【Author】
Chen, Yineng; Li, Ming; Zhu, Xinghui; Fang, Kui; Ren, Qingshan; Guo, Ting; Chen, Xiaoxuan; Li, Cheng; Zou, Zhuoyang; Deng, Yuechao
【Source】INFORMATION PROCESSING & MANAGEMENT
【Abstract】Consortium chain can better combine blockchain technology with market business, so it is adopted by all walks of life and develops at a large scale. Practical Byzantine Fault Tolerance (PBFT) is more suitable for consortium chain, which are partially decentralized, resistant to Byzantine nodes and strong consistency. However, the limited network scale that PBFT can support is not conducive to the large-scale development of consortium chain. Based on the analysis of the working principle and consensus mechanism of PBFT, this study proposed an al-gorithm to improve PBFT: feature grouping and credit optimization Byzantine Fault Tolerance (FCBFT). In this algorithm, a feature grouping model is proposed to optimize the node structure of large-scale consortium chain, which divides large-scale network nodes into different institutions to form independent consensus groups by feature grouping. On this basis, a reputation score reward mechanism is proposed to improve the consensus efficiency of large-scale consortium chains. It introduces a reputation score calculation formula to select high-reputation primary nodes. At the same time, a replacement cycle is established to replace high-reputation nodes with low-reputation nodes, so as to optimize the consensus efficiency of the consortium chain. The experimental results show that FCBFT has shorter delay and higher throughput (TPS) than PBFT, which becomes more obvious as the number of nodes increases. After the number of nodes ex-ceeds the threshold, the TPS of PBFT drops rapidly, while FCBFT can maintain high efficiency and stability. FCBFT's block generation speed is higher than PBFT, and its relative growth rate has steadily increased.
【Keywords】PBFT; Consortium chain; Consensus algorithm; Feature grouping; Credit
【标题】一种针对大规模联盟链的实用拜占庭容错改进算法
【摘要】联盟链能够更好地将区块链技术与市场业务相结合，因此被各行各业采用并大规模发展。实用拜占庭容错（PBFT）更适用于部分去中心化、抗拜占庭节点、一致性强的联盟链。但 PBFT 所能支持的网络规模有限，不利于联盟链的大规模发展。本研究在分析PBFT的工作原理和共识机制的基础上，提出了一种改进PBFT的算法：特征分组和信用优化拜占庭容错（FCBFT）。该算法提出了一种特征分组模型来优化大型联盟链的节点结构，将大型网络节点划分为不同的机构，通过特征分组形成独立的共识组。在此基础上，提出了声誉评分奖励机制，以提高大型联盟链的共识效率。它引入了声誉分数计算公式来选择高声誉的主节点。同时，建立替换循环，将高声誉节点替换为低声誉节点，从而优化联盟链的共识效率。实验结果表明，FCBFT 比 PBFT 具有更短的延迟和更高的吞吐量（TPS），随着节点数量的增加，这一点变得更加明显。节点数量超过阈值后，PBFT 的 TPS 迅速下降，而 FCBFT 可以保持高效率和稳定性。 FCBFT 的出块速度高于 PBFT，相对增长率稳步提升。
【关键词】PBFT；联盟链；共识算法；特征分组；信用
评论