Briefing
The core problem in Byzantine Fault Tolerant (BFT) consensus operating under asynchronous network conditions is the high latency and throughput degradation imposed by the mandatory agreement stage within the Asynchronous Common Subset (ACS) sequence. Falcon addresses this by introducing the Graded Broadcast (GBC) protocol, a novel primitive that allows a block to be directly included in the ACS set, thereby circumventing the costly agreement phase. This fundamental architectural change, coupled with the Asymmetrical Asynchronous Binary Agreement (AABA) for safety and a partial-sorting mechanism for stability, establishes a new performance baseline for ABFT, implying that future decentralized architectures can achieve high throughput and low latency without sacrificing liveness in adversarial network environments.
Context
Traditional Asynchronous BFT (ABFT) protocols, which guarantee liveness even when network delays are unbounded, rely on the Asynchronous Common Sub-seQuence (ACSQ) framework. This framework mandates that every block must pass through an Asynchronous Common Subset (ACS) protocol, which itself contains a resource-intensive agreement stage. This design choice, while ensuring robustness against network manipulation like Distributed Denial of Service (DDoS) attacks, creates an inherent bottleneck, resulting in high latency, unstable block committing, and reduced throughput for all transactions.
Analysis
Falcon’s core mechanism centers on the Graded Broadcast (GBC) , which redefines the block inclusion process. In previous ABFT designs, a block required full, explicit agreement from the network to be considered finalized. GBC introduces a graded, or probabilistic, level of certainty for a block’s broadcast, allowing nodes to include a block in the consensus set before full, explicit agreement is reached. This is achieved by having nodes use the GBC to disseminate blocks with a certain “grade” of confidence.
The protocol then uses the new Asymmetrical Asynchronous Binary Agreement (AABA) only as a fallback or final check to ensure the network maintains safety and consistency, conceptually moving the consensus bottleneck from a synchronous, required-for-every-block step to an asynchronous, as-needed safety layer. This decoupling of dissemination from final agreement fundamentally reduces the good-case latency.
Parameters
- Latency Benchmark (sMVBA) → $6delta$ latency in the good case. This represents the minimum time required for the state-of-the-art Multi-valued Byzantine Agreement protocol to reach consensus, where $delta$ is the maximum network delay.
- Communication Complexity (Previous) → $O(n^2)$ communication complexity in both good and bad cases. This is the message complexity of the state-of-the-art sMVBA protocol, where $n$ is the number of nodes.
- Throughput Improvement → Enhanced throughput. The new agreement trigger allows nodes to wait for more blocks before committing, which reduces block discarding and increases the overall transaction rate.
Outlook
The introduction of the Graded Broadcast primitive establishes a new direction for asynchronous protocol design, prioritizing low-latency liveness under adversarial conditions. Future research will focus on formally proving the tightest possible latency bounds for the GBC/AABA combination and exploring its application in other distributed system contexts, such as decentralized sequencing for Layer 2 rollups. The ability to achieve high throughput and low, stable latency in a fully asynchronous environment is a critical step toward realizing global-scale, resilient, and censorship-resistant decentralized applications in the next three to five years.
Verdict
Falcon’s Graded Broadcast is a foundational re-architecture of Asynchronous BFT, successfully circumventing the inherent latency trade-off to deliver robust liveness and efficiency.
