Research On Scalable Consensus Service Technology For Consortium Chains

Blockchain technology has been used in various scenarios since its inception,thanks to its distributed,security and traceability characteristics.However,compared with established solutions such as distributed database systems,the throughput of the current blockchain is still very small,which significantly hinders the development of the blockchains.As a blockchain system with an access mechanism,the consortium chain has added supervisable features compared to the public chain and has hugely improved transaction processing and transaction confirmation delays compared to the public chain.Whereas,the current consensus algorithm used by the consortium chain only supports single-node serial block generation.For example,Hyperledger Fabric,a representative of the consortium chain,uses the Raft consensus algorithm based on the"leader-follower" model,which generates blocks only within its single Leader node,and which Follower nodes only participate in the consensus process.The feature of serial packing with a single node will cause transaction blockage and severely reduce system performance.To build a scalable consortium chain with enhanced performance,this paper proposes a DHT-Orderer consensus mechanism that supports multiple orderer nodes to generate blocks in parallel based on the idea of DHT(distributed hash table)in the P2P system.The specific research content includes:1.According to DHT's consistent hashing algorithm,all transactions and orderer nodes can be mapped in a single ring using a specific hashing algorithm.The transaction is routed to a specific node in the ring for processing through the DHT algorithm.Therefore,all orderer nodes in the ring can independently generate blocks.At the same time,since DHT supports dynamic j oining and leaving of nodes,DHT-Orderer no longer needs a protocol similar to Raft/Paxos.In order to avoid unbalanced transaction distribution among orderer nodes,virtual orderer nodes can be added to the ring to alleviate the processing and communicating pressure of physical nodes.2.Given the problem of the consistency of the block sequence in the distributed block packing architecture,the node routed by key=0 is set as the mainNode in the DHT-Orderer system,and the rest of the nodes package the transaction into an"unsorted block" and forward it to the mainNode.After the mainNode receives it,the block will be sorted,then the whole block processing is completed.The mainNode accomplishes the sorting task according to the FIFO strategy,which can achieve extremely high performance.3.A pluggable DHT-Orderer prototype system is implemented based on the Hyper-ledger Fabric platform.To verify the performance of DHT-Orderer,performance tests were carried out in terms of block capacity,timeout period,the maximum allowable number of Tx,number of DHT nodes participating in the consensus,number of concurrent transactions,and communication mechanism.The experimental results show that when the parameters approximate(AbsoluteMaxBytes=5MB,PreferredMaxBytes=2MB),the timeout period=2s,and the maximum number of Tx received=500,the system TPS can reach 1700.The DHT-Orderer consensus mechanism performs well in terms of throughput,latency,and scalability,and meets the needs of enterprise-level alliance chains.