EcoShift: Performance-Aware Power Management for Power-Constrained Heterogeneous Systems

2026-04-19

AI Summary

This work addresses the challenge of efficient power reallocation in cluster environments under strict power budgets, where existing power management strategies fall short due to their neglect of the differing performance sensitivities of CPU- and GPU-bound applications to power caps. To overcome this limitation, we propose EcoShift, a novel framework that, for the first time, integrates application-level performance sensitivity to CPU and GPU power allocation into cluster-wide scheduling. EcoShift combines online performance prediction with a dynamic programming-based allocator to enable fine-grained, performance-aware power redistribution. Evaluated on heterogeneous platforms featuring Intel CPUs and NVIDIA A100/H100 GPUs, EcoShift consistently meets total power constraints while achieving up to 6% higher average performance compared to state-of-the-art approaches.

Technology Category

• Machine Learning: Hardware-aware ML
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Planning, Routing, and Scheduling: Applications

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Power-constrained HPC systems increasingly run heterogeneous CPU--GPU applications under strict cluster-wide power limits. Existing cluster-wide power management policies rely on fair-share or utilization heuristics and do not capture application-specific sensitivity to CPU and GPU power caps, leading to inefficient use of reclaimed power. We present EcoShift, a performance-aware cluster-wide power management framework. EcoShift combines online performance prediction with a dynamic-programming-based allocator to distribute reclaimed power across CPU--GPU applications for maximum average performance improvement. Through emulation-based evaluation on two heterogeneous Intel CPU and NVIDIA A100/H100 GPU platforms with diverse CPU--GPU workloads, EcoShift consistently outperforms state-of-the-art policies, achieving up to 6% average performance improvement while preserving the cluster-wide power constraint.

Problem

• power-constrained systems
• heterogeneous computing
• cluster-wide power management
• CPU-GPU applications
• performance sensitivity

Innovation

• performance-aware power management
• heterogeneous systems
• dynamic programming
• online performance prediction
• power capping