DOE supercomputer could aid global warming research

The most powerful supercomputer in the world may be used for global warming research, according to officials at the Energy Department and IBM.DOE’s Office of Science, DOE’s National Nuclear Security Administration (NNSA) and IBM will share the cost of a five-year, $58 million research and development effort to build machinery that will surpass the fastest computer in existence — the IBM BlueGene/L system at NNSA’s Lawrence Livermore National Laboratory.“That could change the landscape of computational research in the next few years both for national security and for the scientific community at large,” said Dimitri Kusnezov, director of the Office of Advanced Simulation and Computing at NNSA.Blue Gene/L is a classified computer used only for nuclear security programs. The Top 500 list of the world’s fastest computers, a semiannual ranking, crowned BlueGene/L the winner earlier this month — for the third consecutive time.Under the agreement, scientists at DOE’s Argonne National Laboratory and Lawrence Livermore, and computer designers at IBM will develop technology capable of modeling predictions for nuclear weapon decay, human genome mapping and environmental change, according to IBM and NNSA officials. The contract does not include purchase of the system.The first incarnation of the technology, BlueGene/P, is in the 2007 budget request for Argonne, DOE spokesman Jeff Sherwood said. Lawrence Livermore has not committed to purchasing the system, Kusnezov said, adding that the laboratory officials will decide in 2008, after evaluating the research.“The R&D allows us to hedge our bets,” he said.The goal of the R&D phase is to produce a software environment for scientific exploration on top of an architecture scalable to hundreds of thousands of low-power processor cores, IBM officials said.“With Blue Gene, we went from working on a thousand processors to a hundred thousand” processors, Kusnezov said. “Now, we’re talking to moving that up to maybe a million. The number of processors we’re talking about using is just hard to fathom.”A successor to BlueGene/P, theoretically dubbed BlueGene/Q, is expected in the 2011 time frame, he said.NNSA and the Office of Science will each contribute $17.5 million, and IBM will contribute $23 million to the R&D effort.Kusnezov said the success of BlueGene/L attracted the interest of officials, governmentwide.“People realize that you can actually buy petascale computing at a more or less affordable price,” he said. The system is energy-efficient and relatively compact, fitting into two rooms — rather than consuming a building, he said.