Design and Evaluation of Self-Checking Single-Version Software Fault Tolerance for Non-Fail-Stop Transient Faults

Transient hardware faults increasingly threaten modern computing systems due to hardware miniaturization and environmental radiation. Traditional fault tolerance relies heavily on multi-version software or hardware redundancy. However, these techniques impose severe resource penalties. This paper presents a comprehensive framework for single-version software fault tolerance, synthesizing and advancing the foundational paradigms introduced by research [3-7]. We introduce a unified, self-checking, and self-protective single-version program architecture designed for non-fail-stop execution environments. By integrating instruction-level and data-level redundancy, adaptive checkpoints, semantic assertions, and program-flow monitoring, our mechanism masks transient faults without secondary code versions. Experimental evaluation via software-implemented fault injection demonstrates a high fault-coverage rate with low performance and memory overhead.