloud services as a hedge against uncertainty in case a fast-growing new product or business unit stalls, Weinman says. “While you’re growing, it’s easy to say you should just buy servers and keep deploying them. But if you do enter into a period of decline, those assets are like concrete shoes.”
Costs That Lurk in the Cloud
Dirst and Bolick agree that cloud computing also can have hidden or unanticipated costs. Bolick says he underestimated the cloud server capacity he would need to achieve an acceptable level of performance because he made the mistake of treating cloud servers as the equivalent of physical servers.
Because a cloud server actually represents a virtualized slice of memory and processor resources from a pool of servers, its performance may not match what you would expect from a dedicated server in your data center or a colocation facility.
“If I was doing it again, I would essentially overengineer the environment in the cloud. Then, once I had the system stabilized, I would downsize it from there,” Bolick says. “Of course, one of the beauties of the cloud is we were able to ramp up quickly,” correcting the performance problem in a few days.
And when transitioning to the cloud, your cost model should allow for a few months of running the new and old systems in parallel, which will lead to a spike in expenses.
Organizations that have adopted principles of service-oriented architecture will be in the best position to take advantage of cloud technologies, Dirst says, while those that have not may need to play catch-up. That might mean training staff, upgrading middleware, or both.
Dirst admits he underestimated the management and monitoring costs associated with cloud apps. Because traditional enterprise-network-management tools don’t do a good job of tracking cloud apps, his department has had to script its own monitoring routines so it knows when a cloud app—or the integration between an app and enterprise systems—stops working.
“That usually costs more than we had expected,” Dirst says. Still, he sees good ROI, combined with the ability to deliver new capabilities faster in the cloud environment. That’s particularly true of SaaS apps, which typically deliver new features instantly two or three times a year. When a traditional software vendor releases a new version, it’s usually three to six months before DeVry can roll it out with confidence.
In the overall cost equation, “Not having to go through that? How does that commercial go? Priceless,” says Dirst.
David F. Carr is a freelance writer based in Florida.