Byzantine-Tolerant Consensus in GPU-Inspired Shared Memory

2025-03-17

AI Summary

Modeling and implementing Byzantine fault-tolerant (BFT) consensus in heterogeneous parallel systems under shared-memory semantics remains challenging, particularly due to the mismatch between conventional synchronization assumptions and GPU-accelerated weak memory consistency. Method: This work formally integrates GPU-style weak memory semantics into distributed consensus theory—departing from classical synchronous or asynchronous shared-memory models—and designs the first provably correct BFT consensus algorithm for shared memory, guaranteeing deterministic agreement under the optimal resilience condition (n geq 3f + 1). Contribution/Results: The algorithm achieves (O(n^2)) message complexity—significantly reducing communication overhead compared to classical protocols—while ensuring full safety and liveness. Its correctness is rigorously verified using TLA+. The key innovation lies in establishing a GPU-inspired shared-memory model that jointly optimizes fault tolerance and high-concurrency efficiency.

Technology Category

• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Machine Learning: Hardware-aware ML

Application Category

• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

In this work, we formalize a novel shared memory model inspired by the popular GPU architecture. Within this model, we develop algorithmic solutions to the Byzantine Consensus problem.

Problem

• Formalize GPU-inspired shared memory model
• Solve Byzantine Consensus problem
• Develop algorithmic solutions within model

Innovation

• GPU-inspired shared memory model
• Byzantine Consensus problem solutions
• Algorithmic development in shared memory