Research On Concurrency Conflict Optimization In Hyperledger Fabric

Blockchain technology has the advantages of decentralization,immutability,and traceability,which have been widely used.Hyperledger Fabric(Fabric)is a popular consortium blockchain platform,which adopts a highly modular architecture and combines the advantages of decentralization of the public blockchain and high performance of the private blockchain.It is suitable for developing enterprise-level applications in a commercial environment.Fabric outperforms other public blockchain platforms like Bitcoin because the transaction processing flow(execution,ordering,and validation phases)in Fabric doesn’t require a high-computation proof-of-work mechanism to reach consensus,and transactions can be executed concurrently in the execution phase.However,compared with traditional distributed systems,there is still a large gap in the performance of Fabric,especially in scenarios with transaction concurrency conflicts,the performance of Fabric is greatly reduced,which hinders the application of Fabric to practical scenarios.Therefore,we have carried out a series of research work on Fabric transaction concurrency,and proposed two optimization methods,detecting conflicts and merging transactions,to improve the performance.The main research contents are as follows:1.To solve the problem of low efficiency caused by identifying conflicting transactions in the validation phase of Fabric,we analyze the causes of concurrency conflicts in detail and propose a method that is detecting conflicts based on the caching mechanism.By maintaining a cache to store unverified transactions at the endorsers,which can detect transaction conflicts by the cache in advance in the execution phase,and the identified conflicting transactions can be terminated early to avoid them occupying system resources,so those valid transactions can be processed by peers faster.Experiments show that this method can improve the concurrent performance of the Fabric system.2.For the semantically legal and conflicting transactions in asset transfer scenarios,we analyze the transaction semantics and the updating process of the state,and propose a method that is optimizing conflicts based on the merging mechanism,which takes maximizing the number of merging transactions as the optimization goal.By preprocessing the read-only transactions and inevitable invalid transactions,the mechanism can reduce the number of potentially conflicting transactions in the block;By obtaining the approximate solution through the merging transaction algorithm,the mechanism can avoid the legal transactions being marked as invalid due to concurrency conflicts.The experimental results show that this method can significantly improve the throughput and the rate of successful transactions in the scenario of concurrency conflicts.