Ilmu Komputer & AI editorial
Hydrozoan: Latency-Adaptive DAG Consensus under Mixed Byzantine and Crash Faults
The core problem
DAG-based consensus protocols have become a leading design for high-throughput Byzantine fault-tolerant (BFT) replication. They achieve great throughput and reach the optimal three-message-delay limit for consensus among validators. However, two-delay protocols come at a cost: they either reduce resilience by requiring larger committees (e.g., -style quorums) or rely on fallback mechanisms that sacrifice the DAG's high throughput.
This paper introduces **Hydrozoan**, the first DAG protocol with a *dual commit path* under a hybrid fault model. The model tolerates Byzantine validators and crashed validators on validators, where is a tunable parameter that sizes the fast quorum. The central idea is that leaders commit in **two message delays** whenever at most validators are faulty, and in **three message delays** otherwise, without extra messages, without view changes, and while allowing multiple leaders per round. Both paths are evaluated on the same DAG, and a novel *graded indirect rule* reconciles them so that every honest validator reaches the same decision.
The authors also present **Optimal-Hydrozoan**, a variant that
Innovation
The geo-distributed evaluation demonstrates that Hydrozoan matches Mysticeti's throughput while committing approximately **25% faster** when the fast quorum fits fast regions. When the fast quorum does not fit, or after more than faults, Hydrozoan falls back to three message delays, whereas existing two-delay protocols stall.
Specifically, the paper reports that under geo-distributed conditions, the latency advantage of the fast path depends on geography: rounds that reach a remote region cost far more than those that do not. This means that the decision of which path is faster is not a property of the protocol but of the deployment's network topology. The knobs allow operators to configure the fast quorum to match the regions where low latency is required.
Optimal-Hydrozoan, the variant that tolerates one more fault on the fast path, is the first construction to match the known lower bound. The authors do not report a specific throughput number beyond matching Mysticeti, but they emphasize that the fallback to three message delays occurs without extra messages or view changes, preserving liveness and safety.
The machine-checked proofs in Lean 4 cover both Hydroz
Why it matters
Hydrozoan's dual commit path addresses a fundamental tension in DAG-based consensus: the trade-off between low latency and resilience. By introducing a hybrid fault model with Byzantine and crash faults, and a tunable parameter , the protocol allows deployments to choose where to place the fast quorum. This is particularly relevant for geo-distributed systems, where network latency varies significantly across regions.
The graded indirect rule is a novel mechanism that reconciles the two paths without requiring validators to run separate protocols or perform view changes. This is crucial for maintaining the DAG's high throughput and avoiding the stalls that plague existing two-delay protocols when faults exceed the fast-path threshold.
The machine-checked proofs in Lean 4 add a strong layer of assurance, which is increasingly important for consensus protocols deployed in adversarial environments. The taxonomy of the work spans Architecture, Cybersecurity, Network, and Cryptography, reflecting its interdisciplinary nature.
One limitation is that the evaluation focuses on geo-distributed conditions and comparison with Mysticeti; broader comparisons with other DAG protocols and adversarial network conditions could further validate the approach. Nevertheless, Hydrozoan represents a significant step toward latency-adaptive consensus that can be tuned to deployment-specific requirements.
Future work could explore dynamic adjustment of based on observed fault patterns, or integration with sharding and other scalability techniques.
Who should read this
Opening member contentโฆ