Improved Raft consensus algorithm based on NSGA-II and K-Means

• Published in: 2024 10th International Symposium on System Security, Safety, and Reliability (ISSSR) pp. 383 - 390
• Main Authors: Yang, Sijia, Tan, Pengliu, Fu, Haowei
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.03.2024
• Subjects: Blockchain; Blockchains; Clustering algorithms; Consensus algorithm; Fault tolerant systems; Leader election; Log replication; Raft consensus algorithm; Scalability; Throughput; Voting
• ISSN: 2835-2823
• DOI: 10.1109/ISSSR61934.2024.00055

Aiming at the problems of inefficient leader election, high leader node carrying pressure, and low scalability of blockchain network of Raft consensus algorithm, an improved Raft consensus algorithm-NK-Raft is proposed. The multi-objective function is established during the leader election phase, and the NSGA-II algorithm is utilized to conduct the leader election directly. By doing this, the election efficiency is increased, and time-wasting several rounds of leader node competition in the Raft consensus process are avoided. In order to decrease the load-bearing pressure on the leader nodes and increase the scalability of the blockchain network, the tightness value is proposed to stratify the network during the log replication phase. The K-Means++ algorithm is then introduced to group the node clusters in the blockchain network. The experimental study's findings demonstrate that, in comparison to the Raft consensus algorithm and the MCRaft algorithm, the NK-Raft algorithm reduces the election and consensus delays of the leader node as the blockchain network scales up, increases throughput, significantly lessens the load-bearing pressure on the leader node, has better scalability, and guarantees the blockchain network's performance.