1997 Congressional Hearings
Special Weapons
Nuclear, Chemical, Biological and Missile

Testimony of Dr. Mitchel B. Wallerstein

Deputy Assistant Secretary of Defense for Counterproliferation Policy

Before the House National Security Committee

on November 13, 1997

The Defense Department views export controls as a critical part of our national security and defense policy. It is important that we deny potential adversaries those technologies and items that would provide them with key military capabilities that would enable them to seriously threaten U.S. military forces and our national security. DOD has always given serious attention to export controls on high performance computers because their capabilities are of significance in applications ranging from acoustic submarine detection to missile defense.

We also recognize, however, that computer technology continues to advance rapidly, and ever-increasing computing capabilities are widely available on a global basis. Computer users no longer have to rely on large, expensive, single vector "supercomputers" to solve complex problems. Rather, advances in microprocessor chip power have given computer manufacturers the ability to offer flexible, adaptable, and powerful workstations to perform high performance computing applications.

Thus, we must recognize that export controls only advance U.S. interests if they are effective. Ineffective export controls only create an illusion of security protection where -- like a Maginot Line -- none exists, and wastes scarce government and industry export control resources with no compensating security benefit.

Before we last revised computer controls in 1995, DOD conducted an unprecedented review of over 600 DOD military programs to determine how the Department uses computers and where our critical performance thresholds lie. The categories of applications investigated included surveillance, target detection and recognition, submarine design, aerodynamic vehicle design, explosives research, weapons guidance, and cryptography. Our findings indicated two clusters of critical DOD applications around 10,000 million theoretical operations per second (MTOPS) and 20,000 MTOPS. We also found that a significant number of applications below 10,000 MTOPS are increasingly being performed on workstations.

As part of a larger study by the Administration (which included all of the departments and agencies with national security, nonproliferation, and export control responsibilities), we also looked at trends in computer technology. We made judgments on the effectiveness of controls by taking the following security and market factors into account: (1) the extent of world-wide production and number of computer units in the field; size, portability and cost; (2) ease of upgrades to higher performance levels without manufacturer support; (3) availability of commodity high performance communications links to network together individual computers; and (4) increasing parallelization of important national security applications.

We concluded at that time that workstations up to 7,000 MTOPS were becoming widely available, and therefore uncontrollable, from a variety of sources and that this level of computing performance could be obtained relatively easily worldwide by a variety of means -- e.g., by buying lower level workstations and upgrading them, or by buying the components and piecing together a high performance computer.

For example, lower level workstations (2 to 4 processors) that are readily available through commercial distribution networks globally can be upgraded easily with uncontrollable plug-in boards to create systems with 64 or more processors that are capable or achieving performance levels in the 7,000 MTOPS range. Numerous foreign manufacturers can purchase microprocessors and other components and build relatively sophisticated machines. Microprocessor performance has more than quadrupled in the past five years -- from 80 MTOPS in 1993 to over 450 MTOPS today. And, new developments in high speed communications -- for example, High Performance Parallel Interconnect (HIPPI) and Fiber Distributed Data Interconnect (FDDI) switches -- provide the means to connect many lower-level workstations together to achieve performance levels above the 7,000 MTOPS range. In fact, many DOD research labs are now using such commercially available workstations and interconnect technology to develop and test weapon systems.

Our findings on national security applications and technology trends were the basis for the consensus USG decision to establish the export control levels on high performance computers that we have in place today. The Department continues to fully support these control levels because we believe they are technically realistic while also meeting our national security and nonproliferation objectives. Specifically, the control methodology is designed to serve two purposes:

  • First, it permits the government to calibrate control levels and licensing conditions to the national security or proliferation risks posed by a specific destination or end-user. The most stringent controls and licensing conditions are focused on countries that pose significant proliferation or other security risks. We have the ability to deny access to U.S.-origin computing power needed for critical military applications to such countries. ( I should note that we continue to maintain our controls under the Enhanced Proliferation Control Initiative under which exporters who know they are selling to end users involved in proliferation projects must obtain a license for any level of computer.)
  • Second, it enhances U.S. national security and preserve the U.S. computer industrial base for defense by ensuring that legitimate computer exports are not impeded.

Finally, Mr. Chairman, I would like to comment on legislative language on high performance computer export controls in the Defense Authorization Conference Bill. I particularly question the advisability of a 180 day Congressional notification period for any export control policy changes in such a fast-moving technology sector as computers. I am a firm supporter of appropriate Congressional consultation on matters of national security. Nevertheless, I would hope that a more reasonable time period of 30 days would suit both your interests and those of the Administration, because that would provide opportunity for appropriate Congressional oversight while allowing simultaneously for changes to keep pace with advancing technology.

I want to underscore the continued commitment of the Department of Defense to strong and effective export controls on high performance computers and other militarily sensitive technologies. However, the approach taken in the Defense Authorization Conference Bill would significantly impair the Presidentís flexibility -- and that of DoD -- to ensure that export control policies and procedures are implemented in a manner that protects our national security interests without damaging the continued economic viability of the U.S. computer industry. I remain committed to working with Congress to address these issues in a manner that maintains the flexibility we need to balance these important national interests.