Leader Election Algorithm in Fault Tolerant Distributed System

Abstract

Reliability and availability are two important aspects of a fault-tolerant distributed system. A replicated database system helps us to improve the availability of a distributed system. In order to maintain the reliability, we need to maintain the consistency among replicas. Consensus protocols are used to achieve coordination among replicas, and leader plays pivotal role in these consensus algorithms. Leader works as a powerful tool to gain consistency and to reduce the overhead in maintaining the coherency among replicas. The leader election process requires consensus among sites to elect one site as a leader of the system. In this article, we present the formal development of leader-based election system in Event-B by proposing an algorithm, leader election using ordered delivery (LEOD). This paper highlights the importance of message ordering at delivery end in designing the algorithm for leader election in distributed systems. We conclude this paper by giving specifications of algorithm in Event-B using RODIN platform.