Performance Modeling and Evaluation of Hyperledger Fabric: An Analysis Based on Transaction Flow and Endorsement Policies
Carlos Melo, Glauber Gonçalves, Francisco A. Silva, André Soares
TL;DR
The paper develops a generalized SPN-based performance model for Hyperledger Fabric 2.5+ and validates it via case studies, focusing on the transaction flow from proposal to commit. It demonstrates how block size, arrival rate, endorsement policies, and gateway deployment shape latency, throughput, and utilization, identifying bottlenecks in endorsement and ordering while showing that distributed endorsements and gateways can stabilize latency at the expense of throughput. The approach provides a framework for service providers to tune Fabric deployments for throughput and latency trade-offs in industrial settings, and it highlights the impact of endorsement policy and gateway configuration on performance. Overall, the work contributes a generalizable modeling framework and empirical insights for configuring Fabric-based services with performance guarantees in mind.
Abstract
Blockchain is a paradigm derived from distributed systems, protocols, and security concepts. However, can blockchain applications provide services in industrial environments, especially concerning performance issues? In blockchains, long response times can impair both user and service experience, and intensive resource use may increase the costs of service provision. The proposed paper tries to answer this question by evaluating the performance of one of the most popular permissioned blockchain platforms, the Hyperledger Fabric (HLF). We provide a framework for performance evaluation based on modeling and experimentation. The results indicate that block size and arrival rate can compromise throughput (by -70%), latency (by +1,500%), and environment utilization (by +28%) and that multiple gateways can reduce latency (by -75%), and throughput (by -60%)