Research On Optimization Of Raft Consensus Algorithm Based On Multi-strategy

Distributed systems have become the primary way to solve traditional centralized systems' processing efficiency and data security problems.A consensus algorithm is the process and rules by which multiple nodes in a distributed system agree on a value.The emergence of blockchain technology has made consensus algorithms a research hotspot at home and abroad again.Combining traditional consensus algorithms and blockchain technology also allows their implementation in various fields.The Raft algorithm is a simplified version of the classic distributed consensus algorithm Paxos,designed based on the replicated state machine model.The algorithm is broken down into smaller sub-problems for better understanding,implementation,and debugging.However,the Raft algorithm also suffers from unstable runtime performance and uncontrollable election results.Given the problems existing in the Raft algorithm,this topic focuses on the leader election mechanism in the algorithm.A multi-strategy-based election optimization scheme for the Raft is proposed and is divided into the active trigger election mechanism and the priority election timeout mechanism according to the applicable phases.The active trigger election mechanism evaluates and votes on the performance and security of the incumbent leader by adding multiple performance and security metrics to the heartbeat request message of the leader,and adding multiple strategies corresponding to them to the heartbeat message processing algorithm of the followers.The priority timeout election mechanism adds performance and stability related to multiple strategies to the timeout mechanism of the election process.It adds multiple strategies based on priority factors to increase the probability of a server with high priority becoming the leader during the leader election process.In order to verify the algorithm optimization scheme,a distributed key-value storage system based on multi-strategy Raft election optimization scheme is developed,based on which multi-dimensional comparison experiments are designed to verify the improved scheme.The experimental results show that the election mechanism proposed in this topic can actively trigger a new round of leader elections when the leader experience performance degradation and centralization problems while ensuring the efficiency and security of the algorithm.It can also modify the timeout duration according to the priority level of the servers to elect the servers with better performance and stability in the cluster,which can improve the operation efficiency of the system.Finally,the algorithm's security,consistency,activity,and centrality issues are discussed.