Distributed Algorithms

Chapter 6 Distributed Consensus with Process Failures

In this chapter we continue the study of consensus problems in the synchronous model, which we began in Chapter 5. This time, we consider the case where processes, but not links, may fail. Of course, it is more sensible to talk about failure of physical “processors” than of logical “processes,” but to stay consistent with the terminology elsewhere in the book, we use the term process. We investigate two failure models: the stopping failure model, where processes may simply stop without warning, and the Byzantine failure model, where faulty processes may exhibit completely unconstrained behavior. Stopping failures are intended to model unpredictable processor crashes. Byzantine failures are ...