Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BALLAST: Bandit-Assisted Learning for Latency-Aware Stable Timeouts in Raft

Qizhi Wang

TL;DR

...

Abstract

Randomized election timeouts are a simple and effective liveness heuristic for Raft, but they become brittle under long-tail latency, jitter, and partition recovery, where repeated split votes can inflate unavailability. This paper presents BALLAST, a lightweight online adaptation mechanism that replaces static timeout heuristics with contextual bandits. BALLAST selects from a discrete set of timeout "arms" using efficient linear contextual bandits (LinUCB variants), and augments learning with safe exploration to cap risk during unstable periods. We evaluate BALLAST on a reproducible discrete-event simulation with long-tail delay, loss, correlated bursts, node heterogeneity, and partition/recovery turbulence. Across challenging WAN regimes, BALLAST substantially reduces recovery time and unwritable time compared to standard randomized timeouts and common heuristics, while remaining competitive on stable LAN/WAN settings.

BALLAST: Bandit-Assisted Learning for Latency-Aware Stable Timeouts in Raft

TL;DR

...

Abstract

Randomized election timeouts are a simple and effective liveness heuristic for Raft, but they become brittle under long-tail latency, jitter, and partition recovery, where repeated split votes can inflate unavailability. This paper presents BALLAST, a lightweight online adaptation mechanism that replaces static timeout heuristics with contextual bandits. BALLAST selects from a discrete set of timeout "arms" using efficient linear contextual bandits (LinUCB variants), and augments learning with safe exploration to cap risk during unstable periods. We evaluate BALLAST on a reproducible discrete-event simulation with long-tail delay, loss, correlated bursts, node heterogeneity, and partition/recovery turbulence. Across challenging WAN regimes, BALLAST substantially reduces recovery time and unwritable time compared to standard randomized timeouts and common heuristics, while remaining competitive on stable LAN/WAN settings.
Paper Structure (81 sections, 1 equation, 11 figures, 22 tables, 1 algorithm)

This paper contains 81 sections, 1 equation, 11 figures, 22 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Motivating observation.
  3. Where the idea comes from.
  4. Problem Setting and Metrics
  5. Why these metrics.
  6. Background
  7. Raft election timeouts
  8. Why static ranges are brittle.
  9. Failure detection and tail latency
  10. Heuristics vs. optimization.
  11. Contextual bandits for online control
  12. Why bandits fit timeout selection.
  13. Design: BALLAST
  14. Action space
  15. Data-driven arm scaling.
  16. ...and 66 more sections

Figures (11)

  • Figure 1: Main scenario (30 seeds): mean recovery time with 95% CI. Lower is better.
  • Figure 2: Main scenario (30 seeds): unwritable fraction with 95% CI. Lower is better.
  • Figure 3: Main scenario: CDF of conditional election latency (time_to_leader).
  • Figure 4: Real etcd/raft prototype: mean recovery time (95% CI), main scenario.
  • Figure 5: Real etcd/raft prototype: unwritable fraction (95% CI), main scenario.
  • ...and 6 more figures