Briefing
The foundational challenge in Byzantine Fault Tolerance (BFT) protocols involves the inherent trade-off between achieving low transaction latency, typically seen in leader-driven systems, and maximizing system throughput, which is characteristic of Data-Acyclic Graph (DAG) architectures. The Angelfish protocol addresses this by introducing a new consensus design that operates across the Leader-DAG spectrum. This breakthrough is realized by strategically decoupling the consensus pacing mechanism from the asynchronous data dissemination layer.
The protocol ensures that nodes can progress to a new round as soon as they obtain an availability certificate, bypassing the bottleneck of leader-driven data serialization. The single most important implication is the realization of a consensus architecture that achieves theoretically near-optimal latency while simultaneously maintaining maximal throughput, thereby resolving a critical performance constraint in high-speed decentralized systems.
Context
Prior to this research, BFT consensus protocols were categorized by a fundamental architectural choice → protocols like HotStuff prioritize minimal latency by using a single, rotating leader to serialize block proposals, which inherently limits throughput due to communication overhead; conversely, DAG-based protocols like Sailfish achieve high throughput by enabling parallel data dissemination but often incur higher latency due to the complexity of the ordering mechanism. This established limitation forced system architects to choose between a fast-finality system and a high-volume system, creating a persistent scalability trilemma constraint in the performance dimension.
Analysis
Angelfish’s core mechanism is a hybrid protocol built upon a certified DAG structure. Conceptually, the protocol introduces a distinct separation of concerns. The DAG layer handles the asynchronous, parallel dissemination of all transaction data, maximizing network bandwidth utilization. The leader’s role is strategically reduced to only pacing the consensus process by proposing a sequence of DAG vertices, which are already available across the network via the DAG.
This is fundamentally different from previous leader-based approaches where the leader was responsible for both proposal and data dissemination. By leveraging the certified DAG to guarantee data availability before consensus finality, Angelfish achieves optimal latency in the common case because the finality step (the leader’s proposal) does not require a subsequent data fetch, which eliminates the extra message delay.
Parameters
- Optimal Latency Baseline → Approximately twice the $3delta$ theoretical lower bound for leader-based protocols.
- Peak Throughput Metric → Matches the maximum observed throughput of the Sailfish protocol.
- Architectural Model → Partially-synchronous certified DAG consensus.
Outlook
This research opens new avenues for architecting decentralized systems where performance is paramount, specifically enabling next-generation financial settlement layers and high-frequency decentralized exchanges. The ability to achieve optimal latency and throughput simultaneously suggests a path toward BFT systems that can rival centralized counterparts in speed without compromising decentralization. The next logical step involves the formal analysis and deployment of the Multi-leader Angelfish extension, which aims to further decentralize the consensus pacing and validate the protocol’s robustness under real-world network instability and high load conditions over a multi-year horizon.
Verdict
The Angelfish protocol provides a foundational blueprint for future BFT architectures by demonstrating the theoretical and practical feasibility of simultaneously achieving optimal latency and maximal throughput in a decentralized system.
