In 2010, the Office of Management and Budget (OMB), the fiscal watchdog within the federal government, issued an order requiring that federal agencies consolidate data centers in order to save money and reduce redundancies. Twenty-four agencies, including elements of the Departments of Homeland Security, Treasury, and Defense, were tasked with identifying data centers under their purview, then to determine which could be closed and at what cost savings. Six years later, these agencies have collectively identified 10,584 data centers, of which 3,125, or about 30 percent, were reported closed as of the end of FY 2015. Another 2,078 were slated to be closed by the end of 2019, for a total of 5,203, or just shy of half the total number.
That would seem to be a significant savings, but a report issued last month by the US Government Accountability Office concludes that of the nine metric targets set by OMB, such as power usage, server utilization, and “virtualization density,” only one was met by half of the agencies, with the other eight falling short of even that halfway mark. Furthermore, the report continues, the agencies as a whole have underreported potential savings; the figure now stands at $8.2 billion, but “planned savings may be higher because 10 agencies that reported planned closures from fiscal years 2016 through 2018 have not fully developed their cost savings goals for these fiscal years.”
The GAO report indicates numerous pathways to improvement, noting, for example, that across the spectrum, only about 5 percent of federal server space is actually being used. OMB has recommended but not yet put in place a uniform metric for server utilization, which would seem to be a first step toward improving the efficient use of computer resources. Perhaps surprisingly, in response, the Department of Defense has stated that it “is moving toward commercial cloud hosting services to enable the migration of workloads to more efficient environments, thus positively affecting the virtualization and density metrics.” How hardened those cloud hosting services are—safe from hacking, that is—is, we imagine, a matter under review inside the Pentagon.
That 5 percent figure, incidentally, comes from 2009, indicating another source of frustration noted by the GAO evaluators—namely, that almost all the agencies in question have been slow to provide relevant, up-to-date data. One issue, GAO observes, is that some of the agencies in question are “decentralized,” so that cumulative data are hard to come by. Another, an objection raised particularly by the Department of Transportation, is that many data centers are housed in leased space without “dedicated metering” or uniform measures of performance and savings. One implication is that owners of leased property, if not the agencies themselves, may be required to put such measures in place.
The report recommends pointedly that all of the agencies take steps to improve their reporting and forecasting. This is especially pressing, the report adds, because of federal FITARA requirements that agencies calculate annual cost savings as part of an overall strategy to improve efficiencies and reduce costs.
The report concludes, somewhat circularly, that “until agencies take action to improve progress against OMB’s metrics, including addressing any challenges identified, they could be hindered in making progress against OMB’s optimization targets.” Yet, as noted above, many agencies have struggled to develop planned data center cost savings and avoidance targets. In other words, if the agencies don’t know what they’re supposed to be saving, they won’t be able to meet OMB’s goals. Or, put more simply: “If it can’t be measured, it can’t be managed.”