REINVENTING NERSC

Founded nearly 25 years ago, the National Energy Research Scientific Computing Center was the first unclassified supercomputer center and was the model for those that followed. Our relocation from Livermore to Berkeley Lab in 1996 set the stage for reinventing NERSC to meet the computational science needs of the next century. The new NERSC provides clients with not only state-of-the-art high-performance computers, but also an extensive range of services aimed at accelerating the pace of scientific discovery. The name change that accompanied the move -- from "Supercomputer Center" to "Scientific Computing Center" -- signaled a new philosophy, one of making scientific computing more productive, not just providing supercomputer cycles.

We have reinvented NERSC around that new philosophy. Our high-performance computing systems and excellent client services -- the NERSC program -- are now complemented by expanded expertise in computational and computing science -- the NERSC Division in Berkeley Lab's Computing Sciences organization. (This report describes both Program and Division activities.) Our goal is to give our clients both the computing and the intellectual resources necessary to conduct breakthrough computational research.

Our physical relocation set the tone for the changes to come. Through careful planning and extraordinarily hard work by the entire staff, the move was completed ahead of schedule and under budget -- with no interruption of services to our clients. During our first full year at Berkeley Lab, we welcomed more than 40 new employees from some of the top computing science organizations in the country to strengthen our scientific computing services. We further expanded our research capabilities by integrating several Berkeley Lab computational and computing science research groups into the NERSC Division. As a result, the Division has roughly the same number of staff engaged in research as we have providing day-to-day services, such as supporting our computing and storage systems.

By doing more with less, we have broadened our services without increasing our Program budget. For example, even though we are operating more machines than before, the technical staff dedicated to our core services is 20 percent smaller than it was three years ago.

In the midst of all these changes, we reached a milestone that finally brought massively parallel processing (MPP) into full production mode -- checkpointing and restart of our Cray T3Es. Checkpointing has been a major goal in the MPP community for the 10 years since the first parallel machine was plugged in, and this is believed to be the first successful checkpointing of an MPP system (see about checkpointing).

Our position as the nation's largest computing resource for unclassified research was confirmed by the November 1997 release of the TOP500 list of the world's most powerful computers (http://www.netlib.org/benchmark/top500.html). Our T3E-900 was listed as No. 5 in the world, and in this country is the No. 1 unclassified system.

We were also gratified to receive a favorable evaluation from an independent group of peers. The Lab Director's Review committee, which met in May, assigned a rating of Outstanding/Excellent to Computing Sciences activities at Berkeley Lab. In their report, the committee wrote, "It is a remarkable achievement to continue to adhere to past high standards while integrating a major new activity and holding the new activity to the same high standards."

These achievements and many other highlights of the year are described in more detail in the following pages. Scientific accomplishments are presented in the "Science Highlights" section.


Next Page
Back to Table of Contents