SkyNomad: On Using Multi-Region Spot Instances to Minimize AI Batch Job Cost

TL;DR

SkyNomad is a multi-region scheduling system that leverages heterogeneity in spot instance availability across regions to significantly reduce AI batch job costs while reliably meeting deadlines.

Contribution

It introduces a novel multi-region scheduling approach that exploits spatial and temporal heterogeneity in spot instances to minimize costs and ensure deadline compliance.

Findings

Achieves 1.25-3.96x cost savings in real deployments.

Performs within 10% of optimal in simulations.

Consistently meets deadlines despite spot instance unpredictability.

Abstract

AI batch jobs such as model training, inference pipelines, and data analytics require substantial GPU resources and often need to finish before a deadline. Spot instances offer 3-10x lower cost than on-demand instances, but their unpredictable availability makes meeting deadlines difficult. Existing systems either rely solely on spot instances and risk deadline violations, or operate in simplified single-region settings. These approaches overlook substantial spatial and temporal heterogeneity in spot availability, lifetimes, and prices. We show that exploiting such heterogeneity to access more spot capacity is the key to reduce the job execution cost. We present SkyNomad, a multi-region scheduling system that maximizes spot usage and minimizes cost while guaranteeing deadlines. SkyNomad uses lightweight probing to estimate availability, predicts spot lifetimes, accounts for migration cost, and unifies regional characteristics and deadline pressure into a monetary cost model that guides scheduling decisions. Our evaluation shows that SkyNomad achieves 1.25-3.96x cost savings in real cloud deployments and performs within 10% cost differences of an optimal policy in simulation, while consistently meeting deadlines.