CEI-Net: An Open-Source Framework for P4-Driven Network Management on the CEI Continuum

This paper presents a Docker and Ansible-based framework for deploying and analyzing network topologies, focusing on performance metrics such as latency, throughput, and packet loss. By leveraging containerized environments, the proposed approach ensures lightweight, scalable, and efficient network emulation while maintaining flexibility in software deployment. The network efficiency and congestion were evaluated under varying payload sizes and architectural conditions. The results show that the cloud-based single topology achieves the lowest latency and highest throughput, making it ideal for centralized data-processing applications. The tree topology (edge hierarchy) introduces moderate latency owing to hierarchical forwarding, whereas the spine-leaf topology (distributed IoT) experiences higher latency variations and retransmissions, indicating increased congestion during large-scale deployments. Existing CEI Continuum frameworks that employ QEMU-based VMs and P4-based Mininet testbeds introduce higher latency and packet loss owing to infrastructure constraints and processing overhead. The results highlight the advantages of containerized network simulations for real-time high-performance networking applications, making them particularly suitable for edge computing and Internet of Things environments. To validate the scalability and reproducibility of our deployment, we implemented the framework on Grid'5000 using Enoslib, which provides programmable access to the bare-metal servers. This architecture enables the deployment of realistic, high-performance networking environments under controlled experimental conditions, supporting the extensive emulation of edge and cloud scenarios across distributed infrastructures.