Will data center history repeat itself?

Anyone dead set on consolidating data centers across government, as the government technology and budget experts in the Obama administration seem to be, might profit from talking with their counterparts in the Clinton administration, which undertook a similar task nearly 15 years ago with depressing results.

Despite some breakthroughs — most notably at the Defense Department — the Clinton team largely failed to arrest the growth in data centers, effectively setting the stage for exponential growth after the Bush administration took over and the federal technology budget boomed.

Consider this: When the Office of Management and Budget issued Bulletin 96-02 in October 1995, the number of data centers was believed to be about 200. The OMB directive ordered all federal agencies to close, consolidate, modernize and/or outsource their data centers to increase efficiency and reduce costs. "Industry experience suggests operational savings of between 30 percent and 50 percent from consolidation when compared with unconsolidated operations," then-Budget Director Alice Rivlin wrote.

Fifteen years later, the number of data centers stands at about 1,100, according to OMB.

Related stories:

Extreme makeover: Cutting data centers down to size

And despite all the advances in technology since then — not to mention several changes of the guard at federal agencies — the Obama administration likely faces the same source of resistance: human nature, or what military types call "protecting your rice bowl."

Inevitably, centralizing data center operations means centralizing power. Whatever the business benefits of consolidation for everyone involved, the power struggle was seen as a zero-sum game.

“The resistance was so great, though the intent was good, [that] very little happened,” said John Ortego, a government consultant who at the time was director of the Federal Systems Integration and Management Center at the General Services Administration.

OMB officials believed that the business case was compelling. Based on a fairly straightforward cost analysis of equipment and personnel costs, they came up with an optimal size for federal data centers, as measured by computing capacity. For example, in the case of IBM-based operations, a data center should have enough systems to handle 325 MIPS, or millions of instructions per second. Data centers running below that threshold were eating up too much of their budgets with overhead costs.

Agencies with small data centers were encouraged to consolidate internally, if they had enough overall capacity, or outsource their operations to another agency or commercial service provider.

OMB officials believed they could reduce the number of data centers from 200 to perhaps 50 within three years, saving several hundred million dollars in the process.

In an interview at the time, Bruce McConnell, then chief of the information technology branch at OMB’s Office of Information and Regulatory Affairs, said the strategy “is one that is hard to achieve, and it will result in substantial consolidation — we recognize that — but it is not unreasonable.”

But consolidation required accommodations that many agency officials simply were not willing to make. The big savings in consolidation would not come from co-locating equipment. To achieve true economies of scale, agencies needed to meld their computing infrastructures to a certain degree.

The prospect of making such changes runs into “some deep-rooted culture issues,” said Jim Flyzik, president of the Flyzik Group, who was chief information officer of the Treasury Department during the Clinton administration.

Treasury and a few other agencies made some progress, but the initiative never gained traction, at least among civilian agencies, Flyzik said. And with the change in administration in 2001, data center consolidation went bust and data center building boomed.

To be sure, a different kind of story played out at the Defense Department. Between 1990 and 2005, the Defense Information Systems Agency reduced the number of mainframe sites from 194 to 4, with 11 other centers providing related computing services.

The difference, of course, is that DOD has a command-and-control culture. The same culture is found in the commercial world, in which consolidation happens “because the chief executive officer said this is going to happen,” Ortego said.

That is not the environment in which OMB operates. In the case of data center consolidation, that is unfortunate, Ortego said. Given all the potential savings, consolidation “should be done,” he added, “but I would not anticipate it being done.”