From failure detectors with limited scope accuracy to system-wide leadership

Abstract

A failure detector is a device that provides the processes with information on failures. The accuracy property of a failure detector defines the type of mistakes it is not allowed to make. The limited scope of the accuracy property restricts it to only a part of the system. /spl diams/S/sub k/ is a class of unreliable failure detectors with a limited scope accuracy. Eventually each process that crashes is suspected by every correct process, and there is a time after which some correct process is never suspected by only k processes. An eventual leader facility (usually denoted /spl Omega/)is a device that eventually provides all the processes with the identity of one of them that is correct. Such a facility is used as a basic service in a lot of fault-tolerant distributed protocols (e.g., asynchronous consensus protocols). This paper proposes a protocol that builds an eventual leader service from any unreliable failure detector of the class /spl diams/S/sub t+1/ where t is the maximum number of processes that can crash during a run. The fact that /spl diams/S/sub t+1/ is easier to build than /spl diams/S or /spl Omega/ and the design simplicity of the proposed protocol makes it attractive.