SGI announced it has signed a "memorandum of understanding" with NASA to collaborate on advanced supercomputer technology development. The agreement covers "cooperative research and development" in areas such as computational modeling for astrobiology and virtual surgery for long-duration space missions, SGI said.
SGI and NASA will also work together to develop ultra-high-speed computers which will be used to design next-generation spacecraft, said Daniel Goldin, NASA administrator, in a statement.
"This agreement is of great importance to key NASA projects such as astrobiology, which will look at fundamental human questions concerning the origin and evolution of life," Goldin added.
"This collaboration will also assist SGI in defining systems appropriate to commercially viable high-performance computing," said Richard E. Belluzzo, chairman and chief executive officer, SGI in a statement.
SGI has already done joint work with NASA in deploying an SGI Origin system which contains 256 processors at NASA Ames.
Specific proposed areas of cooperative research and development between SGI and NASA include:
The development of
large-scale supercomputing applications and support tools optimized for
The development of operating systems technology that increases the utilization of these large systems.
Techniques and features to minimize the impact of component failure.
The RS/6000 SP supercomputer, the same architecture used for IBM's chess-playing "Deep Blue" computer, will help researchers at this center to study areas including Alzheimer's Disease, life-threatening arrhythmias (irregular heartbeat), and the atomic structure of materials leading to the discovery of new materials.
Researchers at North Carolina universities throughout the state will have access to the $20 million system, an upgraded and much larger version of the RS/6000 SP Deep Blue, IBM said.
Delivery of the system, running IBM's "AIX" Unix operating system, is scheduled for the fourth quarter of 1999.
Initially, the system will consist of 320 IBM POWER3 microprocessors connected over a dedicated high-speed network. It will have 1.45 terabytes (trillion bytes) of storage, and peak performance of a quarter teraflop. When completed in late 2000 the system will achieve one teraflop performance with 640 POWER3-II processors and connected with new high-speed technology, IBM said.