JNNIE: the Joint NSF-NASA Initiative in Evaluation

Objective: Evaluate the effectiveness of Scalable Parallel Computing (SPC) systems across a very broad range of scientific and engineering applications, identify the nature and degree of factors which impede efficiency, and determine users' accessibility to SPC systems for applications programming.

Approach: Establish a collaboration between NSF and NASA computational centers to achieve a critical mass through the sharing of test SPC systems, talent in parallel computation, and parallel application codes. Select parallel test applications and port to selected SPC test systems. Determine performance with respect to system size and problem scale. Measure execution overhead and idle times to expose sources of performance degradation. Capture user experiences regarding ease of use of SPC hardware and software systems for achieving useful scientific computations. (see picture, 120K).

Accomplishments:

• Collaboration between NSF and NASA established with key computational centers represented.
• SPC test systems and 22 major scientific and engineering parallel application codes identified.
• Performance and scalability experiments performed on all test applications and systems.
• Instruction mixes, execution overhead, and idle times measured.
• Two surveys conducted on issues associated with ease-of-use and experiences of computational scientists documented.
• Panel session held at Supercomputing '94.

Significance: JNNIE results reveal important trends and characteristics of SPC technology and demonstrate the potential and challenges in exploiting this capability for scientific and engineering applications. These results and conclusions will influence future hardware and software system designers in advancing the quality and utility of this important class of high performance computer systems.

Status/Plans: JNNIE experimental work due for completion at end of calendar year. All major objectives have been achieved. Final report first draft will be completed by March, 1995 and final form in June, 1995. Methods developed for this effort will be applied to future NASA evaluation work and results will be considered for future procurements of SPC testbeds.

Point of Contact:

Dr. Thomas Sterling
Center of Excellence in Space Data and Information Sciences (CESDIS)
Goddard Space Flight Center/Code 930.5
tron@chesapeake.gsfc.nasa.gov
(301) 286-2757

Return to the PREVIOUS PAGE