Distributed Consensus Network: A Modularized Communication Framework and Reliability Probabilistic Analysis

TL;DR

This work introduces a modularized communication framework to analyze the reliability and latency of distributed consensus protocols operating over unreliable wireless links. By representing consensus communication as combinations of basic components (A/B/C) and defining activation probabilities, it yields a probabilistic model with a main theorem for consensus success $P_C$ and tractable approximations for the failure rate $P_F$, along with latency implications. It defines Reliability Gain and Tolerance Gain to quantify how joint failure rates and fault-tolerance levels influence performance, and demonstrates two latency-reduction strategies, including network scaling and optimized power allocation, validated by a RAFT prototype. The results offer theoretical guidance for designing low-failure, low-latency consensus systems in imperfect communication environments and establish a pathway for extending the framework to other protocols like PBFT and HotStuff.

Abstract

In this paper, we propose a modularized framework for communication processes applicable to crash and Byzantine fault-tolerant consensus protocols. We abstract basic communication components and show that the communication process of the classic consensus protocols such as RAFT, single-decree Paxos, PBFT, and Hotstuff, can be represented by the combination of communication components. Based on the proposed framework, we develop an approach to analyze the consensus reliability of different protocols, where link loss and node failure are measured as a probability. We propose two latency optimization methods and implement a RAFT system to verify our theoretical analysis and the effectiveness of the proposed latency optimization methods. We also discuss decreasing consensus failure rate by adjusting protocol designs. This paper provides theoretical guidance for the design of future consensus systems with a low consensus failure rate and latency under the possible communication loss.

Figures (8)

• Figure 1: The comparison of latency in the RAFT between the link loss rate of 0 and 0.05. The system consists of 6 followers and 1 leader with 2 followers crashed. The latency is calculated from the instance's arrival at the leader to its commitment by the leader. The single link latency is set as 0.1s. The instance arrival rate is set as 60 per second.
• Figure 2: Three Representative Communication Designs in Consensus Protocols
• Figure 3: Analytical result and the experimental result of consensus failure rate and latency under different link loss rates when $n=4$.
• Figure 4: Latency in the RAFT system when link loss rate is 0.07 and n is 4, 6, 8, and 10.
• Figure 5: The comparison of latency in RAFT system before optimization Method 2 and after optimization Method 2.

Theorems & Definitions (17)

• Definition 1: Activated Node
• Lemma 1
• Lemma 2
• Theorem 1
• Remark 1
• Remark 2
• Theorem 2
• Corollary 1
• Corollary 2
• Theorem 3