Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mahi-Mahi: Low-Latency Asynchronous BFT DAG-Based Consensus

Philipp Jovanovic, Lefteris Kokoris Kogias, Bryan Kumara, Alberto Sonnino, Pasindu Tennage, Igor Zablotchi

TL;DR

Mahi is presented, the first asynchronous BFT consensus protocol that achieves sub-second latency in a wide-area network setting while processing over 100,000 transactions per second, and its performance is compared to state-of-the-art asynchronous consensus protocols, showcasing Mahi-Mahi’s significantly lower latency.

Abstract

We present Mahi-Mahi, the first asynchronous BFT consensus protocol that achieves sub-second latency in the WAN while processing over 100,000 transactions per second. We accomplish this remarkable performance by building Mahi-Mahi on an uncertified structured Directed Acyclic Graph (DAG). By forgoing explicit certification, we significantly reduce the number of messages required to commit and minimize CPU overhead associated with certificate verification. Mahi-Mahi introduces a novel commit rule that allows committing multiple blocks in each DAG round, while ensuring liveness in the presence of an asynchronous adversary. Mahi-Mahi can be parametrized to either attempt to commit within 5 message delays, maximizing the probability of commitment under a continuously active asynchronous adversary, or within 4 message delays, which reduces latency under a more moderate and realistic asynchronous adversary. We demonstrate the safety and liveness of Mahi-Mahi in a Byzantine context. Subsequently, we evaluate Mahi-Mahi in a geo-replicated setting and compare its performance against state-of-the-art asynchronous consensus protocols, showcasing Mahi-Mahi's significantly lower latency.

Mahi-Mahi: Low-Latency Asynchronous BFT DAG-Based Consensus

TL;DR

Mahi is presented, the first asynchronous BFT consensus protocol that achieves sub-second latency in a wide-area network setting while processing over 100,000 transactions per second, and its performance is compared to state-of-the-art asynchronous consensus protocols, showcasing Mahi-Mahi’s significantly lower latency.

Abstract

We present Mahi-Mahi, the first asynchronous BFT consensus protocol that achieves sub-second latency in the WAN while processing over 100,000 transactions per second. We accomplish this remarkable performance by building Mahi-Mahi on an uncertified structured Directed Acyclic Graph (DAG). By forgoing explicit certification, we significantly reduce the number of messages required to commit and minimize CPU overhead associated with certificate verification. Mahi-Mahi introduces a novel commit rule that allows committing multiple blocks in each DAG round, while ensuring liveness in the presence of an asynchronous adversary. Mahi-Mahi can be parametrized to either attempt to commit within 5 message delays, maximizing the probability of commitment under a continuously active asynchronous adversary, or within 4 message delays, which reduces latency under a more moderate and realistic asynchronous adversary. We demonstrate the safety and liveness of Mahi-Mahi in a Byzantine context. Subsequently, we evaluate Mahi-Mahi in a geo-replicated setting and compare its performance against state-of-the-art asynchronous consensus protocols, showcasing Mahi-Mahi's significantly lower latency.

Paper Structure

This paper contains 23 sections, 25 theorems, 2 equations, 7 figures, 3 algorithms.

Table of Contents

  1. Introduction
  2. System Overview
  3. Threat model, goals, and assumptions
  4. Intuition behind the Mahi-Mahi design
  5. Structure of the Mahi-Mahi DAG
  6. The Mahi-Mahi Protocol
  7. Proposers and anchors
  8. The Mahi-Mahi decision rule
  9. Implementation
  10. Evaluation
  11. Experimental Setup
  12. Benchmark under ideal conditions
  13. Performance under faults
  14. Impact of the number of leader slots per round
  15. Related Work
  16. ...and 8 more sections

Key Result

lemma 1

If in round $r$, $2f+1$ blocks from distinct validators certify a block $b$, then all blocks at future rounds $r'>r$ will have a path to a certificate for $b$ from round $r$.

Figures (7)

  • Figure 1: The structure of the Mahi-Mahi DAG. Left: The structure of a wave, consisting of 5 rounds (Propose, Boost, Boost, Vote, Certify). Right: Waves patterns in the Mahi-Mahi protocol (each round starts a new overlapping wave).
  • Figure 2: Example execution with 4 validators, wave length of 5 rounds and 2 leader slots per round.
  • Figure 3: Comparative throughput-latency performance of Mahi-Mahi, Tusk, and Cordial Miners. WAN measurements with $10$ and $50$ validators. No validator faults. $512$B transaction size.
  • Figure 4: Comparative throughput-latency of Mahi-Mahi, Tusk, and Cordial Miners. WAN measurements with 10 validators. Three faults. 512B transaction size.
  • Figure 5: Impact of the number of leaders per round in Mahi-Mahi. WAN measurements with 10 validators. Zero and three faults. 512B transaction size.
  • ...and 2 more figures

Theorems & Definitions (53)

  • lemma 1
  • proof
  • proof
  • lemma 2
  • proof
  • proof
  • proof
  • lemma 3
  • proof
  • lemma 4
  • ...and 43 more