Skip to main content

​
Narwhal and Tusk: A DAG-based Mempool and Efficient BFT Consensus

This paper introduces Narwhal, a high-throughput mempool protocol, and Tusk, an asynchronous Byzantine Fault Tolerant (BFT) consensus protocol, which together f

​
Abstract

This paper introduces Narwhal, a high-throughput mempool protocol, and Tusk, an asynchronous Byzantine Fault Tolerant (BFT) consensus protocol, which together form a modular and efficient system for distributed transaction processing. The core innovation lies in decoupling the data dissemination and ordering (mempool functionality) from the actual consensus logic. Narwhal organizes transactions into a Directed Acyclic Graph (DAG) structure, providing high-throughput data availability and reliable broadcast across replicas. This DAG-based mempool ensures that all honest nodes agree on a common set of available transactions before consensus takes place. Tusk then provides total ordering for these transactions through an asynchronous BFT protocol, which is resilient to network delays and partial synchrony assumptions. The combined system achieves significantly higher throughput and improved fault tolerance compared to traditional BFT systems, by optimizing the most expensive parts of consensus. This work offers a powerful design for building next-generation high-performance blockchain systems. Authors: George Danezis, Eleftherios Kokoris-Kogias, Alberto Sonnino, Alexander Spiegelman Publication: arXiv preprint Publication date: 2022-01-01

​
Key findings

  • The paper introduces Narwhal (DAG-based mempool) and Tusk (asynchronous BFT consensus) as a modular system.
  • Narwhal decouples transaction dissemination and ordering from the consensus mechanism, ensuring high-throughput data availability.
  • Tusk provides total ordering for transactions based on the data provided by Narwhal, robustly handling network delays.
  • The combined architecture achieves significantly higher throughput and improved fault tolerance for distributed systems.
  • This design offers a novel approach to building highly performant and scalable blockchain systems.

​
Citation

George Danezis, Eleftherios Kokoris-Kogias, Alberto Sonnino, Alexander Spiegelman (2022). Narwhal and Tusk: A DAG-based Mempool and Efficient BFT Consensus. arXiv preprint. https://arxiv.org/pdf/2105.11827.pdf
Canonical knowledge ID: research:narwhal-and-tusk-a-dag-based-mempool-and-efficient-bft-consensus