Objective: To obtain access for ESS NRA Investigations to large scalable testbed systems which have the potential to scale to teraFLOPS performance.
Approach: ESS provides access to ESS/GSFC and ESS/JPL scalable parallel systems, and establishes agreements with CAS centers and non-NASA research labs which own large scalable systems in order to acquire additional machine time. This approach leverages on the substantial capital investments already made by other organizations and also provides Investigators with the use of a broader variety and larger machines than NASA could afford to purchase. As computer cycles on remote systems are obtained, ESS testbed managers work with the remote systems' administrators to establish working arrangements and to facilitate system access for ESS Investigators.
Accomplishments: ESS contributed 33% toward the purchase of a 256-node Cray T3D. The system was delivered and installed at JPL, and ESS Investigators began using the system. ESS Investigators used the ESS-owned 64-node Intel Paragon located at JPL. ESS purchased and configured a 4-node MasPar cluster, with an aggregate total of 32,768 processors. The cluster is located at GSFC. ESS purchased and accepted a 16-processor Convex Exemplar SPP-1000, also located at GSFC. Several Investigators have begun using the machine. ESS received and used a percentage of NASA's allotment of time on the Intel Touchstone Delta at Caltech. ESS received an allotment of time on the CAS-owned Thinking Machines Corporation (TMC) CM-5 located at ARC. Investigator teams (composed entirely of U.S. citizens) used the CM-5 allotment. ESS Investigators are using the TMC CM-5 at NRL. An NDPR was used to pay for the usage. To contain costs, ESS acquired accounts on the 32-node TMC CM-5 at U. Maryland for code development. ESS Investigators received accounts and began using the ANL IBM SP-1 (under the NASA allocation). Several ESS Investigators received accounts on the U. Washington KSR-1, offered on a limited basis by PI-G.Lake. ESS provided a member to the Technical Evaluation Committee for the ARC Cooperative Research Announcement (CRA) leading to the placement and availability of an IBM SP-2 at ARC. ESS has received an allocation and Investigators will begin usage in FY95.
Significance: Exposing the Investigations to a wide variety of scalable systems assists the project to rule out weak contenders for objective reasons consistent with ESS requirements. It also enhances the Investigators' chances of success in solving their Grand Challenge problems by increasing the likelihood that they can find an architecture well-matched to their problem. The larger sizes of these shared machines allows larger problems to be run, a factor which aids the system vendors by allowing the Investigations to test closer to the maximum potential of the machines and uncover weaknesses there, thereby spurring further development of the largest systems and hastening the ultimate construction of a teraFLOPS system.
Status/Plans: In FY95, ESS will release, evaluate proposals and award a Cooperative Agreement Notice to bring a large testbed to GSFC in FY96. ESS will continue to arrange for access to large testbeds for Investigators to test new architectures and architecture scalability.
Point of Contact:
Return to the PREVIOUS PAGE