Performance improvement of the consortium blockchain for financial business applications

Abstract

In preparation for improving the consortium blockchain’s performance for financial business applications, the current authors first surveyed typical records of transaction rates for various financial systems and the performance capacity of several known consortium blockchains at this moment in time. The blockchain itself does not offer any performance advantage, but rather sacrifices performance to achieve tamper resistance. The average transaction rates for blockchain technologies, however, are continuously improving and have already reached around 1,000–2,000 transactions per second (TPS), which is equivalent to the transaction rate for the Zengin System, the common name for Japanese Banks’ Payment Clearing Network. Next, the current authors analysed and improved the transaction rate for Hyperledger, a Linux Foundation Project, and Fabric, one of the open source software tools for implementing consortium blockchain. The current authors found that inefficient message transfer between the platform container and the application container was the main cause of Fabric’s performance bottleneck. Therefore, the current authors have introduced more efficient application programming interface (API) between the containers and improved the transaction rate from 725 to 1,350 TPS (an 86% increase). Finally, the current authors analysed Fabric’s performance under artificial network latency. This paper determines that Fabric running over a wide area network might be required to replace its consensus algorithm to improve its transaction rate because other causes of performance deterioration, including the consensus algorithm, seem to become dominant.