High Performance Computing (HPC) Innovation Service Portal Pilots Cloud Computing (HPC-ISP Pilot Cloud Computing)

摘要

The objective of part 1 of the research effort was to perform three capability demonstrations on a brokerage infrastructure in a commercial cloud computing environment, to engage multiple computing service providers and application software vendors and collect comparative metrics to evaluate the return-on-investment (ROI) argument for high performance computing within the DoD supply chain and to evaluate the startup costs saved of using a broker. The demos were: (1) AltaSim Ghost on Ohio Supercomputer Center, with an estimated ROI of 252X and brokerage startup cost savings of 90% and ten months of time; (2) ANSYS on R Systems, with an estimated ROI of 14.7X and brokerage startup costs savings of 91% and seven weeks of time; (3) Mathematica and Blender on Amazon EC2, with an estimated ROI of 27.4X and brokerage startup cost savings of 96% and one month of time. The objective of part 2 of this research effort was to investigate the use of probabilistic computer architectures to solve computational challenge problems at orders of magnitude lower cost and increased processing capability relative to conventional computer architectures. 1000x+ speedups with 10-30x less power usage using a software- reprogrammable probabilistic video processor were documented.