Documents/DOEER/6: Deliver Computing for the Frontiers of Science/6.3: Supercomputing Architectures

6.3: Supercomputing Architectures

Bring dramatic advances to scientific computing challenges by supporting the development, evaluation, and application of supercomputing architectures tailored to science.

Other Information:

Major improvements in scientific simulation and analysis can be obtained through advances in the design of supercomputer architectures. Most of today’s supercomputers were designed for commercial applications. However, computational science places stringent requirements on supercomputer designs that are often quite different from what arise in commercial applications. To meet the need for effective computing performance in the 100-teraflop range and beyond, we will support the evaluation, installation, and application of new very high-end computing architectures for computational science. Our strategy includes the following emphases: • Develop partnerships with U.S. industry in the near term to adapt current and nextgeneration products to more Computing test beds: Advanced Computing Research test beds evaluate new computing hardware and software, such as Oak Ridge National Laboratory’s IBM Power4 Cheetah (pictured left) and Cray Xl, and Argonne National Laboratory’s IBM/Intel/ Cluster. ORNL fully meet the needs of visionary computational science. • Develop partnerships with the Department of Defense, the Defense Advanced Research Projects Agency (DARPA), and other Federal agencies to evaluate long-term architecture developments at the scale needed for Office of Science computation. • Advance the focused research and development of systems software for radical increases in performance, reliability, manageability, and ease of use.

Indicator(s):