Stress-Testing PBFT and Tendermint Consensus for Blockchain-Enabled Smart Grids under Network Loss and Byzantine Leader Failures

Rahawee, Yazen Firas; Alsharbaty, Firas

doi:10.69513/njitcs.2026.175032.1082

	Stress-Testing PBFT and Tendermint Consensus for Blockchain-Enabled Smart Grids under Network Loss and Byzantine Leader Failures
Al-Noor Journal for Information Technology and Cybersecurity
Articles in Press, Corrected Proof, Available Online from 26 September 2026
Document Type: Original Article
DOI: 10.69513/njitcs.2026.175032.1082
Authors
Yazen Firas Rahawee^* ¹; Firas Alsharbaty²
¹Electrical department , Engineering college , University of Mosul, Mosul- Iraq
²ELECTRICAL DEPARTMENT , ENGINEERING COLLEGE , UNIVERSITY OF MOSUL , MOSUL-IRAQ
Abstract
Blockchain-based Advanced Metering Infrastructure (AMI) can reduce reliance on a single centralized data collector, but practical performance depends on how the consensus protocol behaves when the network is impaired or a leader becomes unavailable. This study compares Practical Byzantine Fault Tolerance (PBFT) and Tendermint in a controlled discrete-event simulator driven by London smart-meter data. A fixed population of 900 valid meter IDs and 19 validator nodes were used. Each reported AMI experiment was repeated over 30 independent runs with 300 sequential transactions per run. A separate leader-recovery benchmark used 10 paired fresh-state trials in each run. Under the baseline condition, with 5% packet loss per transmission attempt and uniformly distributed jitter from 0 to 15 ms, the run-level mean finality latency was 55.789 ms for PBFT and 55.822 ms for Tendermint. In the controlled recovery benchmark, validators 0 and 1 were configured as withholding Byzantine validators and occupied the first two leader positions under 15% packet loss and uniform jitter from 0 to 30 ms. The mean recovery latency overhead was 7042.67 ms for PBFT and 6649.15 ms for Tendermint, a 5.59% reduction, and the PBFT-minus-Tendermint run-level difference was positive in all 30 runs. Within the implemented logical-message model, the mean additional recovery-message count was 543.78 for PBFT and 1155.12 for Tendermint. The loss-free analytical normal-path model shows quadratic message growth for both protocols in the broadcast abstraction used here. The results indicate a trade-off: Tendermint had lower recovery latency under the tested leader-failure and timer configuration.
Keywords
Blockchain; Consensus Algorithms; Practical Byzantine Fault Tolerance (PBFT); Tendermint; Advanced Metering Infrastructure (AMI)

Statistics Article View: 9