Director's Perspective
Looking back, 1998 was another year of positive changes and great achievements for NERSC. When we set out to reinvent NERSC three years ago, our vision was to make NERSC a world leader in high performance computing by balancing excellent services with a first-class research program in computer science and computational science. Our latest accomplishments clearly demonstrate the success of NERSC's vision of a high performance computing center. I have read our annual report with great satisfaction. The report documents not only major progress in all areas at NERSC, but also the accelerated pace of scientific discovery we are enabling in our community of users.
During 1998 we completed our transition to HPSS; combined our two Cray T3Es into a single, 640-processor machine; decommissioned the C90; upgraded the J90 cluster; installed new math and visualization servers; completely overhauled the PDSF; and on top of all this began the procurement process for our next supercomputer. Almost every month there was a major systems change. This rapid pace of innovation was kept without ever dropping our high commitment to quality. A survey of our users confirmed that indeed we are doing extremely well in delivering these resources and services to the DOE Office of Science research community.
We ended the year with a strong showing at SC98, which demonstrated NERSC's intellectual leadership. NERSC researchers brought home the 1998 Sidney Fernbach Award recognizing efforts in computational science, as well as the Gordon Bell Prize for one of our Grand Challenge collaborations between NERSC, Oak Ridge National Laboratory, the Pittsburgh Supercomputing Center, and the University of Bristol. Not only did this team win the prize, but they also were the first to achieve a 1.02 teraflops performance with a real scientific application code producing real scientific results.
Our fast pace will continue throughout 1999. DOE just announced a new allocations policy for NERSC. This will bring major changes, since NERSC will operate more like other DOE national user facilities, with extensive peer reviews and outside advisory boards. The initial NERSC-3 platform will be online and available for early users within the next few months, and will bring major new challenges.
Horst D. Simon, Division Director of NERSC
Finally, there is a proposal to significantly increase the level of federal funding for information technology. Congress is considering a significant increase in the DOE Office of Science budget for this new initiative. NERSC is already deeply involved in the planning process, and we expect major new opportunities to arise within a year that will even further enhance our contributions to the DOE Office of Science mission.
With these exciting times ahead of us, I am again grateful to our DOE Office of Science sponsors for their continued endorsement of all of our ambitious plans. I would like to thank our clients, in particular ERSUG and EXERSUG members, for their continued support in transition times, especially for their willingness to collaborate on the new allocation process and their input to the NERSC procurement. My special thanks and congratulations, however, go to the NERSC staff for their skill, dedication, and tireless efforts to make NERSC the best scientific computing resource in the world.
The 1998 Gordon Bell Prize for best performance of a parallel supercomputer application went to a team of collaborators from DOE's Grand Challenge on Materials, Methods, Microstructure, and Magnetism. Their 1024-atom first-principles simulation of metallic magnetism in iron, with code tested and fine-tuned at NERSC, was the first complete application to break the teraflops barrier. Their performance achievement and research are described below.