HGMSD: Hypergraph-Guided Microservice Deployment with Reinforcement Learning

This repository provides the implementation of HGMSD, a hypergraph-based microservice deployment framework designed for multi-cloud environments. HGMSD models service dependencies, resource states, and network topology as a unified hypergraph, and optimizes service deployment via deep reinforcement learning to reduce image pull latency and cross-subnet communication overhead.

⚙️ Environment Setup

• Python 3.7.6
• TensorFlow 2.8.0
• Numpy 1.19.5

Install dependencies:

pip install tensorflow==2.8.0 numpy==1.19.5

🚀 Usage

To train HGMSD:

python hypergraph_main.py

To evaluate HGMSD with image pull latency:

python hypergraph_pull_main.py

Baselines can be run via:

python RMS_DDPG_main.py
python RMS_pull_main.py
python rsdql_main_pull.py

📈 Results

Key performance metrics:

• Response Delay
• Image Pull Delay
• Total Delay

The experimental results are visualized in the figure/ folder after execution.

📌 Future Work

We plan to enhance the generalization and practical applicability of the framework by formalizing all modeling assumptions and defining quantitative benchmarks.

📜 License

This project is for academic research purposes only. For other usage, please contact the authors.