As an agency, NASA invests a significant amount of resources in the development and advancement of highend computing (HEC) resources and associated technologies to support all of its missions. The demand for HEC resources has increased dramatically in the last several years and continues to grow in response to progressively challenging modeling and simulation requirements.

NASA’s need for more HEC resources extends beyond computing cycles. With a wide range of projects to support all four agency missions — from aeronautics and exploration, to science and space operations — NASA’s HEC resources must be flexible and robust enough to handle different types of compute jobs, and able to accommodate a dynamic environment with constantly changing needs, depending on mission priority and status.

A birds-eye view of the 10,240-processor SGI Altix Columbia supercomputer. (NASA Ames/Tom Trower
For instance, one of the Exploration Systems Mission Directorate’s primary focuses is development and testing of the Orion crew exploration vehicle, which will take human explorers back to the Moon by 2020, replacing the retiring Space Shuttle fleet. This type of design and engineering work is very compute-intensive, requiring multiple job runs to be turned around quickly. Shuttle-related calculations under NASA’s Space Operations Mission Directorate are similar in nature, with a very real time-critical element. In contrast, applications within NASA’s Science Mission Directorate typically involve long-running calculations with many processors; for example, modeling the formation and life of a hurricane over several days, or decadal global weather patterns.

The Aeronautics Research Mission Directorate (ARMD) uses HEC resources for a wide range of simulation tasks. Modeling of complex liquid sprays, chemical kinetics, and formation of chemical species is critical to the design of combustors for fuel-efficient, cleanburning engines ranging from those used on commercial transport aircraft to those being studied for hypersonic airbreathing vehicles for access to space. Distributed computing architectures are also required for the multidisciplinary design and optimization of future aircraft and their propulsion systems.

The NASA Advanced Supercomput-ing (NAS) facility at Ames Research Center currently houses a 14,336-core SGI Altix system named Columbia, which supports more than 1,100 users from all four NASA mission directorates delivering 89 Teraflop/s (TF). One of NAS’ primary charters is selecting and deploying computing architectures such as Columbia that will meet the needs of all missions. NAS continually explores emerging, innovative architectures and systems and determines how to strategically leverage them to provide the most effective computing platforms and environments. Among its most recent acquisitions is a 245-TF SGI ICE cluster named Pleiades, which will provide more than 2.5 times the current high-end computing capability for NASA scientists and engineers. This architecture takes advantage of the advances made in the marketplace by utilizing the Intel Xeon quad-core processor in a highly reliable interconnected system spanning 40 compute cabinets.

Technical discussions must continue taking place regarding leading-edge concepts such as petascale systems, and NASA must be willing to openly explore emerging computing architectures to ensure the right HEC resources are in place to carry out future mission challenges. Near-real-time aerospace design and humans living and working in space are just around the corner.

For more information, or to tell us about your ideas, contact William Thigpen, NASA Ames, at 650-604-1061, or contact the NAS Division Office at 650-604-4502.