Raft Consensus Grouping Mechanism Based on Affinity Propagation Clustering Algorithm

An approach to address the shortcomings in the performance of existing consensus algorithms is the proposal of a Raft consensus grouping mechanism, which is based on the nearest neighbor propagation clustering algorithm: AP-Raft. First, the input eigenvalues of the algorithm are gathered from the network transmission efficiency and the current memory usage rate of each node. Second, the nearest neighbor propagation clustering technique groups all nodes based on the feature values. This allows each group's memory consumption rate and node network transmission efficiency to be approximated within a range. Finally, a leader node is chosen for each group, and to efficiently increase the load among the leader nodes, each leader node is connected using a consistent hash technique. Following experimental research, the AP-Raft method successfully increases the system's throughput and consensus time delay while maintaining data consistency when compared to the conventional Raft algorithm.