ImprovedRaft Algorithm Based on Communication Bottleneck and Log Inconsistency

In order to solve the problems of single Leader communication bottleneck and node log inconsistency, we propose ImprovedRaft algorithm based on Raft. To solve the single Leader communication bottleneck problem, we intro-duce Coordinator, which is empowered to forward the App-endEntries RPC message sent by the Leader node and collect the feedback of the message in the log replication phase, assisting the Leader node to complete the log replication process. In large Raft clusters, the communication bottleneck caused by a single Leader can be mitigated. Aiming at the problem of node log inconsistency, this paper proposes a log accelerated back grouping method, combined with a coor-dinator to make the logs in the group reach agreement quickly. Finally, experimental verification and analysis are conducted on Hyperledger Fabric platform. The experimental results show that compared with the Raft algorithm, ImprovedRaft algorithm reduces the average transaction processing time by 4.6% and improves the throughput by 2.3% for different numbers of transaction requests; and reduces the average transaction processing time by 2.6% and improves the throughput by 2.1 % for different numbers of nodes.