ORNLís Richard Alexander shows the inside of a dual-processor Pentium Xeon computer (which has the same layout as a supercomputer) to Boy Scouts (from left) Morgan Alexander, Liam Holland, Miles Pacheco, and Evan Meredith. Educating youth about computers may attract more talent into the growing area of computational sciences. (Photo by Jim Richmond)
Click here for more photos.
Oak Ridge National Laboratory is home to some of the mostpowerful open, or unclassified, computers in the nation. These recently acquired supercomputers will soon give the Department of Energy's Center for Computational Sciences (CCS) at ORNL a total computing speed approaching 6 teraflops, or 6 trillion arithmetic calculations per second. They have also advanced ORNL's leadership role in scientific discovery that is not possible without high-performance computing.
The IBM Power4 supercomputer (dubbed
Cheetah by ORNL researchers), delivered
to ORNL in stages in late 2001, has a
computing speed of 4 teraflops. The IBM
RS/6000 SP (Eagle) supercomputer and
Compaq Alpha-Server SC (Falcon) system
together offer 1.5 teraflops of computing
power. As a result, researchers can solve
complex scientific problems in a virtual
computational environment. ORNL
researchers also are involved in
"collaboratories," in which scientists from
different sites share data over computer
networks as if they were working side by
The development of new algorithms by
ORNL and University of Tennessee
computer scientists will allow researchers
to solve even more complex scientific problems more efficiently through simulations of
experiments. Terascale computing is a powerful tool for analyzing, understanding, and
predicting scientific phenomena because it serves as a bridge between theoretical
understanding and experiment.
In this issue of the ORNL Review, the first section (Computing Infrastructure) describes
the supercomputers at the heart of CCS and the computational infrastructure that is in
place to support breakthrough computational science. We also describe our tools for
monitoring and evaluating performance-for example, identifying the type of
supercomputer on which a numerical code performs best and guiding decisions on which
type of supercomputer to purchase next. We discuss our development of computer tools to
enable scientists to run their codes more efficiently on supercomputers and the creation of
a Scalable Systems Software Center for the preparation of software that will effectively
manage terascale computational resources. We also discuss our contribution to the
development of the computer industry's leading data-storage system-in terms of capacity
and transfer speed-and Probe, our new storage research facility to improve data storage
and transfer for terascale systems.
Another article discusses our work in devising ways
to improve our ability to send large data files over
the Internet so that supercomputers are not idle
because of delays in data delivery. ORNL is building
a high-speed fiber-optic link that will connect
Laboratory supercomputers with those in Atlanta,
Georgia, and Research Triangle Park near Durham,
North Carolina. Through this network, huge volumes
of data and calculation results will be transferred
among these supercomputers and, later, the new
100-teraflops Blue Gene supercomputer that ORNL
is helping IBM to develop. The Blue Gene machine
will be used to relate protein shapes to diseases. ORNL researchers will contribute their
expertise in developing fault-tolerant algorithms and predicting protein structures.
Visualization tools for supercomputers provide insight into physical phenomena, help
scientists verify calculated results, highlight the unexpected, and enable scientists to
communicate their results more effectively. In this issue, we show that some visualization
tools are being used in a CAVETM virtual reality theater at ORNL, to enable scientists to
interact with predicted phenomena that relate, for example, to stellar explosions and
Our plans for a new building to accommodate all our research groups and our
next-generation supercomputer are described, along with our partnerships with
universities and industry and how they will benefit from the ORNL-University of
Tennessee Joint Institute for Computational Sciences, which will also be housed in a new
The second section of this issue (Discovery by Computing) shows
why computers are needed to enable scientific discovery and why
scientific research that depends on high-performance computing is
important to the nation. Supercomputer-supported research at
ORNL falls into six areas of importance to DOE: astrophysics,
biology, chemistry, climate prediction, fusion energy, and
materials. Other computer-driven research at ORNL addresses
questions concerning energy efficiency and health, as well as
electronic devices that may result from advances in
nanotechnology. In this issue you will learn that
Studies of magnetic materials using supercomputers at ORNL
are paving the way for the next generation of information
technologies, including better digital cameras.
A proposed molecular memory cell that would enable laptop
computer batteries to last 100 times longer than today's
batteries is being modeled computationally on an IBM
supercomputer at ORNL. This machine is also being used to
simulate carbon nanotubes in contact with other components
that may be used in tomorrow's nanoscale electronic devices.
Multidimensional simulations of core-collapse supernovae by
ORNL and its partners could help determine how and why
stars explode and how elements are formed and disseminated
into space. This work in astrophysics could also advance the
understanding of combustion, future climate, fusion energy,
and radiation therapy.
A computational analysis of human and bacterial genomes by
ORNL researchers provides insights into what our genes do. ORNL researchers will
soon be computationally predicting 100 protein structures a day and will evaluate
which compounds could make highly effective therapeutic drugs.
Thanks to computer modeling, a scientific discovery was made that might lead to a
way to save victims of cardiac arrest.
Some of the world's largest global climate models are being run on ORNL's
supercomputers, providing insights for national and international assessments of the
effects of global warming caused by human activities.
Supercomputers can be used to simulate chemical reactions, saving time and money
and improving safety.
ORNL fusion researchers are using supercomputers to understand plasma turbulence,
design a device that could eliminate plasma disruptions, and find ways to get radio
waves not only to heat but also to control the plasma to allow sustained
energy-producing fusion reactions.
In support of energy efficiency research, ORNL researchers are building computer
models of vehicles made of aluminum, high-strength steel, regular steel, and
carbon-fiber composites. Other researchers are developing software tools for
supercomputers to simulate engine exhaust from various lean-burn diesel and
gasoline engine designs as it flows through envisioned catalytic converters designed
to chemically transform pollutants into harmless emissions. This research could lead
to safer and cleaner, energy-efficient cars.
Researchers in ORNL's Computational Sciences and Engineering Division are
developing software in support of efforts to ensure our homeland security.
This issue provides a window into supercomputing at ORNL and shows how
supercomputing helps scientists better "see" various phenomena and material structures
that could lead to more detailed scientific understanding and improved devices and drugs.
Clearly, ORNL is advancing science through advanced computing.
The Department of Energy's Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time.