Leaderless Replication

Introduction

Single-leader and multileader replication strategies involve a replica called a leader that receives writes and sequences the write for the rest of the replicas. The followers then apply the write requests in the same order dictated by the leader. Leaderless replication is a special class of replication strategy that removes the concept of a leader and treats every replica in the database uniformly.

Amazon published a research paper on the characteristics of a distributed leaderless database in, “Dynamo: Amazon’s Highly Available Key-Value Store.” This paper presents the design and implementation of Dynamo, a highly available key-value storage system that runs some of Amazon’s core services. Some of the important aspects of the paper include:

• Partition algorithm

• Replication

• Data versioning

• Handling failures through hinted handoff

• Membership and failure detection

Databases that adopt leaderless replication are also called Dynamo-style databases. Examples of databases that adopt leaderless replication include Cassandra and Voldemort. ...