Circumventing the t < n/2 Read/Write Register Impossibility: the Failure Detector Approach

Abstract

This chapter presents the failure detector class (denoted Σ) that allows us to circumvent the impossibility of building an atomic read/write register in an asynchronous message-passing system in which half or more processes may commit crash failures (system model CAMP_n,t[t ≥ n/2]). (The reader is referred to Section 3.3 for formal definitions related to failure detectors.) This chapter first introduces the class Σ, and shows how it allows us to implement an atomic register for any value of t. Then, it shows that Σ is the failure detector class that provides us with the weakest information on failures that allows an atomic read/write register to be built despite asynchrony and any number of process crashes.