Churn Tolerance Algorithm for State Machine Replication

Abstract

State machine replication is a general approach for building a Byzantine fault-tolerant (BFT) distributed systems like a grid or a cloud or a data center. Various BFT protocols based on state machine replication have been introduced for modern distributed systems to tolerate byzantine failures and thus provide more reliable services. However, most of them have not fully considered the adverse affects of Churn (nodes entering and leaving the system at will) on the correctness and availability of services. In this paper, we propose a new churn tolerance algorithm based on the BFT protocol approach using state machine replication. It is capable of tolerating both Byzantine failures and arbitrary churn when the constraint on the number of faults in the distributed system is satisfied.