Analysis of the effects of soft errors on compression algorithms through fault injection inside program variables

摘要

Data logging applications, such as those deployed in satellite launchers to acquire telemetry data, may require compression algorithms to cope with large amounts of data as well as limited storage and communication capabilities. When commercial-off-the-shelf hardware components are used to implement such applications, radiation-induced soft errors may occur, especially during the last stages of the launcher cruise, potentially affecting the algorithm execution. The purpose of this work is to analyze two compression algorithms using fault injection to evaluate their robustness against soft errors. The main contribution of the work is the analysis of the compression algorithm susceptibility by attacking their data structures (also referred as program variables) rather than the memory elements of the computing platform in charge of the algorithm execution. This approach is agnostic of the downstream implementation details. Instead, the intrinsic robustness of compression algorithms can be evaluated quickly, and system-level decisions can be taken before the computing platform is finalized.