![]() ![]() ![]() ![]() |
|
Earth and space sciences project awards $25.8 million to advance Grand Challenge applications NASA High Performance Computing and Communications (HPCC) Program's Earth and Space Sciences (ESS) project has entered its second phase with cooperative agreements totaling $25.8 million for the development of Grand Challenge applications 10 times faster than today. Resulting increases in complexity and resolution will provide a new understanding of fundamental problems in ESS disciplines. A three-year collaboration among NASA scientists, nine Science Team II investigator teams, and Cray Research will achieve these advances (see illustration, page 22). The challenges being pursued include modeling changes in global climate and the Earth's interior, simulating the evolution and dynamics of stars, probing microgravity environments, and processing remote sensing imagery and signals. For broader benefit, the new computer programs and documentation will be made available to the research community on the World Wide Web's National HPCC Software Exchange. "With 49 gigabytes of memory and 230 gigaflops peak performance, this system will be NASA's leading testbed for scalalble parallel computing, in which a program's speed increases proportionally with the number of processors." -James Fischer, ESS project manager
"This effort will further the earth and space sciences by helping to overcome one of high-performance computing's greatest bottlenecks the lack of usable software for parallel machines," said Lee Holcomb, director of the Aviation Systems Technology Division at NASA Headquarters. "Such computational studies strongly mesh with NASA's observational and theo-retical programs and contribute to our wider mission of scientific research and space exploration." Scientific breakthroughs will be enabled by a 384-processor CRAY T3E supercomputer being placed at the Goddard Space Flight Center (GSFC) as part of a $13.2 million agreement with Cray Research. "With 49 gigabytes of memory and 230 gigaflops peak performance, this system will be NASA's leading testbed for scalable parallel computing, in which a program's speed increases proportionally with the number of processors," said James Fischer, ESS project manager. Cray Research subsequently will assemble a CRAY T3E as large as 1,024 processors to allow 100 gigaflops sustained on investigator applications. Project Goals and History Centering on scalable parallel machines' performance advantages, the ESS Project has sought to demonstrate the potential of balanced teraflops systems for meeting The challenges being pursued include modeling changes in global climate and the Earth's interior, simulating the evolution and dynamics of stars, probing microgravity environments, and processing remote sensing imagery and signals. Grand Challenge application requirements. "Without an accelerated development program, this level of improvement may not be available for 15 to 20 years," said Robert Ferraro, ESS/Jet Propulsion Laboratory (JPL) associate project manager. A 1992 NASA Research Announcement was used to select Science Team I. Eight Principal Investigator (PI) teams addressed Grand Challenges, while a complementary set of 21 Guest Computational Investigators developed scalable algorithmic techniques. Science Team I investigators achieved 10- to 20- fold speed gains over 3 years, with several surpassing 10 gigaflops sustained. Among these was an atmospheric data assimilation routine programmed by the Richard Rood (GSFC Data Assimilation Office) PI team and ESS Project/JPL researchers that attained 18.3 gigaflops on a 512-node Intel Paragon, at least 120 times faster than on a CRAY C90. Accompanying such performance have been scientific discoveries that could not have been made without supercomputers. For instance, the George Lake (University of Washington) PI team's high-resolution cosmology simulations revealed that "galaxy harassment," a term they invented to describe continual mutual bombardment, can explain the rapid evolution of galaxies in clusters observed by the Hubble Space Telescope (see Nature, 2/15/96). Science Team II The selection process for Science Team II began in 1995 with the issuance of a Cooperative Agreement Notice soliciting proposals for PI teams and computing testbeds. Cooperative agreements are closely linked research efforts between NASA and the awardees in which all payments are triggered by meeting prenegotiated milestones. "Though evaluated independently, the investigators and testbed vendor ultimately are driven to collaborate to meet the milestones of 10, 50, and 100 gigaflops sustained performance on each team's application codes," said Lisa Hamet Bernard, ESS deputy project manager for testbeds. Computing Testbed Upgrades Their first milestone is being met on an interim CRAY T3D system (the CRAY T3E's predecessor) with 512 processors and 32 billion bytes of memory placed at GSFC this September. By June 1997, NASA and the investigators will complete transition to the 384-processor CRAY T3E. Access to larger CRAY T3E systems will occur before the program's conclusion in 1999. Time on the computers is being divided among the ESS Project and the CAS Project investigators and other NASA ESS researchers.
![]() For More Information Additional
information about the ESS project and Science Team II investigations is
available on the World Wide Web (http://ess.gsfc.nasa.gov/).
|
|||||||||
INSIGHTS was published by the High Performance Computing and Communications Program office.
Program manager: Bill Feiereisen
Editor, photography, design: Judy Conlon
Color separations: TH Graphics
Web Page design: Brandon Bailey