Avoiding consistency and update latency problem in lazy master-master replication using a coordinator architecture

Master-Master replication enables us to replicate data through distributed writes and distributed reads. This is an advantage over Master-Slave replication. There are two problems we need to address in this paper with regards to master-master replication. The first is its nature of being loosely consistent and the second one is updates needs to be first agreed up on to be committed. In this paper, we will show how master-master replication can be consistent and how to optimize writes using a coordinator.