Classes of Byzantine Fault-Tolerant Algorithms for Dependable Distributed Systems

This thesis concentrates on the design of new algorithms for fault-tolerant systems based on system-level hardware masking redundancy. It is argued that any system in which a reliability improvement of at least a factor 100 is required should be based on system-level hardware masking redundancy. The technique of system-level hardware masking redundancy is applicable in a redundant system consisting of a number of processors, in which the system services are replicated on the different processors, and provides resilience to a limited number of faulty processors in the system. The technique is most effective in a distributed system, since the autonomous nature and geographical distribution of the processors in such a system largely contribute to achieve independency between failures of different processors, which improves the reliability of the system