Luckett & Farley provided architecture and engineering services for the renovation of the Department of Defense’s largest supercomputer in Dayton, OH. The project included extensive infrastructure upgrades and the improvements made to the HVAC are expected to reduce the facility’s energy consumption by 50%. As a result of Luckett & Farley’s specialized team, the enhanced supercomputer not only saves energy for the complex, but is able to solve the most demanding problems science, technology, and evolution has yet to imagine.
DoD’s largest supercomputer begins operations at WPAFB‘Raptor’ system enables scientists and engineers to develop technology solutions in response to warfi ghter and other concerns
By Amy Rollins Skywrighter Staff
The Air Force Research Laboratory’s Defense Supercomputing Resource Center (AFRL DSRC) unveiled the Department of Defense’s largest supercomputer at a ribbon-cutting ceremony March 4 in Bldg. 676, Area B, at Wright-Patterson Air Force Base.
The Cray Inc. XE6 model supercomputer, designated as “Raptor” after the Air Force’s F-22 fighter, will help to facilitate weapon systems design, advance design concepts, accelerate modifi cation programs for critical warfighting systems, enable higher fi delity modeling and simulation studies, and support computational science and engineering applications.
Raptor is powered with 43,712 2.4 GHz AMD Magny Cours computer cores. To put this in perspective, a typical desktop has a single processor, and can contain upwards of eight gigabytes of memory. Raptor has 87.7 terabytes of memory, equal to 89,804.8 gigabytes, and 1.6 petabytes of usable disk space. In addition to the Magny Cours eightcore processors, Raptor boasts the new Gemini interconnect, which allows the processors to communicate with each other at speeds on the order of 16 gigabytes per second.
“This system will provide more than seven times the processing power delivered by the AFRL DSRC last year,” said Joe Sciabica, executive director, AFRL. “I’m sure that many of our AFRL scientists and engineers have seized the opportunity to get early access to this tremendous capability, and they are excited about the potential to solve scientific and engineering problems we have yet to imagine.”
Contained within 30 computer racks, Raptor is capable of performing more than 400 trillion floating point operations per second, making it the 18th fastest computer in the world, said Frank Witzeman, director, AFRL DSRC. “To put it into perspective, Raptor could process about 200 billion IRS Tax Form 1040s in about 1 second – or all of the individual tax forms filed last year in less than 1 millisecond.”
Not only does Raptor operate at speeds almost beyond comprehension, it also utilizes environmentally friendly, “green” technology, incorporating high-efficiency, liquid-cooled cabinetry. The ECOphlex water-cooling system provides more direct and efficient heat exchange.
The AFRL DSRC facility also has been modernized with new power and cooling systems, Mr. Sciabica said, incorporating green, energy-efficient computer center technology, such as free cooling from low outdoor temperatures. This intense effort was necessary for the integration of the new $25 million supercomputer.
“This is a historic occasion where we do get to write history,” said Jeff Graham, senior technical advisor for the Computational Science and Engineering Office, AFRL, and technical director of the DSRC. Mr. Graham acted as master of ceremonies for the ribbon-cutting ceremony event.
Since its inception, the AFRL DSRC has been recognized as a leader in providing specialized services and expertise in supercomputing and computational sciences, Mr. Sciabica said. “It has continued to be a critical element in the Defense Department laboratories’ and test centers’ efforts to conduct a wide range of focused research, development, test and evaluation activities.” It is one of six DSRCs in the Department of Defense.
Mr. Sciabica highlighted a number of projects in which the AFRL DSRC has provided high-performance computing capabilities and expertise:
- To the Navy for its maritime and anti-submarine aircraft flight clearance program, thereby reducing testing costs and risks.
- By the Joint Forces Command to conduct urban battlefield simulations to identify and track threats for training and real-time operations.
- By the Energetic Ionic Liquids project, in which its capabilities designed new fuels to replace the bio-hazardous hydrazine for advanced rocket propulsion applications.
- By the Sensors Directorate’s “Gotcha” radar program, which leveraged the AFRL DSRC for facility space and expertise to support a dedicated supercomputer for near real-time image processing.
- By the Air Vehicles Directorate to pursue physics-based simulations of flapping wings, such as those on insects, to understand and support micro air vehicle technology development.
- By the Propulsion Directorate to investigate hypersonic flight, between Mach 5 and 10, to develop propulsion options for sustained hypersonic speeds of future aerospace vehicles.
“Supercomputers lead to scientific discovery, technology development and engineering knowledge by exploiting the use of higher-fidelity, physics-based simulations to address problems that are too costly, too time consuming or too environmentally dangerous to investigate under conventional experimental or test techniques,” Mr. Sciabica commented.
“In all these cases, scientists and engineers can work in virtual space, without putting anyone is harm’s way,” Mr. Witzeman added.
“Our goal is to continue our relationships with our long-standing ‘super users’ while reaching out to potential new users at the entry level of supercomputing, thereby covering a full spectrum of capabilities from the desktop to the mid-range high performance computing cluster to the massive supercomputer,” he said.
Peter Ungaro, president and chief executive officer of Cray Inc., said, “We see ourselves as building a bridge from the scientists’ and engineers’ minds to something that’s really important for our country.” He noted that it took five years and hundreds of people to build the technology that went into Raptor.
The system arrived at Bldg. 656 on Nov. 8 and was operational and being tested within a week. Mr. Witzeman said to test drive the new system and prepare Raptor for production use, a small number of Capability Application Project, or CAP, and pioneer users were hosted. The CAP users ran very large applications that stressed the limits of the system, whereas the pioneer users ran applications typical of expected, normal workloads.
In both cases, meaningful results were obtained in a very short period of time while the system was being stabilized for general use, Mr. Witzeman said. “In some cases, scientific discovery, advanced technology development or increased engineering knowledge resulted from early access to such a massive resource.”
High-performance computing capabilities will continue to evolve even after Raptor, Mr. Ungaro said. “From today to the end of the decade, we’re going to built supercomputers 1,000 times faster than Raptor.”