Enterprise IT

Supercomputing Facility Saves Energy Costs for NASA

At its high-performance computing facility in Silicon Valley, NASA’s Ames Research Center provides world-class processing capabilities that enable researchers to improve air transportation, study the earth’s interconnected systems and explore the moon and Mars.

To house its most powerful supercomputers, GDIT helped the agency design an innovative approach to implement one of the most energy-efficient computing facilities in the world.

The supercomputing capabilities at NASA Ames needed to be upgraded to accommodate additional equipment, but the existing building didn’t have enough power. Upgrading the current space with additional power and cooling infrastructure would had been very expensive and the replacement of legacy equipment still remained useable. High performance computers also use massive amounts of energy to cool heat-generating hardware components, driving high energy costs.

GDIT developed an alternate approach for cooling the HPC system, using evaporative cooling to take advantage of the local climate of Northern California. The new supercomputing facility is modular, consisting of small, segmented units fitted with high performance computers, rather than a traditional data center. Refined in GDIT’s HPC Center of Excellence, this site-specific approach increased power capacity while lowering energy costs to cool the equipment.

Through this collaboration, NASA Ames has gained the ability to fully cool its supercomputing hardware 38 percent more efficiently than the industry standard, resulting in significant cost savings. And virtually all of the energy consumed is used for computing, instead of a significant portion being diverted for cooling.

The energy savings and space savings may allow the agency to install new equipment in the future that could be used for scientific breakthroughs.

sq. ft. foundation, about the size of a football field
modular data centers can be supported
megawatts of HPC compute equipment can be supported
lower energy usage than industry average for supercomputing facilities