Majority Consensus Protocol for Strong Consistency and High Availability

This paper presents a new approach to coordinating clusters of fail-over storage servers. The proposed protocol is designed for supporting large scale on-line storage services which positioned for strong consistency and high availability. Without sacrificing much performance, the protocol requires only one round-trip message communication with redundant servers. A distributed file system with single centralized master extended with this protocol confirms its efficiency as tens of thousands of operations per second and high availability of 99.9995%.