Performance Diagnosis and Optimization for Hyperledger Fabric
Shenbin Zhang, Hua Zhu Song, Bingfeng Pi, Jun Sun, Kazuhiro Yamashita, Yoshihide Nomura · 2020
Hyperledger Fabric is a permissioned blockchain platform which can solve the trustless problems among enterprises. However, the limited transaction throughput prevents the further use of Fabric platform. The performance of Fabric network depends on many factors, such as network parameters, node numbers and hardware limitations. In order to improve the performance of runtime Fabric network, this paper proposed a solution for performance diagnosis and optimization. We conduct lots of performance testing and collect the analysis rules of performance bottleneck. For runtime network, we monitor the performance data, node resources and network parameters in real time. Then we diagnose the performance health and analyze the reasons which cause the bottleneck. Finally, we dynamically tune the network parameters to adapt the network environment. Besides, we provide some suggestions for network maintainer to improve the performance.