Performance Modeling of PBFT Consensus Process for Permissioned Blockchain Network (Hyperledger Fabric)

TLDR: This paper model the PBFT consensus process using Stochastic Reward Nets to compute the mean time to complete consensus for networks up to 100 peers and conducts sensitivity analysis over a variety of system parameters to examine the performance of larger networks.

Abstract: While Blockchain network brings tremendous benefits, there are concerns whether their performance would match up with the mainstream IT systems. This paper aims to investigate whether the consensus process using Practical Byzantine Fault Tolerance (PBFT) could be a performance bottleneck for networks with a large number of peers. We model the PBFT consensus process using Stochastic Reward Nets (SRN) to compute the mean time to complete consensus for networks up to 100 peers. We create a blockchain network using IBM Bluemix service, running a production-grade IoT application and use the data to parameterize and validate our models. We also conduct sensitivity analysis over a variety of system parameters and examine the performance of larger networks