Research On Optimization Of Consensus Algorithm In Blockchain Network

Due to the unique advantages of decentralization,tamper proof and high traceability,blockchain technology has become the key breakthrough object of the new generation of information technology innovation at home and abroad.At present,blockchain application is difficult to be implemented,one of the more important reason is the bottleneck of consensus mechanism.This thesis proposes two improved consensus algorithms for the blockchain consensus mechanism with different error types.The improvements are as follows:1?To improve the non Byzantine fault-tolerant consensus algorithm,this thesis proposes an improved raft algorithm PB-RAFT(parallel batch raft)based on parallel batch processing.Aiming at the time waste caused by the linear operation of the original raft algorithm and the performance problems caused by the large number of distributed nodes and high concurrency in practical application,PB-RAFT algorithm introduces pipeline and batch processing in the process of processing client requests and log replication.In the process of leader election mechanism of the original raft algorithm,when there are isolated nodes in the cluster,it will disturb the normal election of the algorithm.PB-RAFT algorithm introduces the provote process in the leader selection process,and sets a leader priority value for the nodes in the cluster,which is integrated into Provo In the process.Finally,the PB-RAFT algorithm is tested and compared with the original other algorithm through simulation experiments.The experimental results show that the number of leader elections triggered by the improved algorithm in the same time is significantly lower than that triggered by RAFT.At the same time,the PB-RAFT algorithm is also lower than that of raft algorithm in terms of consensus time.2?In order to improve Byzantine fault tolerance consensus algorithm,this thesis proposes an efficient consensus algorithm ER-PBFT(electronic representative Byzantine fault tolerance)based on the participation of regional representatives.In the original PBFT algorithm,too many nodes in the three-stage consensus process will lead to a sharp increase in the number of communication times of message delivery and the consumption of network resources,which is easy to cause network congestion.ER-PBFT divides the whole blockchain cluster into regional sub clusters according to the geographical location,and introduces the node credit score and credit level conversion mechanism to select regional representative nodes Participating in the global consensus greatly reduces the number of nodes participating in the PBFT consensus process,solves the performance bottleneck problem caused by the three-stage consensus of the original PBFT algorithm in the face of distributed applications with large-scale nodes,and greatly improves the efficiency of the algorithm.Finally,the ER-PBFT algorithm is tested and compared with the other algorithm through simulation experiments.The experimental results show that the improved algorithm is superior to the other algorithm in terms of delay,throughput and resource consumption.There are 40 figures,5 tables and 82 references in this thesis.