Election-Quorum-Based Coordinator Election Algorithm for Distributed Systems

Abstract

Today mostly all applications are designed to be distributed. The main intention behind it is to avoid losses due to the failure of the central entity in centralized systems. The key element in distributed systems is coordinator or leader entity. Coordinator plays critical role in distributed systems, specially for performing tasks like simplifying process synchronization and load balancing. Hence electing one of the entities as a coordinator is an important problem. But in practice, coordinator election process in distributed systems is vulnerable to faults and failures hence it is necessary to design a coordinator election algorithm to tolerate faults and failures. In this paper, the notion of fault tolerant Election Quorum (further termed as ‘EQuorum’) method for coordinator election is introduced. The challenges in the existing algorithms are identified and proposed an EQuorum-based fault tolerant coordinator election algorithm. This algorithm can tolerate failure of communication links and entities. The number of messages, time to terminate algorithm and failover time of the proposed algorithm are identified, evaluated and analysed. Thus, this paper proposes a fault-tolerant coordinator election algorithm for partially synchronous distributed systems which outperforms in message and time performance measures. It is observed that EQuorum outperforms as compared to existing election algorithms and provides additional feature of fault tolerance.