High-Performance Computing: Advanced Research Projects Agency Should Do More to Foster Program Goals

Citation
Government Accountability Office, High-Performance Computing: Advanced Research Projects Agency Should Do More to Foster Program Goals (IMTEC-93-24) (May 17, 1993) (full-text).

Overview
Pursuant to a congressional request, GAO reviewed the Advanced Research Projects Agency's (ARPA) High Performance Computing (HPC) Program, focusing on: (1) ARPA distribution of advanced computers to research sites; (2) ARPA interaction with the research community; (3) the balance between HPC software and hardware investments; and (4) the barriers that impede technological progress and prevent ARPA from achieving program goals.

GAO found that: (1) although ARPA actively involves itself with the research community by providing researchers with quick access to early prototypes of massively parallel processing machines (MPP), ARPA placement of the computers at research sites has been unjustifiably biased toward two contractors; (2) HPC technological progress has been slow due to limited ARPA interaction with the research community; (3) ARPA does not provide adequate research guidance or contractor selection information, has not set research goals, does not publish detailed summaries or progress reports, and fails to provide researchers with performance data; (4) the ARPA goal of increasing computing power by 1996 is in jeopardy, since HPC software development has not kept pace with hardware development; and (5) ARPA needs to increase its focus on HPC system software development.