Performance and Resource Trade-Offs of Blockchain Consensus Algorithms in Blockchain-Based IIoT Forensic Evidence Management

Abstract

Blockchain-based forensic evidence management for Industrial Internet of Things (IIoT) environments requires tamper-evident custody records and accountable interorganizational access. Yet, Hyperledger Fabric evaluations often use generic workloads or treat retrieval as a passive query, limiting evidence on ordering behavior for audit-bearing transactions. This study compares Raft and SmartBFT in Fabric 3.1 through submitEvidence metadata registration and retrieveEvidenceAndLog accountable access. Hyperledger Caliper experiments vary batch/load, inter-orderer RTT, orderer count, and crash-stop orderer unavailability, evaluating outcome and successful throughput, successful-completion latency, transaction outcomes, CPU, memory, and orderer traffic. At batch 300 and 200 TPS, Raft provides approximately 22.4% and 18.8% higher successful throughput and 89.4% and 90.6% lower latency for submit and retrieve, while SmartBFT records 35.2% and 30.3% higher traffic per successful transaction. At ten orderers, Raft records 90.5% and 91.5% lower aggregate orderer CPU/TPS. SmartBFT is more sensitive to the 100 ms RTT profile and shows larger memory peaks under stress. Crash-stop behavior is heterogeneous and operation-dependent, with unavailable-node placement, timing, and protocol-state evolution affecting interpretation. Throughput and latency alone are therefore insufficient. Raft is suitable under strong administrative trust and crash-fault assumptions, whereas SmartBFT is appropriate when independently administered orderers face credible Byzantine threats.