Byzantine Consensus in the Random Asynchronous Model

TL;DR

This work introduces the random asynchronous model, a relaxation of the classic asynchronous setting that eliminates adversarial scheduling while retaining unbounded delays and Byzantine faults. It demonstrates Byzantine consensus feasibility under thresholds $n=3f+1$, $n=2f+1$, and $n=f+2$ using deterministic correct-process logic and probabilistic progress, accompanied by impossibility results that map the model’s limits. The key approach replaces adversarial scheduling with a random scheduling mechanism that ensures progress with positive probability, placing the model between fully asynchronous and synchronous paradigms and offering insights distinct from partial synchrony. The contributions include formal model definitions, protocol constructions for different resilience regimes, and bounds that clarify which tasks become solvable under non-adversarial random scheduling. The practical impact lies in providing a theoretical foundation for observed non-adversarial scheduling behavior in real-world BFT deployments and guiding protocol design under probabilistic timing assumptions. All results are stated with explicit probabilistic guarantees and thresholds, emphasizing the nuanced landscape between classic asynchronous and synchronous models.

Abstract

We propose a novel relaxation of the classic asynchronous network model, called the random asynchronous model, which removes adversarial message scheduling while preserving unbounded message delays and Byzantine faults. Instead of an adversary dictating message order, delivery follows a random schedule. We analyze Byzantine consensus at different resilience thresholds ($n=3f+1$, $n=2f+1$, and $n=f+2$) and show that our relaxation allows consensus with probabilistic guarantees which are impossible in the standard asynchronous model or even the partially synchronous model. We complement these protocols with corresponding impossibility results, establishing the limits of consensus in the random asynchronous model.