Raft-PLUS: Improving Raft by Multi-Policy Based Leader Election with Unprejudiced Sorting

Raft is a fast, scalable, understandable consensus algorithm widely used in distributed systems. The Leader handles client requests and interacts with other servers to reach consensus, so stable, reliable, powerful crucial for the cluster. We designed policy-based voting mechanism make elected as reliable possible. In order improve asymmetric relationship between Followers Leader, we trigger new round of election actively that node can transform into Follower under certain conditions enhance symmetry servers. Our proposed Raft-PLUS makes possible through four policies three opposition election. To verify effectiveness algorithm, configure different on 12 simulate process explain process. demonstrate advantages built key-value stores based Raft-PLUS, tested performance normal abnormal states. Experimental results show has similar write throughput conditions. Regarding quality network changes, average 40% higher than algorithm. Leader’s CPU usage fluctuated; was 38% Raft.