[Congressional Record Volume 147, Number 11 (Monday, January 29, 2001)]
[Senate]
[Pages S636-S638]
From the Congressional Record Online through the Government Publishing Office [www.gpo.gov]

      By Mr. BINGAMAN (for himself, Mr. Craig, Mr. Schumer, and Mrs. 
        Murray):
  S. 193. A bill to authorize funding for Advanced Scientific Research 
Computing Programs at the Department of Energy for fiscal years 2002 
through 2006, and for other purposes; to the Committee on Energy and 
Natural Resources.
  Mr. BINGAMAN. Mr. President, I rise today to introduce a bill 
authorizing the Secretary of Energy to provide for the Office of 
Science to develop a robust scientific computing infrastructure to 
solve a number of grand challenges in scientific computing. This bi-
partisan bill, which is referred to as the ``Department of Energy 
Advanced Scientific Computing Act'' is co-sponsored by Senators Craig, 
Schumer, and Murray. Before discussing this program in detail, let me 
briefly frame the proposed effort. First, I will outline the tremendous 
advances made in the last decade for scientific computing. Second, I 
will give a few examples of the ``grand challenges'' in scientific 
computing. Third, I will discuss how the proposed program at the Office 
of Science will give our nation's scientists the tools to meet these 
grand challenges. I will conclude by demonstrating how this program 
integrates with defense related computing programs at the DOE and 
across the interagency.
  Experts agree that scientific computing R&D is at a critical 
juncture. If the breakthroughs proceed as predicted, the information 
age could affect our everyday lives far beyond what we nonexperts 
currently grasp. It is terribly important that we, as a nation, ensure 
that the U.S. maintains a leadership role in scientific computing R&D. 
If we fall beyond in this rapidly changing field, our nation could lose 
its ability to control the national security, economic and social 
consequences from these new information technologies.
  What are the possible breakthroughs in scientific computing that 
merit such strong programmatic attention? Within the next five years we 
expect that advanced scientific computing machines will achieve peak 
performance speeds of 100 teraflops or 100 trillion

[[Page S637]]

arithmetic operations per second; that is 100 times faster than today's 
most advanced civilian computers. To put things in perspective, the 
fastest Pentium III available today can perform about 2 gigaflops (2 
billion operations per second), so a 100 teraflops machine is about 
50,000 times faster than today's fastest Pentium III. We call this new 
wave of computing ``terascale computing''. This new level of computing 
will allow scientists and engineers to explore problems at a level of 
accuracy and detail that was unimaginable ten years ago. I will discuss 
the scientific and engineering opportunities in more detail later. 
First, let me discuss some of the challenges in terascale computing.
  The major advance that led to terascale computing is the use of 
highly parallel computer architectures. Parallel computers send out 
mathematical instructions to thousands of processors at once rather 
than waiting for each instruction to be sequentially completed on a 
single processor. The problem we face in moving to terascale computers 
is writing the computer software that utilizes their full performance 
capabilities. When we say ``peak'' speeds we mean the ability to use 
the full capability of the computer. This happens very rarely in 
parallel computers. For example, in 1990 on state-of-the-art Cray 
supercomputers with about eight processors, we could obtain, on the 
average, about 40-50 percent of the computer's ``peak'' speed. Today, 
with massively parallel machines using thousands of processors, we 
often obtain only 5-10 percent of the machine's ``peak'' speed. The 
issue is how to tailor our traditional scientific codes to run 
efficiently on these terascale parallel computers. This is the foremost 
challenge that must be overcome to realize the full potential of 
terascale computing.

  Another problem we face as we move to terascale computing is the 
amount of data we generate. Consider the following. Your PC, if it is 
one of the latest models, has a hard drive that will hold about 10 
gigabytes of data. If we successfully begin to implement terascale 
computing, we will be generating ``petabytes'' of data for each 
calculation. A petabyte of data is one million gigabytes or the 
equivalent of 100,000 hard drives like the one on your PC. A teraflop 
machine user will make many runs on these machines. But raw data isn't 
knowledge. To turn data into knowledge, we must be able to analyze it--
to determine what it is telling us about the phenomena that we are 
studying. None of the data management methods that we have today can 
handle petabytes data sets. This is the second challenge that must be 
overcome.
  And, many more challenges exist.
  To make effective use of today's and the future's computing 
capability we need to establish a scientific program that is radically 
different from what researchers are used to today. Future scientific 
computing initiatives must be broad multi-disciplinary efforts. 
Tomorrow's scientific computing effort will employ not only the 
physicist who wishes to probe the minute details of solid matter in 
order to say, built a better magnet, it will include a computer 
scientist to help ensure that the physicist's software makes efficient 
use of the terascale computer. Terascale computing will also require 
mathematicians to develop specialized routines to adapt the solution of 
the physicist's mathematical equations to these parallel architectures. 
Finally, terascale computers will require specialists in data 
networking and visualization who understand how to manage and analyze 
the massive amounts of data.
  I note these problems to highlight the complexities of tomorrow's 
scientific computing environment from the common information 
technologies that we employ today. However, because computing 
technology moves at such a rapid rate, elements of the issues that I 
have described will surely impact us in the near future. Given the 
impact information technologies have had only in ten years, it is 
important that we, as a nation, lead the initiative in these 
breakthroughs so that we can positively control the impact that the 
these revolutionary technologies will have on our economy and the 
social fabric of our Nation.
  What are the important problems that we expect terascale computing to 
address? We call these problems ``Grand Challenges''. Terascle 
computing will enable climate researchers to predict with greater 
certainty how our planet's climate will change in the future, allowing 
us to develop the best possible strategies and policy for addressing 
climate change. Terascale computing will help chemists understand the 
chemical processes involved in combustion, which will translate into 
more efficient, less polluting engines. Terascale computing will allow 
material scientists to design nanomaterials atom by atom, which will 
lead to stronger, yet lighter and hence more energy efficient 
materials. Terascale computing will assist nanoscience researchers by 
simulating atom manipulation before undertaking complex and expensive 
experiments. Nanotechnology will lead to whole new generations of 
computer chips, information systems, and stronger, yet lighter 
materials. Finally, terascale computing will enable biologists to 
understand the structure of the proteins encoded in the human genome, 
which will lead to better medicines and health for our citizens. These 
fundamental grand challenge problems are now addressable with the 
recent advances in scientific computing. Due to the impact the grand 
challenge problems will have on our lives, we as a nation, must take 
the lead in their investigation.
  What are the elements of the proposed effort? The program I propose 
will build on the Department of Energy's decades of leadership in high 
performance computing and networks to ensure that terascale computing 
and petabyte data visualization becomes a positive force for the U.S. 
The proposed program has four parts. The first part is the 
establishment of core teams of researchers who specialize in the grand 
challenge problem itself. An example of a core team is one made up of 
geologists and geochemists allied with computer scientists and applied 
mathematicians to write large software programs associated with oil 
exploration or the diffusion of waste in the subsurface. The scientific 
simulation software created by these core teams will be the ``engines'' 
that drive the scientific discovery process. The second element of the 
program enhances the research efforts in computer science and 
computational mathematics that underlie this software development 
effort. These specialists will ensure that the core teams effectively 
use massively parallel computers--not at the current 5-10 percent but 
at 50 percent of the computer's peak running speed. These specialists 
will also develop the software to manage and visualize the petabytes of 
data that the core teams, as well as the next generation of 
experimental facilities, generate. Third, this program will fund 
specialists to develop the networking and electronic collaboration 
software that will allow researchers all across the U.S.--in national 
laboratories, universities, and industry to routinely use petabyte data 
sets. This new networking capability will translate quickly to the 
private sector in the areas of medicine, business transactions, and 
education over the internet. Fourth, this program will fund the unique 
computer hardware required for scientific investigations of the ``Grand 
Challenges'' on a continuing basis. Many of the grand challenge 
problems will benefit from specialized computers. This program will 
fund such specialized computers. For instance, IBM will build in the 
year 2004 or 2005 a unique 1000 teraflops (1000 trillion operations per 
second) computer called ``Blue Gene''. Blue Gene will be 500,000 times 
faster than your desk PC. This machine will be used by DNA researchers 
to predict the structure of proteins and in doing so allow drugs and 
medicines to be optimized before they are commercially produced. We 
propose to place these one-of-a-kind computers at national user 
facilities and make them available to U.S. researchers in national and 
government laboratories, universities, and industry.

  In summary, we are proposing a program that will substantially 
advance our understanding of complex scientific phenomena that affect 
our daily lives. At the present we cannot fully understand these 
phenomena; it is critical that we master it in our national interest so 
to benefit our nation and its people.

[[Page S638]]

  Overall, this program will integrate into other DOE advanced 
computing efforts and into our national strategy for advanced 
scientific computing. In FY01, the DOE National Nuclear Security 
Agency, NNSA, funded the Accelerated Strategic Computing Initiative or 
ASCI at $477 million dollars. ASCI's mission--to develop the capability 
to simulate the safety and surety of the nuclear weapons in our 
stockpile--is critical to the security of our nation. The ASCI program 
is a focused and classified program with one primary user--the nuclear 
weapons community. Its problems revolve around materials and plasmas 
undergoing rapid changes from a nuclear explosion. The Advanced 
Scientific Computing Program I am proposing is unclassified and covers 
many other areas of science critical to the long term well being of the 
nation. This program will involve interaction between researchers at 
the nation's national and federal laboratories, universities, and 
industry. That is not to say that there will be no integration between 
these two worthy and important efforts. Both efforts involve terascale 
computers, so clearly we expect that many of the central tools common 
to both in terms of hardware design and underlying software for 
networks and visualization will be shared. Both programs will benefit 
by the two diverse communities working towards the common goal of 
terascale computing. And, the NNSA will be able to infuse fresh ideas 
from the universities and industry on parallel architectures and data 
visualization into their efforts in ensuring the surety of our nation's 
nuclear weapons stockpile.
  Within the U.S. Government, this effort will fall under the purview 
of the National Coordinating Office for Computing, Information and 
Communications, ``NCO/CIC''. This Office is charged with coordinating 
government-sponsored information technology research programs across 
all of the government agencies. The NCO/CIC provides a forum for DOE to 
coordinate its scientific computing program with information technology 
programs in NSF, DOD, NASA, NIH, NOAA, and other government agencies 
interested in high-performance computing. Although the DOE program is 
focused on its energy, environmental, and scientific missions, many 
benefits will be derived by coordinating its activities with related 
computing activities in other agencies. Finally, I note that in our 
national implementation plan for ``Information for the Twenty First 
Century'', the NSF and the DOE were given the leadership for ``Advanced 
Scientific Computing for Science, Engineering and the Nation''. The 
program I have outlined supports that role.
  In summary, I have outlined a scientific computing program that will 
advance our ability to understand complex but important physical, 
chemical, and biological phenomena. Advancing our understanding of 
global climate change will lead to a better understanding on the 
relationship between our energy consumption and the climate on our 
planet. Mastering materials and chemical processes at an atomic level 
will enhance U.S. industrial competitiveness in many areas such as 
energy efficient materials manufacturing and develop new computer chip 
technologies. Understanding the flow of contaminants in the groundwater 
will help develop better strategies for cleaning up DOE's sites and 
help commercial oil and gas extraction. Predicting the structure of 
proteins will lead to more effective drugs with minimal side effects. 
Beyond solution of the ``Grand Challenges'' are the advancements that 
will be made in advanced computing and networking technologies which 
will benefit users in areas as diverse as medicine and business. These 
problems are of national significance to the health of our citizens and 
our future economy in the 21st century.
                                 ______