Analysis And Improvement Of PBFT Consensus Mechanism In Fabric

Blockchain technology solves the problem of untrustworthiness in the traditional model by virtue of decentralization.The consistency of the data of each node in the blockchain must be maintained by the consensus mechanism,but the imperfection of the consensus mechanism scheme seriously affects the performance of the blockchain.This paper mainly studies and analyzes the problems existing in the PBFT consensus mechanism used in Fabric0.6,and proposes solutions.Through the analysis of the PBFT consensus mechanism,it is found that there are two main problems.First,the primary node is selected in order.This selection method allows those nodes with low hardware performance and poor performance in the consensus to have equal opportunities to be selected as the primary node.Therefore,in the case of a large number of incorrect nodes,the primary node is likely to be frequently replaced.This will reduce the efficiency of the system.Second,The amount of message traffic between nodes is large.In the PBFT consensus process,it needs to broadcast to the whole network,and the traffic reaches O(n2),which limits the scalability of the system.This paper proposes an improved practical Byzantine fault tolerant algorithm IPBFT consensus mechanism to solve the above two problems.For the primary node to select in order,in this paper,a reliability evaluation strategy is designed.The nodes with good performance and excellent performance in the consensus are put into the candidate queue of the primary node,and then the primary node is selected only in this reliable queue.This can ensure that the node with high reliability serves as the primary node,reducing the frequency of occurrence of the primary node replacement,and making the system run more stably.In view of the large traffic volume of the PBFT consensus mechanism,the IPBFT consensus mechanism does not use the whole network broadcast,but each verification node sends a vote and confirmation message to the current primary node in the PREPARE and COMMIT phases of the consensus process.When the primary node receives responses from two-thirds of the nodes that are greater than the total number of nodes,they are aggregated by the aggregate signature algorithm,and then each verification node can verify that the majority node has voted correctly by verifying the aggregated signature,thus simplifying Consensus process to reduce traffic.In the laboratory simulation environment,this paper conducts a comparative experiment on the proposed IPBFT consensus mechanism and the PBFT consensus mechanism in Fabric,and tests it from three aspects:fault tolerance,transaction throughput and system scalability.According to the experimental results,the following conclusions are obtained:Regarding the system transaction throughput,IPBFT can significantly improve the transaction throughput of the stable operation of the system while ensuring the efficiency of the original query.Regarding the scalability of the system,as the number of nodes increases,the rate of decline in IPBFT transaction throughput is significantly slower than that of PBFT,so it has better scalability.Regarding the fault tolerance of the system,the IPBFT consensus mechanism can achieve the same fault tolerance as the PBFT consensus mechanism,that is,the fault tolerance rate is 33%.