Grand Or Not, It Still Sounds Like a Challenge To Me

While I'm not quite sure what differentiates a "challenge" from a "grand challenge", I have to admit that a "grand" challenge does sound more exciting than your run-of-the-mill everyday "challenge."

Of course, the first grand challenge that comes to mind is the DARPA Grand Challenge that pits autonomous ground vehicles against each other. This year's DARPA Grand Challenge will consist of simulated military supply missions in a mock urban area. However, that won't be happening November 3, 2007, with the location to be announced on August 10, 2007.

But DARPA doesn't have a monopoly on grand challenges. Lawrence Livermore National Lab, for instance, has its own "Grand Challenge" in which it provides scientific computing facilities for researchers. In the current challenge -- oops, make that "grand challenge" -- the Lab has allocated 83.7 million CPU hours to 17 research projects ranging from astrophysics, chemistry, materials, and biosciences to earthquake and climate simulation on Lab supercomputers.

Grand Challenge recipients have been allocated time on Atlas, a 44-teraFLOP machine, and Thunder, a 22-teraFLOP machine. Atlas is built around 9, 216 AMD dual-core CPUs (that's 1,152 nodes with 8 processors/node), 16 GB memory/node, 18,432 GB of total RAM, Infiniband interconnect, the Lustre parallel file system, and all running on Linux.

Thunder, on the other hand, is built around 4,096 Intel Itanium2 CPUs (1,024 nodes with 4 processors/node), 8 GB memory/node, 8,192 GB of total RAM, Quadrics switch interconnect, the Lustre parallel file system, and running on Linux.

Research projects were selected by the Lab’s Deputy Director for Science and Technology, the Laboratory Strategic Program Board, and the Laboratory Science and Technology Office. To be considered, proposals had to "address a grand-challenge-scale, mission-related problem that promises unprecedented discoveries in a particular scientific and/or engineering field of research, and if successful, will result in high-level recognition by the scientific community at large."

Over the last 10 years, high-performance computing resources dedicated to unclassified institutional research have increased more than a thousand fold, from 72 giga (72,000 million) FLOPS in 1997 to 81 teraFLOPS today.

"High performance computing is today an integral part of the scientific process and can serve to advance science in areas of vital interest to the country and the global community, including energy, bioscience and medicine, and climate change," said Dona Crawford, associate director for Computation at the Lab. "The new high performance computing resources we have put in place for unclassified research leverage the supercomputing expertise and infrastructure we have in place for stockpile stewardship. The Atlas Grand Challenge projects will in turn benefit our national security missions."

Posted by Jon Erickson at 09:21 AM  Permalink