Strategies for fault-tolerant, space-based computing: Lessons learned from the ARGOS testbed

Abstract

The Advanced Space Computing and Autonomy Testbed on the ARGOS Satellite provides the first direct, on orbit comparison of a modern radiation hardened 32 bit processor with a similar COTS processor. This investigation was motivated by the need for higher capability computers for space flight use than could be met with available radiation hardened components. The use of COTS devices for space applications has been suggested to accelerate the development cycle and produce cost effective systems. Software-implemented corrections of radiationinduced SEUs (SIHFT) can provide low-cost solutions for enhancing the reliability of these systems. We have flown two 32-bit single board computers (SBCs) onboard the ARGOS spacecraft. One is full COTS, while the other is RAD-hard. The COTS board has an order of magnitude higher computational throughput than the RAD-hard board, offseting the performance overhead of the SIHFT techniques used on the COTS board while consuming less power.

Tolerating Soft Errors in Processor Cores Using CLEAR (Cross-Layer Exploration for Architecting Resilience)

“…Additionally, certain DUEs, such as those which may cause a program crash, will not be detectable by CFCSS, or other software techniques, since execution may abort before a corresponding software check can be triggered. The relatively low resilience improvement of CFCSS has been corroborated in actual systems as well [52].…”

“…Additionally, certain DUEs, such as those which may cause a program crash, will not be detectable by CFCSS, or other software techniques, since execution may abort before a corresponding software check can be triggered. The relatively low resilience improvement of CFCSS has been corroborated in actual systems as well [52].Error Detection by Duplicated Instructions (EDDI) provides instruction redundant execution via compiler modification [53]. We utilize EDDI with store-readback [54] to maximize coverage by ensuring that values are written correctly.…”

Plug-and-play design approach to smart harness for modular small satellites

“…Most software-based approaches are aimed at detecting faults, some of them apply redundancy to high-level source code by means of automatic transformation rules, whereas some others use instruction redundancy at the low level (assembly code) to reduce the code overhead and performance degradation, and improve the detection rates [22][23][24][25][26][27]. The AraMiS software was developed using a highlevel software radiation-hardening technique completely developed in house.…”

Survey and future directions of fault-tolerant distributed computing on board spacecraft

“…Redundant execution can be done at instruction level or task level. In instruction level redundant execution [54], each instruction of the executed program is duplicated and, after each duplicated instruction, results are compared for errors. On the other hand, in task level redundant execution, a software task is executed twice or more in time to avoid temporary faults.…”