Automating Staged Rollout with Reinforcement Learning

TL;DR

This paper proposes a multi-objective reinforcement learning approach to automate staged software rollout, balancing rapid deployment and risk mitigation by considering multiple stakeholder metrics.

Contribution

It introduces a novel multi-objective reinforcement learning framework specifically designed for automating staged rollout processes in software deployment.

Findings

Demonstrates effective balancing of deployment speed and failure risk

Shows potential for reducing downtime during rollout

Provides a dynamic, automated decision-making system

Abstract

Staged rollout is a strategy of incrementally releasing software updates to portions of the user population in order to accelerate defect discovery without incurring catastrophic outcomes such as system wide outages. Some past studies have examined how to quantify and automate staged rollout, but stop short of simultaneously considering multiple product or process metrics explicitly. This paper demonstrates the potential to automate staged rollout with multi-objective reinforcement learning in order to dynamically balance stakeholder needs such as time to deliver new features and downtime incurred by failures due to latent defects.