Mysticeti: Latency-Optimal DAG Consensus

Mysticeti represents the 2025 "state-of-the-art" in Byzantine Fault Tolerant (BFT) consensus. It achieves the theoretical minimum latency for ordering transactions in a decentralized network, significantly improving upon previous DAG-based designs like Narwhal and Bullshark.

1. Why the Paradigm Shift: From Certified to Uncertified

Before 2024, DAG-based consensus relied on Certified DAGs (e.g., Bullshark).

The Mysticeti Solution: Uncertified DAGs

Mysticeti removes the certification requirement. Nodes simply sign their own blocks and share them. The "voting" happens implicitly through the references in the DAG structure.

2. Theoretical Lower Bound: The 3-Round Proof

BFT theory proves that 3 message delays is the absolute lower bound for reaching consensus in an asynchronous environment with f failures.

  1. Round 1: Propose (Node shares its block).
  2. Round 2: Reference (Other nodes include the block in their DAG).
  3. Round 3: Commit (The DAG structure becomes deep enough to guarantee ordering).

3. Architecture: Mysticeti-C and Mysticeti-FPC

Modern implementations (notably in the Sui Network) use a hybrid approach to maximize both throughput and latency:


External Deep Dive:

See Also: