Cray’s XT Jaguar has raced ahead of its competitors and promises to accelerate the pace open scientific research.
Supercomputing has never been any more super than it is now at the Oak Ridge National
Laboratory (ORNL) in Oak Ridge, Tennessee. In November 2008, ORNL researchers announced that
the recently upgraded Cray XT Jaguar supercomputer offered a peak performance of 1.64 quadrillion
calculations per second—or 1.64 petaflops—making it the most powerful computer ever seen for
open scientific research.
Jaguar is the second supercomputer to break the “petaflop barrier,” the first being the IBM
Roadrunner system installed at Los Alamos National Laboratory (LANL) in Los Alamos, New Mexico
(see “Roadrunner Keeps an Eye on the Nation’s Nuclear Arsenal,” page 82). The new Jaguar uses
45,376 of the latest quad-core Opteron processors from AMD and features 362 trillion bytes (362
terabytes) of memory and a 10-petabyte file system. The machine has 578 terabytes per second of
memory bandwidth and an unprecedented input/output bandwidth of 284 gigabytes per second to
greatly facilitate the movement of data into and out of processors.
The system has already demonstrated its prowess, boasting the only two real (i.e., nonbenchmark)
scientific applications to date that have exceeded a petaflop. One, which investigates high-temperature
superconductors, achieved an unprecedented 1.35 petaflops to win the prestigious Gordon Bell Prize,
awarded to recognize the world’s fastest scientific computing application. The second, which simulates
the electronic structure of systems such as magnetic hard drives, achieved 1.05 petaflops. Both
achievements came from teams led by ORNL’s Thomas Schulthess.
Jaguar also took top honors in the annual High-Performance Computing (HPC) Challenge awards,
which recognize excellence in handling computing workloads. Jaguar took first place in two of the
four categories (speed in solving a dense matrix of linear algebra equations and sustainable memory
bandwidth) and third place in another (speed in executing the Global-Fast Fourier Transformation, a
common algorithm used in many scientific applications).
Jaguar also recently has been aiding in the search for new drugs against Alzheimer’s disease. In 2007,
ORNL researchers Edward Uberbacher and Phil LoCascio used 100,000 processor-hours on Jaguar to
investigate the mechanisms by which a new class of drugs acts. The drugs, called caprospinols, may stop the growth of Alzheimer’s fibrils and even disassemble the threadlike fibers. According to Uberbacher, it’s
the first time molecular dynamics has been used to simulate the mechanisms these drugs use to interact
with Alzheimer’s fibrils and reduce their growth. The findings, he adds, will provide researchers with new
ideas about how to improve the drugs that treat the disease.
ORNL originally contracted with Cray in 2004 to build a supercomputer for the researchers and
scientists who use the laboratory’s national supercomputing user facility. Since then, Cray and ORNL
researchers have rapidly upgraded Jaguar, and it is now the undisputed leader of the pack for open
scientific research. Jaguar has been put to work on a variety
of petascale research, including designing more efficient car
engines, probing the mysterious dark matter pervading the
universe, reproducing propagating shockwaves in a supernova
and simulating new superconducting materials.
The supercomputer may play a key role in handling
the simulations required for the next report from the
Intergovernmental Panel on Climate Change, the vanguard
organization that predicts the climate’s future. “Climate
change is one area where you don’t want to do the experiments,” explained Thomas Zacharia, ORNL’s
associate laboratory director for computing and computational sciences. “Instead, you want to model and simulate what’s happening. And you want the highest-fidelity calculations to minimize uncertainty.
That way, the policy decisions that have to be made are based on the most comprehensive, most sound
scientific calculations.”
Petascale Earth climate simulations enable scientists to incorporate increasingly complex global
models that take into consideration such parameters as atmospheric wind flow, ocean currents and
land and sea ice melt. The models help scientists better understand the ebb and flow of climate
processes and feedback, such as the global water cycle, which
in turn produces information on the timescales useful to resource
managers and planners. Only supercomputers on a par with
Jaguar can provide models with the kind of predictive skill
needed for making decisions about the future of Earth’s climate.
Supercomputer experts predict that Jaguar and other
petascale systems will transform science, acting as a “third pillar”
to discovery, one that complements the centuries-old pillars of
theory and experimentation. That is, supercomputers and their ability to simulate complexity now offer
another broad avenue toward understanding the universe. “Supercomputing, in many ways, touches all
aspects of humanity,” Zacharia reflected. “It has great power to accelerate progress. In fact, I think of
supercomputers as time machines. For the first time, they allow us to predict the future.”
|