Dependable Multiprocessor

Toward the first supercomputer in space

Sponsor: NASA/JPL


Project title: Dependable Multiprocessor (DM), previously known as the environmentally adaptive fault-tolerant computer (EAFTC)

Abstract - As satellites and other space platforms continue to evolve, mission requirements for earth and space science expand with more powerful sensors featuring enhanced resolution, which quickly leads to in-situ processing demands far in excess of conventional embedded architectures. In partnership with Honeywell Space in Clearwater FL, researchers at the University of Florida are conducting a comprehensive research project to investigate and develop for NASA the first supercomputer in space. A broad range of basic and applied research challenges in computer engineering must be overcome to achieve a dependable, versatile, and powerful system architecture that can adapt to the harsh environment of space while supporting high-end applications in signal, image, and communications processing.

Space-based radar (SBR), multi-spectral imaging, and other high-performance applications are creating orders of magnitude more data than limited-bandwidth satellite downlinks can support. As such, mission planners are demanding advanced space computing with more onboard payload processing to improve autonomous operations and reduce the amount of data sent over downlinks. Future platforms will need to be inexpensive, flexible, scalable, and able to support reduced development schedules while saving size, weight, power, etc. In addressing these challenges, Honeywell Space and the University of Florida are developing for NASA an advanced new satellite payload architecture that features general-purpose processors, memory, reconfigurable FPGA resources, and high-speed networking for spaceborne high-performance computing (HPC) with both fault-tolerant infrastructure and hardware-reconfigurable processing, among other components. The future goal for this platform is to allow scientists to migrate parallel application code developed on ground-based HPC systems to a space platform with dozens of computational nodes in a seamless fashion.

This project focuses on key research tasks associated with technology development efforts for the Space Technology 8 (ST-8) mission of the New Millennium Program (NMP) at NASA. The ST-8 mission is slated for launch in February 2009 and will feature a payload for advanced technology. As cited by NASA administration in their 1/28/05 announcement, "these technological capabilities will provide orders of magnitude in performance compared to the state-of-the-art technologies used in NASA satellites. Future NASA science and exploration missions will benefit greatly from these technological capabilities. The ST-8 project will effectively develop these technologies from the early stages to flight readiness, and then validate them in space prior to using them in NASA sponsored scientific missions." Research and development results in this project will ultimately culminate in a successful system launch and mission, establishing the first supercomputer in space.

Several facets of research are the primary emphases in this project, all in terms of new methods and from them new software and hardware infrastructure for space. The first is fault-tolerant computing, so that commodity processing subsystems (i.e. commercial off-the-shelf, COTS) with their many performance and cost advantages can be effectively deployed in the harsh environment of space. The second is parallel computing, so that these subsystems can work concurrently and achieve the performance and usability levels required of a space-based, high-performance computer system. The third is reconfigurable computing, so that maximum efficiency in performance and electrical power can be realized with a highly adaptable system by leveraging new technologies in field-programmable hardware.


Research Thrust Areas at UF for 2006-2008 Research Thrust Areas at UF in 2005-2006 (TRL5: baseline DMM for space)
Related Links and Materials