Configuration Driven Performance Analysis of Hyperledger Fabric: An Empirical Study

Abstract

Hyperledger Fabric has emerged as a leading permissioned blockchain platform for enterprise applications due to its modular architecture, fine-grained access control, and support for scalable transaction processing. However, configuring Fabric for optimal performance in real-world deployments remains a non-trivial task, as multiple parameters jointly influence latency and throughput. This paper presents a comprehensive empirical evaluation of Hyperledger Fabric performance, focusing on the combined impact of endorsement policies, block size, batch timeout, and peer scalability. A structured experimental methodology is adopted, comprising three test campaigns that analyze (i) the effect of endorsement policies, (ii) the tuning of block size and timeout, and (iii) scalability with respect to the number of peers per organization. Performance is assessed using percentile-based latency metrics (P50, P95, P99) and transaction throughput under controlled yet realistic deployment conditions. The results show that the OR endorsement policy consistently outperforms the AND policy in terms of both latency and throughput, while a block size of 100 transactions with a timeout between 50 and 100 ms achieves the best performance trade-off. Furthermore, the network maintains stable performance up to 2–4 peers per organization, beyond which scalability degradation becomes pronounced. Based on these findings, the paper derives practical configuration guidelines for enterprise-grade Hyperledger Fabric deployments. The presented results provide actionable insights for performance tuning and contribute to a deeper understanding of Fabric’s scalability behavior in multi-peer environments.