[Previous] | [Session 16] | [Next]
R. Sengupta, J.D. Offenberg, D. Fixsen (Raytheon ITSS), J.C. Mather (GSFC), R. Hanisch, M. Nieto-Santisteban (STScI), D. Katz, P. Springer (JPL)
An alternative to using expensive, high power, radiation hardened processors, or special purpose architectures, is to use commercial-off-the-shelf (COTS) scalable, low power, fault tolerant, high performance computation in space. New spaceborne applications would run in embedded high-performance computers and return analysis results to earth; not just raw data. Most of the faults caused by the radiation environment of the regions of space for which REE (Remote Exploration and Experimentation) is developing a system (Deep Space, Low Earth Orbit) are transient, Single Event Upsets (SEUs). Some of these faults can cause errors at different application levels. System and applications software can potentially detect and correct some or many of these errors. Here we discuss different software fault tolerance approaches such as replication, voting and masking with a focus on Algorithm-Based Fault Tolerance (ABFT). Combined approaches of software and hardware like fault avoidance, redundancy and reconfiguration are discussed. These approaches would show the tradeoffs between reliability, power, cost and computation power for spacecraft in a low-to-moderate radiation environment. This task is funded by NASA's REE project under Dr. Robert Ferraro.