Data Center Cooling Simulation Framework

A comprehensive open-source framework for simulating and optimizing data center cooling systems, combining AI workload simulation with thermal modeling and control strategies.

Overview

This repository provides a complete simulation environment for data center cooling optimization, featuring:

• AI Datacenter Simulation Engine: Large-scale AI training workload simulation with GPU-aware thermal analysis
• AlphaDataCenterCooling: Virtual testbed for evaluating data center cooling control strategies
• Telemetry & Monitoring: Prometheus and Grafana integration for real-time metrics
• Dashboard: Web-based visualization and control interface

Repository Structure

.
├── ai-datacenter-sim/          # AI workload simulation and telemetry stack
│   ├── SimAI/                  # SimAI large-scale training simulator
│   ├── simulation/             # Facility simulation (RDHx + CHW plant dynamics)
│   ├── adapters/               # Integration adapters for cooling systems
│   │   ├── alpha_adapter/      # AlphaDataCenterCooling adapter
│   │   └── simai_adapter/      # SimAI workload adapter
│   ├── monitoring/             # Prometheus and Grafana configuration
│   ├── dashboard/              # Frontend dashboard and API
│   └── telemetry/              # Telemetry ingestion utilities
│
├── AlphaDataCenterCooling/      # Virtual testbed for cooling system optimization
│   ├── AlphaDataCenterCooling_Gym/  # Gymnasium environment interface
│   ├── Resources/              # Model files and initialization data
│   └── docs/                   # Documentation and figures
│
└── [Documentation files]        # Project documentation and guides

Quick Start

Prerequisites

• Docker and Docker Compose
• Python 3.8+ (for local development)
• Git

Running the Simulation Stack

1. Clone the repository:

   git clone <repository-url>
   cd datacenter-cooling-sim

2. Configure environment variables:

   # Copy .env.example to ai-datacenter-sim directory
   cp .env.example ai-datacenter-sim/.env
   # Edit ai-datacenter-sim/.env and change Grafana admin credentials for production

3. Start the AI Datacenter Simulation:

   cd ai-datacenter-sim
   docker-compose up -d

   Note: Docker Compose automatically reads the .env file in the same directory. The Grafana credentials (GRAFANA_ADMIN_USER and GRAFANA_ADMIN_PASSWORD) are loaded from this file.

   This starts:

   • Prometheus (metrics collection) on port 9090
   • Grafana (visualization) on port 3000
   • AlphaDataCenterCooling service on port 5001
   • Alpha adapter on port 8085
   • SimAI adapter
   • Dashboard API on port 8001
   • Dashboard frontend on port 5174

4. Access the services:

   • Grafana Dashboard: http://localhost:3000 (default: admin/admin)
   • Prometheus: http://localhost:9090
   • Dashboard: http://localhost:5174
   • AlphaDataCenterCooling API: http://localhost:5001

Running AlphaDataCenterCooling Standalone

cd AlphaDataCenterCooling
docker-compose up

See AlphaDataCenterCooling/README.md for detailed usage instructions.

Documentation

• AI Datacenter Sim README: Detailed guide for the simulation engine
• Telemetry Documentation: Telemetry setup and usage
• AlphaDataCenterCooling README: Cooling system testbed documentation

Key Features

AI Datacenter Simulation

• SimAI integration for large-scale AI training workload simulation
• GPU-aware thermal modeling
• Network topology simulation
• Workload trace analysis

Cooling System Simulation

• AlphaDataCenterCooling virtual testbed
• Gymnasium-compatible environment for RL/control algorithms
• REST API for external integration
• Real-time disturbance updates

Monitoring & Visualization

• Prometheus metrics collection
• Grafana dashboards
• Custom web dashboard
• Real-time telemetry ingestion

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│   SimAI         │────▶│  SimAI Adapter   │────▶│  Pushgateway    │
│   Workloads     │     │                  │     │                 │
└─────────────────┘     └──────────────────┘     └────────┬────────┘
                                                           │
┌─────────────────┐     ┌──────────────────┐             │
│ AlphaDataCenter │────▶│  Alpha Adapter   │──────────────┼─────┐
│ Cooling         │     │                  │              │     │
└─────────────────┘     └──────────────────┘              │     │
                                                           ▼     ▼
                                                  ┌─────────────────┐
                                                  │   Prometheus     │
                                                  │   (Metrics DB)   │
                                                  └────────┬────────┘
                                                           │
                                    ┌──────────────────────┼──────────────────────┐
                                    ▼                      ▼                      ▼
                            ┌──────────────┐    ┌──────────────┐    ┌──────────────┐
                            │   Grafana    │    │ Dashboard API│    │   Dashboard   │
                            │              │    │              │    │   Frontend    │
                            └──────────────┘    └──────────────┘    └──────────────┘

Development

Building Components

Each component can be built independently:

# Build Alpha adapter
cd ai-datacenter-sim/adapters/alpha_adapter
docker build -t alpha-adapter .

# Build SimAI adapter
cd ai-datacenter-sim/adapters/simai_adapter
docker build -t simai-adapter .

# Build dashboard
cd ai-datacenter-sim/dashboard/frontend
npm install
npm run build

Running Tests

See individual component READMEs for testing instructions.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Third-party components have their own licenses:

• SimAI: Apache License 2.0
• Astra-Sim (Alibaba Cloud fork): MIT License
• AlphaDataCenterCooling: See AlphaDataCenterCooling/README.md for citation information

Citation

If you use this framework in your research, please cite:

@misc{datacenter-cooling-sim2025,
  title={Data Center Cooling Simulation Framework},
  author={Kardashev Labs},
  year={2025},
  url={https://github.com/kardashev-lab/datacenter-cooling-sim}
}

Third-Party Components

This framework builds upon the following open-source projects and research:

SimAI - Large-scale AI training simulation framework:

• Repository: https://github.com/aliyun/SimAI
• Paper: SimAI: Unifying Architecture Design and Performance Tuning for Large-Scale Large Language Model Training with Scalability and Precision (NSDI'25 Spring)
• License: Apache License 2.0
• Based on: Astra-Sim (https://github.com/astra-sim/astra-sim) extended by Alibaba Cloud

AlphaDataCenterCooling - Virtual testbed for data center cooling optimization:

• Repository: https://github.com/wfzheng/AlphaDataCenterCooling
• Citation:

@article{wu2025alphadatacentercooling,
  title={AlphaDataCenterCooling: A virtual testbed for evaluating operational strategies in data center cooling plants},
  author={Wu, S. and Zheng, W. and Wang, Z. and Chen, G. and Yang, P. and Yue, S. and Li, D. and Wu, Y.},
  journal={Applied Energy},
  volume={380},
  pages={125100},
  year={2025}
}

Support

For questions and issues:

• Open an issue on GitHub
• Check the documentation in each component's README
• Review the telemetry documentation for setup help

Acknowledgments

• SimAI Team (Alibaba Cloud) for the large-scale AI training simulation framework
  • Repository: https://github.com/aliyun/SimAI
  • Paper: NSDI'25 Spring
• Astra-Sim Team (Georgia Tech & Facebook) for the original simulation framework
  • Repository: https://github.com/astra-sim/astra-sim
• AlphaDataCenterCooling Authors for the cooling system testbed
  • Repository: https://github.com/wfzheng/AlphaDataCenterCooling
  • Paper: Applied Energy, 380, 125100 (2025)
• The open-source community for excellent tools and libraries