Fast Follower Recovery for State Machine Replication

摘要

The method of state machine replication, adopting a single strong Leader, has been widely used in the modern cluster-based database systems. In practical applications, the recovery speed has a significant impact on the availability of the systems. However, in order to guarantee the data consistency, the existing Follower recovery protocols in Paxos replication (e.g., Raft) need multiple network trips or extra data transmission, which may increase the recovery time. In this paper, we propose the Follower Recovery using Special mark log entry (FRS) algorithm. FRS is more robust and resilient to Follower failure and it only needs one network round trip to fetch the least number of log entries. This approach is implemented in the open source database system OceanBase. We experimentally show that the system adopting FRS has a good performance in terms of recovery time.