Research On Transaction Concurrency Conflict Based On Hyperledger Fabric And Realization Of Prototype System

Due to the characteristics of decentralization,tamper-proof and trustless,blockchain technology has attracted much attention.As a distributed ledger database,blockchain technology can well solve the trust problem in the context of big data.However,compared with traditional databases,blockchain technology still has some shortcomings in terms of performance,which is one of the main reasons why this technology cannot completely replace traditional databases.Hyperledger Fabric(Fabric),as a representative of the consortium chain,has been widely adopted due to its architectural design that makes it more efficient in processing data than the public chain.Although the throughput of Hyperledger Fabric is much higher than that of the public chain and other alliance chains,it is still not enough to meet the needs of some commercial applications,and there is still a certain distance from large-scale commercial use.This is because there are a large number of parallel conflicts in the system,causing the system to reject many transactions,thereby reducing throughput.Therefore,this paper focuses on solving the problem of concurrency conflicts in the Hyperledger Fabric system in order to improve the system response speed and throughput.The main work content of this paper includes the following three parts:(1)For "intra-block conflicts",this paper points out and proves the system reliability problems of the existing solution Fabric++through experiments,and proposes a bitmap-based BMR algorithm and a topological sort-based TR algorithm based on it.The BMR algorithm and the TR algorithm rely on the analysis of the global optimal solution and the local optimal solution,and use the greedy idea to solve the "conflict within the block" problem.When they receive a transaction,they will conduct conflict analysis on it to decide whether to keep the transaction.(2)For "inter-block conflicts",this paper points out the problems existing in the existing solution LMLS,and through the analysis of the transaction processing flow,on the basis of the BMR and TR reordering algorithms,a GTCL mechanism is proposed.After the reordering algorithm and before the verification phase,it locks the keys involved in the write set of the transaction to be updated,and broadcasts the information to all peer nodes through the sorting node for caching,realizing the internal locking mechanism of the system.(3)Design and implement a transaction prototype system based on the improved Hyperledger Fabric,which simulates the basic business of the bank,including account creation,deposit and withdrawal,transfer and other functions.In terms of deployment,Docker and Docker-Compose technologies are used to simplify the process,and the improved solutions such as GTCL mechanism and reordering algorithm are integrated into Hyperledger Fabric,so that users do not need to perform additional learning or operation,thereby reducing the difficulty of using the system.Improved usability.Experiments show that the improved scheme proposed in this paper can effectively solve the problem of concurrency conflicts in Hyperledger Fabric.Regardless of low-conflict scenarios or high-conflict scenarios,the improved system can achieve greater throughput,especially in highconflict scenarios.At the concurrency level,the total throughput and successful transaction throughput of the system can be effectively increased in the case of high concurrency,and the performance of the original Fabric can be equal to that of the original Fabric in the case of low concurrency.