High Performance Computing Meets Energy Efficiency

High Performance Computing Meets Energy Efficiency

NREL's most powerful energy research tool is also one of the most energy efficient data centers in the world.

A screen shot of a computer-generated simulation showing the turbulent nature of wind turbine wakes. Enlarge image

This computer-generated simulation shows the turbulent nature of wind turbine wakes. The simulation helped uncover potential differences in output between downstream 'waked' turbines and upstream turbines.
Simulation by Patrick J. Moriarty and Matthew J. Churchfield, NREL

The new High Performance Computing Data Center at the National Renewable Energy Laboratory (NREL) hosts high-speed, high-volume data processing capabilities that support the breadth of NREL's research. This research leads to increased efficiency and lower costs for important renewable energy technologies, including wind and solar energy, energy storage, and the large-scale integration of renewables into the Smart Grid.

Reflecting the lab's focus on energy efficiency, the data center's innovative design helps reduce energy use and, in turn, the typically high operational expense, making it one of the most energy efficient data centers in the world.

"Supercomputing" Advances Technologies

Housed in NREL's new Energy Systems Integration Facility, the High Performance Computing Data Center's powerful processing capabilities make it the fastest system in the world dedicated to advancing energy efficiency and renewable energy technologies.

Sometimes referred to as a "supercomputer," this system executes a massive number of calculations at ultra-fast speed for computer simulations and modeling used to deepen understanding of the material, biological, and chemical processes involved in clean energy technologies and integrating these technologies into larger energy systems.

A photo of the mechanical area for a data center. Enlarge image

The mechanical area for the data center houses the pumps and heat exchangers that are part of the center's energy-efficient liquid cooling design.
Photo by Dennis Schroeder, NREL

By early summer of 2013, the data center's computing system will run at petaflop-scale (1 million billion calculations per second) processing speeds. This ultra-high-speed processing will be capable of unprecedented modeling and simulation of materials and processes, some of which would otherwise be too expensive, too lengthy, too dangerous, or otherwise impossible to study by direct experimentation.

"Modeling and simulation capability is key to advancing technologies," says Steve Hammond, director of NREL's Computational Science Center. "Models and simulations allow researchers to dive deeper into subjects, and can give us a greater understanding than we'd gain through direct observation."

Exactly how will this remarkable high-speed computer processing be used? "We'll solicit requests and allocate large computational capabilities to specific projects," Hammond explains. "We'll be able to meet computational challenges that we haven't been able to address before because of a lack of adequate computer processing capabilities."

Innovative Energy Efficiency Built In

NREL's High Performance Computing Data Center will set a new standard of energy efficiency for data centers, which typically consume large amounts of electricity. The data center is designed for an annualized average power usage effectiveness (PUE) rating of 1.06 or better. The PUE is a ratio of the power the entire data center uses divided by the amount of the power required for the actual computational and data processing.

In contrast with the new system's projected PUE, and according to the Environmental Protection Agency's Energy Star Program (2009), the average data center operates with a PUE of 1.90 or greater. In addition, NREL's data center is designed to re-use nearly all of the waste heat it generates. No other existing data center achieves similar energy-savings benchmarks.

Warm Water Liquid Cooling

FLOPS is an acronym for FLoating point OPerations per Second—a critical measure of computing power and speed. A teraflop is one trillion calculations per second; a petaflop is a quadrillion, or a thousand trillion, calculations per second. The processing capabilities of NREL's High Performance Computing Data Center will soon be at the petaflop scale, enabling a new level of large-scale, high-speed processing.

When computer chips run heavy processing loads, they produce a large amount of heat. A variety of cooling techniques are used to keep these chips at steady operational temperatures. NREL has adopted a holistic approach to data center operations. Of particular note, the high performance computing systems in the data center are kept from overheating by using a unique warm liquid-cooled approach. Liquid has approximately 1,000 times the cooling capacity of air; therefore circulating liquid to cool computer chips is much more efficient than fans circulating air. While many data centers use water for cooling, most use chilled water, which is more energy intensive than a warm water approach.

Waste Heat Re-Use

The data center also incorporates an innovative and energy-saving waste-heat capture and re-use process that uses the heat the high performance computing generates. After water circulates through heat exchangers to capture waste heat from the computing system, the heated fluid then serves as the primary source of heat for laboratory and office spaces in the building. Excess heat can also be exported to adjacent NREL buildings and, in winter months, piped under walkways to melt dangerous ice. The data center's process is projected to deliver substantial energy savings for the entire NREL campus.

Setting the Course for Tomorrow

Traditional data centers use approximately 3% of all electricity consumed in the United States alone—about the same amount used by the entire airline industry. NREL's High Performance Computing Data Center's state-of-the-art, energy efficient building design and operations technologies reduce electricity consumption and pave the way for data centers of the future.

The High Performance Computing Data Center will provide a multi-faceted basis for simulating future integrated energy innovations.

According to Carolyn Elam, ESIF manager, NREL is developing a core competency in supercomputing that will drive development of the next generation of systems simulation. This competency also will make the experience-driven refinements of systems integration simulations, operations, and controls available for future energy-system architecture, policy, and investments. In essence, the data center will work as an open laboratory to further explore how large-scale high performance computing systems should be managed and integrated into a broader energy ecosystem.

Learn more about the ESIF High Performance Computing Data Center.

—Kathryn Ruckman

Top of page