LabNotes - October 2013
NETL's Morgantown Supercomputer Sets a High Bar for Energy Efficiency
Most of us do our part to reduce energy consumption by some minuscule measure, even if it’s to simply save a few bucks rather than save the planet. Consider this: a single Google search consumes electricity equivalent to turning on a 60W light bulb for 17 seconds. In 2011, datacenters were estimated to be responsible for 1.3 percent of the world’s total electrical consumption. The engines that run these staples of our digital planet—known as supercomputers, or high-performance computing (HPC) resources—actually consume massive amounts of electricity. And, it’s a serious problem that is escalating quickly as our dependence on these technologies skyrockets.
The massive HPC that powers the NETL SBEUC is cooled using revolutionary methods to reduce energy requirements below DOE standards and far below the global average.
As a result, scientists have been hard at work optimizing energy efficiency of HPC systems, and the Simulation-Based Engineering User Center (SBEUC) at the National Energy Technology Laboratory’s (NETL) Morgantown, W.Va. facility is one of the finest examples of this effort. Similarly to how the fuel efficiency of cars is measured in miles per gallon, computing efficiency is measured in Power Usage Effectiveness (PUE), or the amount of power that is actually used by the computing equipment, versus that used in cooling and other energy demands. Because these large, tractor-trailer size facilities can generate substantial heat, in many cases, a large proportion of energy is used to cool the system. For example, most datacenters operate at about 1.6 PUE, meaning that for every 100w of computing electricity, 60w is used to cool the system. The DOE’s stricter standard wants that number at 1.4 PUE or better, with 1.2 PUE being optimal.
The NETL HPC center in Morgantown is addressing this problem on two primary fronts: reducing HPC energy consumption, and enabling researchers to devise more efficient fossil energy production and usage. One of the world’s largest HPC systems, the SBEUC is a critical lynchpin in U.S. strategic fossil-fuel research and energy policy.
Researchers at the SBEUC found a way to cut energy consumption by revolutionizing its cooling methods. “The 1.2 PUE target is a tall order, but it was part of our specification,” said David Speed, SBEUC Operations Manager with URS Energy & Construction, which configured the SBEUC for NETL, built upon the “Ice Cube” line of HPC systems by SGI. “The system we’ve built has achieved a PUE of 1.04.... This is far under the target, ranking us on par with some of the most efficient systems in the world. ”
But the SBEUC’s success doesn’t end with its cutting-edge cooling system. It also offers researchers massive parallel computing power to conduct virtual experimentation, enabling scientists to measure in days what it used to take weeks or months to determine.
“The SBEUC is accelerating R&D that would address environmental concerns around power generation, speed the development of new cutting-edge technology, and enable us to troubleshoot and optimize existing technology,” said Chris Guenther, director of the SBEUC at NETL. “All of the work that once would have been done manually can now be done on this computer, faster than we could have ever done before. It’s all about making research into the next generation of fossil fuel systems faster, cheaper and safer. ”
For more information on the NETL SBEUC, please see the article on the NETL Features page.