Research On Byzantine Consensus Algorithm In Asynchronous Environment

In a distributed system,its consistency is difficult to achieve because of a series of problems such as unreliable network communication and possible failure of nodes.The byzantine consensus algorithm is the core research content of high-availability distributed systems,which enables the system to provide correct and consistent services in the event of various errors.At the same time,asynchronous systems are more practical than synchronous systems because they do not rely on time-based assumptions.Therefore,the study of Byzantine consensus algorithm in asynchronous environment has very significant theoretical and practical significance.Interaction consistency(IC)is one of the earliest consensus issues.Most of the earlier studies focused on synchronous systems with very strong synchronization assumptions.However,in the current Internet,the actual distributed system is asynchronous in most cases,so the traditional synchronous interactive consistency algorithm is not applicable.In a purely asynchronous environment,since it is impossible to distinguish whether a node is a failed node or a slow node,the interaction consistency in a pure asynchronous environment can not be solved.In view of the above situation,this thesis proposes an efficient interactive consistency algorithm(IC,RBC-ABVA)in asynchronous environment(non-pure asynchronous)based on the work of Diamantopoulos et al.The algorithm sends the proposed value of the correct node to all nodes in the network by Reliable Broadcast Protocol(RBC),and directly agrees on the proposed values of all nodes by asynchronous binary vector Byzantine Consensus Protocol(ABVA),which improves the traditional solution of parallel execution of multiple binary consensus protocols,the overall message complexity is only O(N~3),which is obviously better than the algorithm proposed by Diamantopoulos et al.The experimental results show that the algorithm proposed in this thesis has obvious advantages in performance.The vector consensus problem weakens the validity definition in the interaction consistency problem,making the vector consensus solvable in a purely asynchronous environment.In this thesis,the efficient asynchronous Byzantine consensus algorithm HoneyBadgerBft(vector consensus)is deeply analyzed.Based on this,a more efficient asynchronous Byzantine consensus algorithm HBBFT-improved is proposed.The algorithm decouples the traditional asynchronous common subset protocol(ACS),adds a storage layer and sorts the transaction just like PBFT,so that in the multi-round execution scenario,the reliable broadcast of the node transaction does not need to wait for the completion of the consensus phase,improving the throughput of the algorithm to a certain extent.Abandon the way of negotiating the common coin value of each round alone by every asynchronous binary consensus protocol(ABA)instances in the node,and choose to have all ABA instances running in parallel in the node to negotiate the common coin of each round together,in order to reduce the number of threshold signature operations which is very time-consuming and the number of messages that need to be transmitted,reducing the latency of the algorithm to some extent.In this thesis,the correctness and security of the HBBFT-improved algorithm are theoretically analyzed and proved,and the actual performance evaluation is carried out through experiments.The experimental results verify that the algorithm proposed in this thesis has certain performance advantages in multiple rounds of execution scenarios.