Analysis of Reliability and Structural Complexity for Various Implementations of a Finite State Machine Resistant to Soft Failures
Abstract
Introduction: Up-to-date design rules used in computer engineering make hardware unreliable when working under radiation. A hit of a charged particle causes a “soft failure” - a situation when hardware elements remain in usable condition but the information transmitted or stored in the memory is corrupted. We need to develop new circuitry solutions which would increase the resistance of hardware (especially finite states machines) to soft failures. Purpose: The estimation of reliability characteristics for various redundant structures of a Moore automaton in case of a flow of soft failures. Results: The comparison between the known Moore automaton prototypes has shown that redundant internal memory and self-recovery of corrupted data in the memory allow us to significantly increase the duration of trouble-free functioning, without making the implementation more sophisticated as compared to non-recoverable structures. In case of a soft failure, a memory bit remains functional, and the valid state of the system can be restored by overwriting the corrupted bit with a valid one when the recovery is over. An unrecoverable failure can occur only if several instances of redundant units of the memory are corrupted simultaneously during a singe recovery period. If we manage to ensure a sufficiently small recovery period in the system, it will dramatically prolong its non-failure operating time. Practical relevance: The obtained results allow you to determine the ways of developing fault-tolerant finite state machines resistant to soft failures: blocking the propagation of a soft failure in the memory, periodic self-recovery of corrupted bits, and introducing additional hardware to detect and register failures.Published
2017-06-21
How to Cite
Egorov, I., & Melekhin, V. (2017). Analysis of Reliability and Structural Complexity for Various Implementations of a Finite State Machine Resistant to Soft Failures. Information and Control Systems, (3), 34-46. https://doi.org/10.15217/issn1684-8853.2017.3.34
Issue
Section
Information and control systems