The five initial sites using the production grid are Space and Missile Defense
Command in Alabama, Tank-Automotive Research, Development and Engineering Center
in Michigan, Naval Research Laboratory in Washington D.C., Naval Oceanographic
Office in Mississippi and White Sands Missile Range in New Mexico.
The production grid is founded on Platform's LSF MultiCluster, which the HPCMP
selected as the basis for its meta-queuing initiative. LSF MultiCluster is
software that controls all computing resources and coordinates jobs across
multiple geographic locations.
"Our production grid is moving forward because of the features and capabilities
of LSF MultiCluster. It's a fundamentally new and better way to deliver high
performance computing services to our researchers and engineers through a
uniform queuing environment," said William Reidy, a contractor supporting the
HPCMP. "After just a few months, we're seeing dramatically increased
accessibility and will ultimately be able to provide 1.1 million supercomputer
processor hours per year. Our hardware and software utilization is more
effective, and we are able to recover cycles that would otherwise go unused."
The production grid also utilizes Platform FTA (File Transfer Agent) to ensure
reliability when transferring files between locations. Platform's FTA and LSF
both operate in heterogeneous environments -- a critical aspect of production
grids.
The High Performance Computing Modernization Program delivers world-class
commercial, high-end, high-performance computational capability to the
Department of Defense and facilitates the rapid application of advanced
technology into superior warfighting capabilities. With 21 centers located in 14
states and the District of Columbia, the program is creating a pervasive culture
among DoD scientists and engineers where they routinely use advanced
computational environments to address the DoD's most complex and challenging
problems. More information on the DoD HPCMP can be found on line at http://www.hpcmo.hpc.mil/.