SL-Chain: Parallel Execution Model of Hyperledger Fabric Chaincode in Shared-log Environment

Abstract

Two main challenges of the blockchain system are the inconsistency of the execution results of the distributed application (smart contract) executed on the nodes that individually maintain the state, and the performance scalability. Blockchain systems often consist of multiple nodes that share application state. However, in an architecture in which each node maintains independent storage updated through blocks, it is difficult to maintain the state of the nodes consistently and achieve high performance. Hyperledger Fabric introduced an execute-order-validate (EOV) architecture to overcome the limitations of blockchain systems that followed the existing order-execute architecture. Nevertheless, there are many concerns that the performance of permissioned blockchain is insufficient to apply to real world business applications. In this thesis, we propose Shared-log based Chain (SL-Chain), parallel execution model of Hyperledger Fabric chaincode in shared-log environment that overcome the performance scalability problem of the blockchain system. SL-Chain is the first blockchain system which introduces shared logs to execute distributed applications. Furthermore, we introduce a sequencer to handle conflicts between transactions competing for one entry in the shared log. SL-Chain executes user transactions in parallel with multiple nodes, and the state consensus among multiple nodes is performed based on a shared log protocol. For implementation, we adopted vCorfu as a shared log protocol. The experimental evaluation shows that SL-Chain achieves 30,000 TPS (Transactions Per Second) of with multiple adapter servers.