StarReact: Detecting Important Network Changes in BFT Protocols with Star-Based Communication

Threshold signatures have improved the scalability of BFT protocols by replacing all-to-all broadcast with star-based communication. On the flip side, this approach renders network re-organization and performance optimization more costly, because information can only be exchanged between the leader and all other nodes.We present StarReact,an extension for BFT protocols that relies on star-based communication to collect and disseminate quorum certificates. This extension allows all nodes of the system to measure and evaluate the network state by monitoring the messages disseminated by the leader. Each node decides independently if performance degradation justifies a leader change. By deploying a median filter, the measured commit latency of the system is evaluated by each node to determine if a network change has a lasting effect that warrants a change in leadership or should be ignored. We showcase how StarReact is able to identify important network changes for different deployment scenarios and explain how view change executions should be adapted to different environments to improve commit latency and potentially throughput for such systems.