DC-Powered Data Centers: Sooner or Later?
A small Silicon Valley testbed run by Sun Microsystems and Lawrence Berkeley National Lab has achieved significant power savings through the use of DC power distribution instead of AC (grid) power. The study found that energy savings of 10 to 20 percent were possible by the more efficient movement of power through the facility.
The study is seen as a step towards convincing data center end users that DC distribution is a workable alternative to AC power, which must be converted to DC within the data center, with more than 20 percent of total wattage often being lost in the process. The LBL team noted that DC power distribution “has not yet made significant inroads into many data centers because the technology is unfamiliar to many facilities engineers.”
With power issues dominating data center budgets and a real-world implementation available, is a shift to DC power around the corner? One of the leading advocates of DC data centers is Peter Gross, the CEO and chief technology officer of EYP Mission Critical Facilities, an engineering firm that has participated in many high reliability projects. Gross outlined the case for DC data centers at the DataCenterDynamics event in New York in May.
“DC distribution in data centers has been somewhat controversial,” Gross noted. “The DC solution is not new, unique or exceptional. But it’s very difficult for most people to envision such a drastic change from the way we do things now. It will take some time for the culture to change to be ready to accept this.”
The Sun/LBL web site notes that “with the advent of servers on the market that can operate with either AC or DC, it is possible to use the DC powering approach, thus eliminating extra power conversion steps and losses.” EYP believes DC distribution can provide 20 to 35 percent savings – even larger than the gains seen in the California test environment.
Gross said the high-end data center user base has diverged in recent years, with financial institutions rarely exceeding 180 watts/SF, while search engine and supercomputing data centers at 200 watts/SF and beyond (with some nearing 400w/sf). The second group uses high-density clusters, providing redundancy that makes the reliability of individual components less critical.
“Reliability is still important, but accomplished differently,” said Gross. “The two big decision drivers for these users is the cost of energy and the ability to support high-density environments. The cost of electricity of these users is poised to exceed the cost of the IT equipment. Servers are cheap, but power is expensive. This is a delicate but important shift in our industry.”
The question-and-answer session following Gross’ presentation indicated lots of interest in the concept, but also lots of pointed questions and reservations about the transition. A number of attendees expressed concern about trying to pitch a DC conversion to IT executives at their companies. Many seemed interested in seeing how it would work in someone else’s data center. Thus, the Sun/LBL studies are likely to be closely studied.
“We believe DC is not only the efficient way to go, but the reliable way to go,” said EYP’s Kfir Godrich, who presented along with Gross at the NYC event. “We believe you’ll see this more and more, but it’s going to take some time for people to migrate from AC to DC servers.”