Attention-Informed Surrogates for Navigating Power-Performance Trade-offs in HPC

TL;DR

This paper introduces a novel surrogate-assisted multi-objective Bayesian optimization framework that uses attention-based embeddings of job telemetry to efficiently navigate power-performance trade-offs in HPC scheduling, achieving better results with less data.

Contribution

It is the first to apply embedding-informed surrogates within a MOBO framework for HPC scheduling, improving Pareto front quality and reducing training costs.

Findings

Higher-quality Pareto fronts for runtime-power trade-offs.

Significant reduction in training costs.

Improved stability of optimization results.

Abstract

High-Performance Computing (HPC) schedulers must balance user performance with facility-wide resource constraints. The task boils down to selecting the optimal number of nodes for a given job. We present a surrogate-assisted multi-objective Bayesian optimization (MOBO) framework to automate this complex decision. Our core hypothesis is that surrogate models informed by attention-based embeddings of job telemetry can capture performance dynamics more effectively than standard regression techniques. We pair this with an intelligent sample acquisition strategy to ensure the approach is data-efficient. On two production HPC datasets, our embedding-informed method consistently identified higher-quality Pareto fronts of runtime-power trade-offs compared to baselines. Furthermore, our intelligent data sampling strategy drastically reduced training costs while improving the stability of the results. To our knowledge, this is the first work to successfully apply embedding-informed surrogates in a MOBO framework to the HPC scheduling problem, jointly optimizing for performance and power on production workloads.