Fault-tolerant naming and mutual exclusion

Abstract

We present solutions to two classical problems concerning distributed systems in which some sites or processes can possibly have byzantine faulty behavior. We first study the naming problem (how to give each site of a network an unique identifier). We are naturally led to make some supplementary assumptions about the synchrony of message passing, the connectivity of the underlying graph and the existence of a special site, provided with a digital signature, for initiating the protocol. The solution that we present uses three waves of messages between the initiator and any other site. Then, we solve the mutual exclusion problem with particular assumptions about the behavior of byzantine processes. The solution implements each critical section as a separate segment, whose address (necessary to access it) is “hidden”. A process must reconstruct this address before entering its critical section, involving the cooperation of a number of other processes. For the two problems, protocols are given and their complexity is estimated.