Demonstration of the remote exploration and experimentation (REE) fault-tolerant parallel-processing supercomputer for spacecraft onborad scientific data processing

This paper is the written explanation for a demonstration of the REE Project's work to-date. The demonstration is intended to simulate an REE system that might exist on a Mars Rover, consisting of multiple COTS processors, a COTS network, a COTS node-level operating system, REE middleware, and an REE application. The specific application performs texture processing of images. It was chosen as a building block of automated geological processing that will eventually be used for both navigation and data processing. Because the COTS hardware is not radiation hardened, SEU-induced soft errors will occur. These errors are simulated in the demonstration by use of a software-implemented fault-injector, and are injected at a rate much higher than is realistic for the sake of viewer interest. Both the application and the middleware contain mechanisms for both detection of and recovery from these faults, and these mechanisms are tested by this very high fault-rate. The consequence of the REE system being able to tolerate this fault rate while continuing to process data is that the system will easily be able to handle the true fault rate.