Update (Dec 2025): We added support for load balancing over the QUIC protocol.

---

P4Kube: In-Network Load-Aware Load Balancer for Kubernetes

P4Kube: In-Network Load Balancer for Kubernetes
Read the conference paper here

Citation

If you use P4Kube in your research, please cite our paper:

title={P4Kube: In-Network Load Balancer for Kubernetes},
author={Grigoryan, Garegin and Penkowski, Kevin and Kwon, Minseok},
booktitle={2025 IEEE 22nd Consumer Communications & Networking Conference (CCNC)},
pages={1--6},
year={2025},
organization={IEEE}
}

---

Getting Started

To run the code and reproduce our experiments:

1. Access the FABRIC testbed
   Ensure you have a FABRIC account and access to JupyterLab on FABRIC.

2. Clone P4Kube_setup.ipynb inside JupyterLab on FABRIC:

   wget https://raw.githubusercontent.com/gareging/P4Kube/refs/heads/main/P4Kube_setup.ipynb
   

Launch the Notebook

Open P4Kube_setup.ipynb in JupyterLab on the FABRIC testbed and follow the step-by-step cells to

• reserve the slice,
• setup and configure networking using bmv2 P4 switch emulator,
• setup and configure the Kubernetes cluster.

Testing QUIC

On master, run

kubectl apply -f quic.yaml

On the client:

sudo apt install libssl-dev python3.9 python3.9-venv python3.9-dev -y;
sudo apt install python3-pip -y;
pip3 install aioquic wsproto;
git clone https://github.com/aiortc/aioquic.git;
cd aioquic/examples;
python3 http3_client.py --insecure https://<vip_of_quic_deployment>:4433/;

Experimenting

Clone P4Kube_experiments.ipynb inside JupyterLab on FABRIC:

wget https://raw.githubusercontent.com/gareging/P4Kube/refs/heads/main/P4Kube_experiments.ipynb

Follow the instructions cell-by-cell.

For any questions, please email to grigoryan@alfred.edu.