Performance Analysis of Endorsement in Hyperledger Fabric Concerning Endorsement Policies

Abstract

As a typical representative of a permissioned blockchain system, Hyperledger Fabric has garnered substantial attention in recent years. Despite the application and promotion of Hyperledger Fabric in fields such as smart grid and smart healthcare, challenges persist with respect to its performance, especially its transaction latency. High latency always discourages Hyperledger Fabric from latency-sensitive applications. In this paper, we focus on the execution phase of the unique Execute-Order-Validate architecture of Hyperledger Fabric and conduct a comprehensive analysis of its endorsement policies. We summarize three theorems and give corresponding mathematical proofs based on the definition of the endorsement policy. To better analyze the generation of latency during the endorsement process, we further developed a theoretical model using queuing theory. Subsequently, we conducted multiple experiments on Hyperledger Fabric v2.0 for performance evaluation. With the experiment results, we discuss the reasons for the performance differences between different endorsement strategies and between equivalent logical expressions. Eventually, this paper offers some suggestions for endorsement policy selection which can provide a reference for the application of Hyperledger Fabric in the actual production environment.