Abstract: Blockchain faces scalability issues, and the main solution currently is to use sharding to increase transaction processing capacity. However, the reduction in the number of nodes within a single shard after sharding can lead to centralization and gathering of Byzantine nodes within the shard, which can compromise network security. To address this issue, a sharding protocol that combines reputation and jump consistent hash for blockchain is proposed. Firstly, a node reputation level model is designed, which assigns different reputation levels to nodes based on their behavior in the network, thus reducing the influence of Byzantine nodes. Secondly, a reputation-based jump consistent hash sharding partial reconstruction algorithm (RJSPR) is proposed, which uses reputation values as an important consideration for shard reconfiguration, reducing the cost of shard migration and effectively improving the network security after sharding. Finally, a reputation-enhanced Raft consensus algorithm (RERaft) is proposed for in-shard consensus, which has a certain degree of Byzantine fault tolerance. Experimental results show that the RERaft algorithm can effectively improve the consensus efficiency of the system while possessing a certain degree of Byzantine fault tolerance. Compared to common Byzantine fault tolerant algorithms, it has better scalability.

Key words: blockchain, sharding technique, reputation mechanism, jump consistent hash, consensus protocol

摘要： 区块链面临扩展性问题，目前主要的解决方法是采用分片技术提升交易的处理能力，但分片过后单个分片内节点数量降低，容易造成分片内中心化与拜占庭节点聚集，对网络的安全性产生危害。针对此问题，提出了一个结合信誉跳跃一致性哈希的区块链分片协议。设计了一种节点信誉等级模型，通过节点在网络中的行为赋予节点不同的信誉等级，降低拜占庭节点在网络中的话语权；提出了基于信誉的跳跃一致性哈希分片部分重构算法（reputation-based jump consistent hash sharding partial reconstruction，RJSPR），将信誉值作为节点进行分片重配置的重要考量，减少分片迁移的开销，并且有效提升分片后的网络安全性；提出了一种基于信誉的改进Raft共识算法（reputation-enhanced Raft，RERaft）用于片内共识，使其拥有一定程度的拜占庭容错能力。实验结果表明，RERaft算法在具备一定拜占庭容错能力的同时可以有效提升系统的共识效率，相比于常见的拜占庭容错算法，具有更好的可扩展性。

关键词: 区块链, 分片技术, 信誉机制, 跳跃一致性哈希, 共识协议