New framework uses multi-agent RL for carbon-aware AI data centers

Researchers have developed a hierarchical carbon-aware multi-agent reinforcement learning (CA-MARL) framework to manage energy consumption in AI data centers (AIDCs). This framework aims to reduce carbon emissions by optimizing AI training and inference workloads. The system includes a workload manager agent for spatial allocation of jobs across AIDCs and local AIDC agents for temporal job shifting, GPU allocation, and cooling system control. AI

IMPACT This framework could lead to more sustainable AI infrastructure by optimizing energy usage and reducing the carbon footprint of AI operations.

RANK_REASON The cluster contains a research paper detailing a new framework for AI data center energy management. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• AI Data Centers
• Hierarchical Multi-Agent Reinforcement Learning for Carbon-Aware AI Data Centers in Power Distribution Systems
• IEEE 33-node power distribution system
• local AIDC agents
• National Cancer Institute
• workload manager (WM) agent

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →