History-based weighted average voter: a novel software voting algorithm for fault-tolerant computer systems

Abstract

Voting algorithms have been widely used in the realisation of fault-tolerant systems. We propose a new software voting algorithm which uses the history record of redundant modules to compute the final output. Two versions for the novel algorithm are introduced In the first version any module result is assigned to a weighting value such that module results with higher history record value, are assigned to a higher weighing value than those with lower history record value. In the second version of the novel voter, those module results which have a history record value, less than the average record value, are allocated a weight of zero and removed from the contribution toward the voter output. Furthermore, a novel method for creation of a history record of modules is proposed. Empirical results show that both versions of the novel voter give higher safety performance than the Standard Weighted Average voter with permanent and transient errors.