Migrating software to a new data center could take a big bite out of consolidation's anticipated payback.
There is no doubt that the federal government can reduce many of its IT-related costs by decreasing the number of data centers it operates and drastically increasing the efficiency of those that remain. The benefits of using fewer, centralized servers — and thus less energy to power and cool the equipment — are well understood. Those are the selling points of Federal CIO Vivek Kundra’s governmentwide data-center consolidation plan, which he says can save more than $3 billion.
The business case gets a bit murkier when it comes to the costs associated with making consolidation happen. It’s not that the transition costs are impossible to nail down, but it’s far from certain that every government agency has a good handle on everything that consolidation entails.
The area with the biggest potential for hidden costs is the software that provides the business functionality that agencies depend on — everything from e-mail and case management systems to procurement and financial applications. How much would it cost, for instance, to rework potentially hundreds of software applications so that they can continue to operate on a new platform in a new location?
Some observers believe the migration costs could be high enough to be a deal-breaker. One Federal Computer Week reader, commenting online about an article on Kundra’s IT management reform plans, wrote: “Hardware has always been chump change when it comes to overall IT dollars. What goes in the box is the real cost. The reprogramming alone during these consolidations eat up much more [money] than any projected savings.”
People can debate whether the recoding expenses would really be that high, but most will acknowledge that software is a serious cost-side variable that it would be foolish to take lightly.
Part of the uncertainty stems from the fact that many IT executives don’t know exactly how much they can squeeze their hardware during consolidation and still provide a satisfactory software experience for end users, said Michael Daconta, chief technology officer at Accelerated Information Management.
Agencies are sitting on excess hardware capacity they thought they needed based on their end users’ original performance requirements. If they are lucky, CIOs can consolidate by using new virtualization technology to run most of their applications on far fewer servers, in effect calling users’ bluff.
But if it turns out that some users’ applications do in fact need that reserve hardware capacity to absorb usage spikes and if agencies don’t want to own that spare hardware, they will have to buy that capacity on a utility basis from the cloud, Daconta said. Then you are talking about potentially rewriting applications to work in the cloud, and that means big upfront expenses.
“It’s not clean math,” Daconta said. “Which option you take and how much real versus imagined savings you get depends on understanding your utilization and current efficiency and what can move and not move.”
There are other ways to mitigate software migration costs. The Agriculture Department is in the middle of a consolidation effort that will reduce 43 data centers to seven. In the project’s early stages, officials recognized that older client/server applications that depended on fast local-area network speeds would suffer if they had to traverse a wide-area network to reach a server relocated to a distant data center.
Rather than possibly incurring substantial upfront costs to rewrite those apps for a Web-based environment, USDA officials deployed a more economical Citrix Systems terminal emulation solution that allows those apps to function across a WAN until normal refresh schedules call for an application upgrade, said Jim Steven, USDA’s data center consolidation program manager.
“We knew it was a nonstarter to come in and say, ‘You need to spend $200,000 reprogramming your apps,’ because the money is just not there,” Steven said.
Nevertheless, as Steven and others point out, consolidation will no doubt require agencies to spend some money to save more money. Determining what those amounts will be is the hard part.