Performance Improvement of Ordering and Endorsement Phase in Hyperledger Fabric

Abstract

In the fourth industrial revolution, blockchain is one of the popular technologies. There are many benefits such as decentralization, persistency and security. A blockchain network can be public or private. Among them, the private blockchain model is applied to the business model. It can be a consortium of networks and is efficiently handled using smart contracts. However, there are many problems in applying the blockchain technology to various environments. One of the concerns is that it does not provide enough performance for many transactions [1]. This paper focuses on improving the performance of Hyperledger Fabric which is reported to be the fastest private blockchain platform [2]. Fabric has various component such as chaincode(smart-contract), endorsing peer, ordering service and committing peer. We propose two optimization method about endorsing peer and ordering service. Firstly, we have improved the procedure of read transactions by distinguishing between read and write transactions in the endorsement phase. We achieved 60% better performance than original read transaction processing. Secondly, we create a new consensus protocol without using an external project. We decreased latency by 80% through ordering service optimization. By combining two optimizations, we improved the overall latency and throughput performance by 20%.