High Performance Computing in the U.S. the Next Five Years

Proceedings, 6th International Meshing Roundtable, Sandia National Laboratories, pp.3, October 1997

MESHING
RESEARCH
CORNER

Horst D. Simon:
Mail Stop 5OB-4230
Director, NERSC Division
Lawrence Berkeley National Laboratory
Berkeley, CA 94720, U.S.A
simon@nersc.gov.

Keynote Speaker
The years 1994- 96 brought on a remarkable change in the world of high performance computing both in the U.S. and consequently worldwide. Microprocessor based parallelism was successful far beyond the original expectations, and over a short period of time, highly parallel machines based on commodity microprocessors became the norm. This rapid transition had some unforeseen consequences: the high performance computing is now dominated by a small number of vertically integrated companies, where supercomputing is no longer the dominant business. I will offer some comments what the changes in the business world win mean for the supercomputer center managers and users.

Reinventing and re-engineering were not only buzzwords for the business world. All major U.S. super computer centers went through a period of major change and reevaluation in the 1995 -1997 time frame. NSF recompeted its four national center; DOD completed a modernization program resulting in four new consolidated centers; DOE competed the energy research supercomputing centers via the Grand Challenge competition. I will use the example of NERSC to show what impact these changes had on supercomputer centers and their mode of operation for next five years.

This change in the market place was accompanied by changes in the major federal programs in the U.S. The BPCC program ended in 1996, and new efforts are under way. However, the current driving force is the ASCI program in DOE. The ASCI "Red" machine at Sandia delivered the first Tflop/s calculation ever in late 1996. In the third part of my presentation I will review ASCI and other new programs in the U.S. such as Internet and compare their goals to the earlier HPCC.

About the Keynote Speaker:
Before becoming Director of Berkeley Lab's NERSC Division, Horst Simon was with the Advanced System Division of Silicon Graphics (SGI), where he managed SGI's university and research laboratory programs. From 1987 to 1994, Horst led a research department at the Computer Sciences Corporation at NASA Ames Research Center. Groups in his department studied parallel applications, scientific visualization, and numerical grid generation. Simon holds a Diploma in Mathematik from the TU Berlin, Germany, and a PhD in mathematics from UC Berkeley. His primary research interests are in the development of high-performance algorithms for vector and parallel machines.