MH-Raft: an efficient and low-latency consensus algorithm for distributed systems via MOEA/D and hybrid hierarchical clustering

• MH-Raft introduces a hierarchical node management and election framework that leverages MOEA/D to minimize election latency by evaluating multi-dimensional node attributes. • A rigorous tightness definition enables optimal mediator node selection through a hybrid clustering algorithm, adaptively partitioning the network and optimizing mediator–follower mappings. • Comprehensive experiments show that MH-Raft reduces consensus latency by 14.87%–34.45% and boosts average throughput by 30.43% compared to conventional Raft. • The proposed algorithm offers a scalable and low-latency consensus solution for large-scale distributed systems and blockchain applications.