Hydrofoil: Reducing Downtime by Combining Leader-Based and Randomized Replicated State Machines

Abstract

This thesis introduces Hydrofoil, a new linearizable replicated state machine that combines both leader-based and randomized leaderless approaches. While the leader is active, Hydrofoil is able to perform as efficiently as other leader-based protocols such as Raft. In addition, by leveraging the power of Ben-Or, a randomized consensus algorithm, Hydrofoil is able to make progress even when the leader is slow, or during the period when a new leader is being elected.