BYZANTINE FAULT-TOLERANCE THROUGH APPLICATIONS ORIENTED SPECIFICATION.
McMillin, Bruce M., Ni, Lionel M. · Rare & Special e-Zone (The Hong Kong University of Science and Technology) · 1987
In parallel-processing systems, fault tolerance may be achieved through the use of fault-tolerant algorithms which make use of implicit redundancy in the data. The authors introduce a class of predicates, D, useful as metrics for the purpose of fault identification in fault-tolerant algorithms. Byzantine Agreement is used both to mask failures and to provide, by means of D, a reliable (consistent) basis for distributed identification of nonmaskable failures. Byzantine Agreement is chosen over PMC (Preparata-Metze-Chein)-based techniques, which do not provide a complete diagnostic basis for the Byzantine fault case. The coupling of Byzantine Agreement achieves a high-level application-oriented abstraction for the fault detection problem.