SustainDC: Benchmarking for Sustainable Data Center Control
🤖 AI Summary
Datacenter energy consumption exacerbates climate change, necessitating sustainable control strategies. Method: This paper introduces the first multi-agent reinforcement learning (MARL) benchmark platform for sustainable datacenter operations. It integrates thermodynamic modeling, time-varying grid carbon intensity, geographic–climatic diversity, and heterogeneous workloads to jointly optimize workload scheduling, cooling control, and battery management. Contribution/Results: We propose a unified, scalable, and tightly coupled MARL evaluation framework—addressing the lack of standardized benchmarks for sustainable intelligent operations. Implemented in Python and compatible with mainstream algorithms (e.g., MAPPO, QMix), the platform demonstrates significant reductions in energy consumption and carbon emissions across diverse real-world configurations. Empirical analysis further exposes critical limitations of existing methods in cross-domain generalization and multi-objective coordination. The benchmark provides a reproducible, comparable foundation for advancing green AI infrastructure.
Application Category
📝 Abstract
Machine learning has driven an exponential increase in computational demand, leading to massive data centers that consume significant amounts of energy and contribute to climate change. This makes sustainable data center control a priority. In this paper, we introduce SustainDC, a set of Python environments for benchmarking multi-agent reinforcement learning (MARL) algorithms for data centers (DC). SustainDC supports custom DC configurations and tasks such as workload scheduling, cooling optimization, and auxiliary battery management, with multiple agents managing these operations while accounting for the effects of each other. We evaluate various MARL algorithms on SustainDC, showing their performance across diverse DC designs, locations, weather conditions, grid carbon intensity, and workload requirements. Our results highlight significant opportunities for improvement of data center operations using MARL algorithms. Given the increasing use of DC due to AI, SustainDC provides a crucial platform for the development and benchmarking of advanced algorithms essential for achieving sustainable computing and addressing other heterogeneous real-world challenges.
Problem
Research questions and friction points this paper is trying to address.
Benchmarking MARL algorithms for sustainable data center control
Optimizing energy use in data centers via multi-agent reinforcement learning
Addressing climate impact of data centers through AI-driven solutions
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multi-agent reinforcement learning for data centers
Custom Python environments for benchmarking
Optimizes workload, cooling, and battery management
🔎 Similar Papers
No similar papers found.
