Correspondent computing for software implementation fault tolerance
P.-N. Lee, J. Blankenship · 2002
Correspondent computing, which achieves diversity through correspondent operations and uses a decision algorithm which is based on correspondence checking rather than majority voting or assertion checking, is proposed. The significance of this is that multiple diverse software versions are produced from a single algorithm and the assertions used in the correspondence test are more exact and easier to formulate. The ability of the correspondent computing method to generate correspondent versions systematically is further enhanced by the introduction of 'partial' reciprocal operations. Their role in creating reciprocal correspondent program modules tolerant to software implementation faults is explored in an actual program which implements a well-known triangle algorithm.>