The NASA New Millennium Program (NMP) Space Technology 8 (ST8) Dependable Multiprocessor (DM ) project is the current incarnation of the long-held DoD and NASA desire to migrate high-performance COTS supercomputing to space. Many DM system concepts are based on related technologies developed and demonstrated over the past three decades on several DARPA (Defense Advanced Research Projects Agency), NASA, and DoD programs including Space Touchstone, Remote Exploration and Exploration (REE), and Improved Space Architecture Concept (ISAC). The Dependable Multiprocessor (DM) is a processing system embodying an architecture and a methodology which enable COTS-based, high performance, scalable multi-computer systems to operate in space environments. DM technology encompasses fault tolerance middleware which manages a cluster of high performance COTS processors while providing enhanced SEU-tolerance, and supports MPI (Message Passing Interface) parallel and distributed processing. The goals of the DM project were to develop technology that is transparent to the underling platform, able to support upgrades in both hardware and software, and is easy to port applications from the laboratory to space. Over the past 4 years, DM has been demonstrated to be technology-independent and easy to use, not only for parallel and distributed applications, but also for user-selectable SEU-tolerant modes of operation. DM technology has been successfully demonstrated with heterogeneous COTS processors, heterogeneous COTS operating systems (VxWorks and Linux), several state-of the-art multicore processors, and FPGAs. The high speed interconnect between COTS data processing nodes in the DM testbeds is TCP/IP over Ethernet, but the physical interconnect medium could also be parallel or serial RapidIO or wireless. This presentation provides an overview of DM technology and the status of DM's TRL6 technology validation.
Document date April 16, 2009.