A new Heuristic to Discriminate Transient from Intermittent Faults

Effective discrimination between transient and permanent faults is a very important practical problem in (dependable) system design. A count-andthreshold mechanism named a-count, designed to discriminate between transient faults and intermittent faults in computing systems, is presented in an enhanced embodiment. It retains enough simplicity to allow exhaustive analysability through simple models. It is shown that the introduction of two operating thresholds, instead of the single one present in the basic scheme already known, both improves the performance figures of the mechanism and eases the designer's task of tuning the internal parameters.