[Paper Review] Dynamic Practical Byzantine Fault Tolerance and Its Blockchain System: A Large-Scale Markov Modeling

Yan-Xia Chang, Quan‐Lin Li+2 more|arXiv (Cornell University)|Oct 25, 2022

Blockchain Technology Applications and Security4 citations

TL;DR

This paper proposes a dynamic Practical Byzantine Fault Tolerance (PBFT) consensus mechanism where voting nodes dynamically join and leave the network, modeling the system as a large-scale multi-dimensional Quasi-Birth-and-Death (QBD) process. Using RG-factorization and Markov modeling, it develops efficient algorithms to compute throughput and performance metrics, revealing trade-offs between security and scalability under varying node dynamics and failure rates.

ABSTRACT

In a practical Byzantine fault tolerance (PBFT) blockchain network, the voting nodes may always leave the network while some new nodes can also enter the network, thus the number of voting nodes is constantly changing. Such a new PBFT with dynamic nodes is called a dynamic PBFT. Clearly, the dynamic PBFT can more strongly support the decentralization and distributed structure of blockchain. However, analyzing dynamic PBFT blockchain systems will become more interesting and challenging. In this paper, we propose a large-scale Markov modeling technique to analyze the dynamic PBFT voting processes and its dynamic PBFT blockchain system. To this end, we set up a large-scale Markov process (and further a multi-dimensional Quasi-Birth-and-Death (QBD) process) and provide performance analysis for both the dynamic PBFT voting processes and the dynamic PBFT blockchain system. In particular, we obtain an effective computational method for the throughput of the complicated dynamic PBFT blockchain system. Finally, we use numerical examples to check the validity of our theoretical results and indicate how some key system parameters influence the performance measures of the dynamic PBFT voting processes and of the dynamic PBFT blockchain system. Therefore, by using the theory of multi-dimensional QBD processes and the RG-factorization technique, we hope that the methodology and results developed in this paper shed light on the study of dynamic PBFT blockchain systems such that a series of promising research can be developed potentially.

Motivation & Objective

• To address the challenge of modeling PBFT consensus in dynamic networks where nodes continuously join and leave, increasing decentralization but complicating performance analysis.
• To develop a scalable stochastic model that captures the dynamic behavior of voting nodes and transaction processing in PBFT-based blockchains.
• To provide computationally efficient methods for evaluating key performance metrics, especially throughput, in dynamic PBFT systems.
• To analyze the impact of critical system parameters—such as node arrival/departure rates, voting speed, and Byzantine failure rates—on system performance and security.
• To lay a theoretical foundation using multi-dimensional QBD processes and RG-factorization for future research on adaptive and secure blockchain consensus mechanisms.

Proposed method

• Model the dynamic PBFT system as a large-scale continuous-time Markov process with state space reflecting the number of active voting nodes and transaction states.
• Formulate the system as a multi-dimensional Quasi-Birth-and-Death (QBD) process to capture the complex transitions between states during the prepare, commit, and reply phases.
• Apply the RG-factorization technique to solve the steady-state distribution of the QBD process, enabling efficient computation of performance metrics.
• Derive two effective algorithms for computing the transaction throughput of the dynamic PBFT blockchain system based on the solution of the QBD process.
• Integrate system parameters such as node arrival rate (p), departure rate (β), and voting rate (γ) into the Markov model to analyze their impact on performance.
• Use numerical experiments to validate theoretical results and illustrate the sensitivity of throughput to key system parameters.

Experimental results

Research questions

• RQ1 How does dynamic node churn (arrival and departure of voting nodes) affect the throughput and stability of PBFT-based blockchain systems?
• RQ2 What is the impact of varying node arrival and departure rates on the performance and security of dynamic PBFT consensus mechanisms?
• RQ3 How do different transaction processing rates (e.g., prepare, commit, reply) influence the overall system throughput in a dynamic environment?
• RQ4 What trade-offs exist between system throughput and security when increasing the number of voting nodes or Byzantine node ratios?
• RQ5 Can the RG-factorization technique be effectively applied to solve the steady-state distribution of large-scale multi-dimensional QBD processes in dynamic PBFT systems?

Key findings

• Throughput (TH) increases with higher node arrival rates (p), indicating that faster onboarding of honest nodes improves system efficiency.
• Throughput increases with higher block-pegging or rollback rates (β), showing that faster recovery from failures enhances transaction processing speed.
• There exists a critical threshold (p̃₀) such that increasing the voting rate (γ) decreases throughput when p < p̃₀, but increases it when p > p̃₀, revealing a non-monotonic performance trade-off.
• The model reveals a fundamental trade-off: higher throughput is achievable with fewer nodes, but this reduces security if the proportion of Byzantine nodes is high.
• Numerical validation confirms that the theoretical throughput model aligns with simulation results across varying parameter settings, confirming the accuracy of the proposed algorithms.
• The RG-factorization-based solution method enables efficient computation of performance metrics in large-scale dynamic PBFT systems, making real-time analysis feasible.