A Load-balanced Byzantine-fault-tolerant Consensus Algorithm With Efficient View Change

Since Satoshi Nakamoto proposed the "decentralized" Bitcoin,blockchain technology has received more and more attention.As the core technology of the blockchain,the Byzantine consensus algorithm ensures the security of the blockchain in the presence of malicious nodes.The master-based Byzantine consensus algorithm uses the master node to broadcast requests from clients to achieve consensus,which can have a higher consensus efficiency.However,the master node needs to undertake more calculations and network traffic,so there will be an imbalance of load between the master node and the slave node.Due to the unbalanced load,the probability that the master node collapses and the system needs to elect a new leader increases accordingly.Most master-based Byzantine consensus algorithms require a lot of redundant synchronization operations when re-electing a new master node,which leads to a lower view change efficiency.Aiming at the problem of unbalanced load between the master node and the slave node,this paper proposes a load-balanced log replication protocol.In the log replication protocol,not only the master node but also the slave nodes could start the consensus process.This log replication protocol makes the broadcast pressure and calculation pressure of the master node evenly distributed to all nodes,avoiding the unbalanced load between the master node and the slave node.Through experimental analysis,the proposed log replication protocol has the characteristics of load balancing,high throughput,low client latency and high scalability.Aiming at the problem of low view change efficiency of Byzantine consensus algorithm,this paper proposes an efficient view change protocol.In the view change protocol,the elected new master node has the most complete log list,which could avoid the redundant synchronization process caused by the missing log entries in the master node.After the election of the master node is completed,the view replacement protocol obtains the status of the lagging slave nodes through the query of the master node,and then all the lagging log entries from the node are packaged and sent at once.In this way,the new master node will only synchronize log entries that are lagging behind,instead of synchronizing all log entries after the previous checkpoint,which greatly avoids redundant synchronization operations.Through experimental analysis,the view change protocol proposed in this paper has lower view change time,lower CPU calculation and lower network traffic,and it also has better scalability.