FYP (For Your Ping)

An LLM-powered SDN controller that translates natural language intents into automated network configurations with real-time monitoring and optimized resource allocation.

Comment

Inspiration

Configuring Software-Defined Networking (SDN) policies can be tedious and error-prone, often requiring deep technical expertise. With the growing adoption of intent-based networking, I was inspired to explore how Large Language Models (LLMs) could bridge the gap between human-friendly instructions and low-level network configurations. The vision is to make network resource allocation more intuitive, efficient, and scalable.

What it does

FYP (For Your Ping) is a work-in-progress system that:

  • Lets users express high-level intents like “prioritize video traffic” in natural language.
  • Translates these intents into ONOS Intent Framework JSON configurations.
  • Simulates the results in a Mininet-based virtual network.
  • Provides a dashboard for real-time monitoring of traffic, devices, and policies.
  • Lays the groundwork for optimization algorithms that balance bandwidth, routing, and load.

How we built it

  1. Architecture Design
    Designed a modular system with four main components:

    • NLP Module: Uses gpt-oss to parse natural language and generate structured intents.
    • Optimizer: Plans to allocate resources dynamically.
    • Controller Integration: Communicates with ONOS via REST APIs to push intents.
    • Visualization Dashboard: React-based UI to display network state and allow intent submission.

  2. Technology Stack

    • Backend: Python scripts for parsing, ONOS integration, and simulations.
    • Frontend: React, Tailwind CSS for a clean UI.
    • Network Simulation: Mininet and ONOS Controller.
    • Containerization: Docker for isolated testing and deployment.

  3. Early Prototyping
    Built static UI components, basic ONOS scripts, and tested prompt-based intent parsing.

Challenges we ran into

  • ONOS Learning Curve: Understanding its intent framework and APIs took significant time.
  • Complex Networking Setup: Simulating multi-node topologies in Mininet while ensuring ONOS communication.
  • LLM Integration: Experimenting with prompt design for accurate translation of network intents.
  • Time Constraints: Couldn’t complete full optimization logic and end-to-end integration.

Accomplishments that we're proud of

  • Clear system architecture design showing feasibility of LLM-driven SDN resource management.
  • Initial React dashboard prototype for intuitive intent submission and network monitoring.
  • Successfully connected ONOS and Mininet to control simulated network flows.
  • Learned to containerize components for easier testing.

What we learned

  • Deepened understanding of SDN concepts, ONOS architecture, and Mininet simulation.
  • Gained experience in LLM integration and prompt engineering for domain-specific tasks.
  • Learned to design and structure large projects under strict time limits.
  • Improved knowledge of Docker networking and REST API communication.

What's next for FYP (For Your Ping)

  • Complete end-to-end integration: Full pipeline from natural language intent → ONOS JSON → live simulation.
  • Develop optimization algorithms for smarter bandwidth, routing, and load balancing.
  • Add error handling and intent validation to ensure safe network changes.
  • Implement metrics visualization (latency, throughput, fairness) in the dashboard.
  • Explore fine-tuning gpt-oss with SDN-specific datasets for higher accuracy.
  • Deploy a fully containerized, reproducible version for public testing.

Built With

Share this project:

Updates