Workshop on Enabling Technologies for Peta(FL)OPS Computing

Objective: Conduct and produce the first comprehensive assessment of the field of Peta(FL)ops computing systems. Establish a baseline of understanding of its opportunities. challenges, and critical elements. Set near-term research directions to reduce uncertainty and enhance our knowledge of this field that will lead to U.S. preeminence in developing, producing, and applying Peta(FL)ops-scale computing systems.

Approach: Conduct intensive 3-day workshop to focus key members of the HPC community on critical issues driving feasibility of Petaflops computing. Establish working groups in areas of applications and algorithms, device technology, parallel architectures and system structures, and system software and tools.

Accomplishments: Workshop conducted February 22-24, 1994 at Doubletree Hotel, Pasadena, CA. More than 60 invited contributors from industry, academia, and government participated with sponsorship from NASA, DOE, NSF, NSA BMDO ARPA.

The workshop findings 1) identified applications of economic, scientific, and societal importance requiring Petaflops scale computing, 2) determined technical barriers to achieving effective Petaflops computing, 3) revealed enabling technologies critical to implementation of Petaflops computers, 4) derived research issues central to realizing Petaflops capability, and 5) set research agenda for near-term work focused on immediate questions confronting the field.

Significance: The findings of this workshop established the field of Petaflops computing as a legitimate domain of inquiry and determined the feasibility of such systems within the next two dozen years. It revealed the importance of the integration of disparate device technologies including optical and superconducting as well as advanced semiconductor. The workshop identified many applications requiring Petaflops capability that will be important to US science and industrial needs and revealed dominant resource scaling characteristics determining system cost and reliability.

Status/Plans: The report is to be published as book by MIT Press. A workshop entitled "The Petaflops Frontier" is being organized at Frontiers of Massively Parallel Computing '95 Conference.


Point of Contact:


Return to the PREVIOUS PAGE



curator: Larry Picha