Rorqual: Speeding up Narwhal with TEEs

TL;DR

Rorqual targets the Narwhal Mempool in a partially synchronous network of $n \ge 3f+1$ peers, where up to $f$ may be Byzantine, by embedding protocol logic inside TEEs to speed vertex inclusion in the DAG and reduce communication, while offering privacy benefits against MEV. The main approach uses TEEs for enclaved execution with unbiased randomness and remote attestation, paired with Reed-Solomon erasure coding to disperse vertices so that $n-f$ acknowledgments suffice. The paper demonstrates latency improvements to a single message delay ($Δ$) in the best case, graceful degradation under faults, and strong accountability, along with compatibility with Shoal and Bullshark and resilience to adversarial scheduling. The results show notable gains in throughput and latency, with robust security properties and practical MEV implications for DAG-based blockchain systems.

Abstract

In this paper, we introduce Rorqual, a protocol designed to enhance the performance of the Narwhal Mempool by integrating Trusted Execution Environments (TEEs). Both Narwhal and Roqual are protocols based on a Directed Acyclic Graph (DAG). Compared to Narwhal, Rorqual achieves significant reductions in latency and increases throughput by streamlining the steps required to include a vertex in the DAG. The use of TEEs also reduces the communication complexity of the protocol while maintaining low computational costs. Through rigorous analysis, we demonstrate the protocol's robustness under both normal and adversarial conditions, highlighting its improvements in throughput, latency, and security.

Theorems & Definitions (26)

• lemma 1
• proof
• theorem 1
• proof
• theorem 2
• proof
• theorem 3
• proof
• lemma 2
• proof