"High Performance Computing, Now and in the Future"

                        by

                Olaf O. Storaasli
         Distinguished Research Scientist
             Future Technologies Group
         Computational Sciences Directorate
             Oak Ridge National Laboratory

When I performed Structural Analysis Computations at NASA Langley in the 70's
people were astounded by our achievment of Millions of FLoating Point
Operations/Sec (MFLOPS), truly High_performance Computing, then. Not until
1989, was the next level of performance (1000 MFLOPS) achieved by a NASA-ODU
team (Storaasli-Nguyen-Agarwal) who received Cray's 1st GigaFLOP Award (1
billion FLoating Point Operations/Sec) who achieved 1.334 GFLOPS performance
for the static analysis of a "gigantic" Space Shuttle Solid Rocket Booster with
54K equations. In 1998 another Cray T3E at ORNL broke the TeraFLOP (1,000,000
MFLOPS).  Not until 2008 did DOE computers (at LANL and ORNL) exceed 1 PetaFLOP
(1,000,000,000 MFLOPS), a billion times faster than the early NASA computers.
See details on ORNL's Jaguar at:

http://www.energy.gov/discovery/the_supercomputing_fast_lane.html

DOE views Supercomputers as the 3rd leg of scientific discovery along with
theory and experiment and that an increasing number of significant
discoveries will be performed on supercomputers.

  Dr Storaasli will describe the current top fleet of supercomputers (mostly
DOE), what they consist of (hardware and software), and what applications
they solve.   In addition, based on his research in the Future Technologies
Group at ORNL, Dr. Storaasli will project what the next level of
supercomputers (ExaFLOP) will look, including their architecture, software,
tools and applications and what performance levels they will achieve.

Dr. Storaasli is an internationally-known expert on parallel methods for
computational analysis on High-Performance Computers (HPC) having conducted
research in this area for over thirty years.  Before parallel computers were
sold, Dr. Storaasli led a NASA hardware-software-applications team to develop
the Finite Element Machine (Wikipedia), one of the 1st parallel computers. As a
NASA senior computational scientist, his research developed new parallel
algorithms to solve large systems of matrix equations for aerospace structures,
electromagnetic and acoustic applications.  These were first developed to
harness cutting-edge NASA supercomputers (1st Cray GigaFLOP award and NASA
Software of the Year Award), then parallel computers and recently even
faster accelerators sped by Field-Programmable Gate Arrays (FPGAs).

At NASA he edited six books and authored over 100 works in computational
mechanics including static and dynamic structural analysis, eigenvalue &
optimization methods, interdisciplinary analysis, data management, and
parallel-vector structural analysis methods on supercomputers.  He received
NASA.s prestigious Floyd Thompson Fellowship for post-doctoral research at the
Norwegian University of Science & Technology in Trondheim and Det Norske
Veritas in Oslo, Norway, in 1984-85.  He has been invited to give dozens of HPC
lectures including 16 overseas.  He conceived and led a NASA Creativity and
Innovation team to explore FPGAs as an alternative to traditional processors to
accelerate scientific and engineering computing.  This led to NASA.s $15M award
for his reconfigurable scalable FPGA computer project for space applications
(image compression and robotics).

Current Assignment
Dr. Storaasli joined ORNL.s Computer Science & Mathematics Division.s Future
Technologies Group in 2005 to explore new algorithms-architectures to
accelerate scientific supercomputing by harnessing FPGAs.  His research
evaluated multiple FPGA systems at ORNL(6), Cray(6), Edinburgh University (64)
and the Naval Research Lab (150).  Speedups of 10-100X over traditional
processors resulted for key Department of Energy (DOE) analyses
(climate/weather, molecular dynamics, genomics, large matrix equation
solution). His research spun off DOE (Science Alliance and SBIR) and USAF
research awards.  Dr. Storaasli currently leads a cross-cutting ORNL team
planning to apply rapid FPGA genomics to future low-cost personalized
medicine.