Research On Consensus Mechanism Based On Witeness Node Selection And Replacement Scheme

The blockchain technology has been widely used in the cloud computing,the personal identity management,the supply chain traceability,and the medical health information management fields,because it is in the process of continuous improvement and has advantages such as decentralization,tamper proof characteristic and anonymity.As the key technology of blockchain technology,the efficiency of consensus mechanism affects the performance of blockchain applications directly.At present,the mainstream consensus mechanisms are not perfect in terms of block throughput and the fairness of block generation rights obtained by nodes.Therefore,how to design a consensus mechanism that meets the above requirements is worthy to study.To tackle the problems of low block throughput and poor nodes fairness caused by the witness nodes selection scheme in the existing consensus mechanisms,this thesis designs a consensus mechanism based on the multi-parameter witness nodes selection scheme.In this consensus mechanism,the back propagation neural network algorithm utilizes multiple performance parameters to predict the nodes' speed of block generation.According to the predicted values,multiple nodes with high speed of block generation are selected to constitude the witness nodes set.If there are several nodes with the same block generation speed,the consensus mechanism will preferentially select nodes that have been chosen less frequently as witness nodes during the nodes selection,to increase the probability of block generation rights obtained by a few nodes and ensure the nodes fairness.In the subsequent consensus process,these witness nodes are responsible for generating blocks in turn,to speed up the block generation and improve the block throughput.Secondly,when the witness nodes fail,the system needs to wait until the block generation time limit is reached,and the next witness node regenerates the block,therefore low block throughput occurs in the existing consensus mechanisms.To address the problem of low block throughput,this thesis designs an improved consensus mechanism based on the replacement of failed witness nodes.During the nodes selection,the consensus mechanism will select multiple witness nodes and wait nodes based on the speed of block generation and selected times before.When the witness nodes fail,these nodes are replaced by wait nodes directly,and the block generation is continuous.Therefore,the waiting time for nodes replacement is reduced,and the block throughput is increased.To avoid the failed nodes are selected for generating blocks again,they are removed from the witness nodes set,at the same time.Finally,the consensus mechanism proposed in this thesis is compared with the several existing mature consensus mechanisms,such as the workload proof consensus mechanism,the stack proof consensus mechanism and the delegated stack proof consensus mechanism.The experimental results show that compared with the above consensus mechanisms,the consensus mechanism proposed in this thesis provides a higher block throughput.Compared with the delegated stack proof consensus mechanism,it improves the nodes fairness,and provides a higher block throughput when nodes fail.