BunchBFT: Across-Cluster Consensus Protocol

TL;DR

BunchBFT is a scalable Byzantine fault-tolerant consensus protocol that uses a cluster-based hierarchical approach to significantly improve throughput and performance in distributed systems.

Contribution

It introduces a novel cluster-based design with hierarchical communication and decentralized leader election, enhancing scalability and throughput in BFT systems.

Findings

Achieves up to 10x throughput improvement over MirBFT.

Proven to satisfy safety and liveness properties.

Demonstrated effective performance in wide-area network settings.

Abstract

In this paper, we present BunchBFT Byzantine fault-tolerant state-machine replication for high performance and scalability. At the heart of BunchBFT is a novel design called the cluster-based approach that divides the replicas into clusters of replicas. By combining this cluster-based approach with hierarchical communications across clusters, piggybacking techniques for sending messages across clusters, and decentralized leader election for each cluster, BunchBFT achieves high performance and scalability. We also prove that BunchBFT satisfies the basic safety and liveness properties of Byzantine consensus. We implemented a prototype of BunchBFT in our PaxiBFT framework to show that the BunchBFT can improve the MirBFT's throughput by 10x, depending on the available bandwidth on wide-area links.