Digital Twin for Software Defined Networks (DT4SDN)

---

---

Prof. Granelli Fabrizio

Group: De Marco Matthew, Lo Iacono Andrea, Revrenna Jago

Table of Contents

• Digital Twin for Software Defined Networks (DT4SDN)
  • Idea of the Project
  • Architecture Overview
  • Requirements
    • Hardware Requirements
    • Software Requirements
  • Getting Started
    • Option 1: Windows Setup (Vagrant)
    • Option 2: macOS Setup (Multipass)
  • Running the System
    • Step 1: Start the Physical Twin
    • Step 2: Start the Digital Twin
    • Step 3: Generate Traffic
  • ML Models
  • Project Structure
  • API Reference
  • Links
  • Team Members
  • Acknowledgments
  • License

Idea of the Project

---

This project implements a Digital Twin for monitoring and predicting traffic behavior in Software Defined Networks (SDN). The system consists of:

1. Physical Twin (PT): A Mininet network with OpenFlow switches controlled by a Ryu controller
2. Digital Twin (DT): A real-time monitoring dashboard with ML-based traffic prediction

The Digital Twin continuously polls the Physical Twin for traffic statistics, stores them in a SQLite database, and uses trained Seq2Seq LSTM models to predict future traffic patterns 60 seconds ahead.

Key Features:

• Real-time traffic monitoring via web dashboard
• ML-based traffic state classification (NORMAL, ELEVATED, HIGH, CRITICAL)
• Multi-scenario model support (Normal, Burst, Congestion, DDoS, Mixed)
• Live prediction graph visualization
• Scenario-based model switching from UI

Architecture Overview

---

+---------------------------+          REST API          +---------------------------+
|      PHYSICAL TWIN        |  <--------------------->   |       DIGITAL TWIN        |
|---------------------------|                            |---------------------------|
| Mininet Network           |                            | Orchestrator              |
| OpenFlow Switches         |                            | Flask Web Dashboard       |
| Ryu Controller            |                            | SQLite Database           |
|                           |                            | PyTorch ML Models         |
+---------------------------+                            +---------------------------+
     192.168.56.101                                           192.168.56.102

Requirements

---

Hardware Requirements

• Host Machine: Windows 10/11 or macOS
• RAM: Minimum 8GB (16GB recommended)
• CPU: Multi-core processor with virtualization support
• Disk: 30GB free space

Software Requirements

Windows:

• VirtualBox 6.1+
• Vagrant 2.3+

macOS:

• Multipass 1.10+

Both VMs require:

• Ubuntu 20.04
• ComNetsEmu framework (official repo)
• Python 3.8+
• PyTorch, Flask, NumPy, Scikit-learn

Important

For ComNetsEmu installation, refer to the official repository: https://git.comnets.net/public-repo/comnetsemu . ComNetsEmu installation is a fundamental and mandatory step to replicate our work in your machine. ComNetsEmu must be installed in both the VMs (the one used for the Physical Twin and the one used for the Digital Twin)

Getting Started

---

Option 1: Windows Setup (Vagrant)

1. Install Prerequisites

   • Install VirtualBox
   • Install Vagrant

2. Clone the Repository

   git clone https://github.com/MattDema/Networking_DT4SDN.git
   cd Networking_DT4SDN

3. Start the Digital Twin VM

   vagrant up comnetsemu_dt

4. SSH into the VM

   vagrant ssh comnetsemu_dt

5. Set up Python environment (inside VM)

   cd ~/Networking_DT4SDN/setups
   chmod +x setup_vm.sh
   ./setup_vm.sh

Tip

The Vagrantfile automatically installs ComNetsEmu from the official repo, clones this repository, and sets up dependencies.

---

Option 2: macOS Setup (Multipass)

Creating the VMs:

# Physical Twin VM
multipass launch 20.04 --name comnetsemu-vm --cpus 1 --mem 1G --disk 5G

# Digital Twin VM  
multipass launch 20.04 --name digital-twin --cpus 2 --mem 4G --disk 20G

Setting up the Physical Twin:

multipass shell comnetsemu-vm

1. Install ComNetsEmu from the official repository
2. Install dependencies:

   pip3 install requests

3. Clone the repository:

   git clone https://github.com/MattDema/Networking_DT4SDN.git

Setting up the Digital Twin:

multipass shell digital-twin

1. Install ComNetsEmu from the official repository
2. Install dependencies:

   sudo apt update
   sudo apt install python3-pip -y
   sudo pip3 install flask requests numpy scipy joblib torch scikit-learn pandas

3. Clone the repository:

   git clone https://github.com/MattDema/Networking_DT4SDN.git

Running the System

---

Step 1: Start the Physical Twin

Open two terminals in the Physical Twin VM:

Terminal 1 - Ryu Controller:

cd ~/Networking_DT4SDN/setups
chmod +x start_ryu.sh
./setups/start_ryu.sh

Terminal 2 - Mininet Topology:

cd ~/Networking_DT4SDN/src/utils
python3 topology_generator.py
# Follow prompts to create network (e.g., 3 switches, 6 hosts, tree topology)

---

Step 2: Start the Digital Twin

1. First, get the Physical Twin IP:

   # On host machine
   multipass list  # or check Vagrant network config

2. Update the PT_IP in the orchestrator:

   cd ~/Networking_DT4SDN/src/controllers
   nano orchestrator.py
   # Change: PT_IP = os.getenv('PT_IP', 'YOUR_PT_IP_HERE')

3. Also update in src/web_interface/app.py

4. Start the Orchestrator:

   cd ~/Networking_DT4SDN
   python3 src/controllers/orchestrator.py

5. Access the Dashboard at: http://DT_IP:5000

---

Step 3: Generate Traffic

In the Mininet terminal, use iperf to simulate different traffic patterns:

Normal Traffic:

h1 iperf -c 10.0.0.2 -u -b 500K -t 60

Burst Traffic:

h1 iperf -c 10.0.0.2 -u -b 25M -t 15 &

Congestion Simulation:

h1 iperf -c 10.0.0.2 -u -b 8M -t 90 &
# Wait 20 seconds, then:
h2 iperf -c 10.0.0.3 -u -b 8M -t 70 &

DDoS Simulation:

h1 iperf -c 10.0.0.2 -u -b 29M -t 400

Warning

Press Ctrl+C to stop traffic and observe the model returning to NORMAL state.

ML Models

---

The system uses two types of ML models:

Model Type	Purpose	Architecture
Classifier	Predict traffic state	BiLSTM
Seq2Seq	Predict 60s future values	CNN + BiLSTM

Available Scenarios:

• normal - Trained on normal traffic patterns
• burst - Trained on burst traffic
• congestion - Trained on network congestion
• ddos - Trained on DDoS attack patterns
• mixed - Trained on all scenarios (default)

Switching Models via UI: Use the dropdown in the dashboard header to switch between scenario models in real-time.

Switching Models via API:

curl -X POST http://DT_IP:5000/api/models/switch \
  -H "Content-Type: application/json" \
  -d '{"scenario": "ddos"}'

Project Structure

---

Networking_DT4SDN/
├── src/
│   ├── controllers/
│   │   └── orchestrator.py       # Main entry point
│   ├── web_interface/
│   │   └── app.py                # Flask dashboard + API
│   ├── database/
│   │   ├── db_manager.py         # SQLite operations
│   │   └── schema.sql            # Database schema
│   ├── ml_models/
│   │   ├── state_predictor.py    # Classifier model
│   │   ├── seq2seq_predictor.py  # Seq2Seq model
│   │   └── training/             # Training scripts
│   └── utils/
│       └── topology_generator.py # Mininet topology creator
├── models/                       # Trained model files (.pt)
├── setups/
│   ├── start_ryu.sh              # Ryu controller startup
│   └── setup_vm.sh               # VM setup script
├── Vagrantfile                   # Windows VM configuration
└── requirements.txt              # Python dependencies

API Reference

---

Endpoint	Method	Description
/	GET	Dashboard HTML page
/api/prediction	GET	Current traffic prediction
/api/models	GET	List available models
/api/models/switch	POST	Switch to different scenario model
/stats	GET	Database statistics

Links

---

• Presentation PDF

Team Members

---

• Matthew De Marco (matthew.demarco@studenti.unitn.it)

• Andrea Lo Iacono (andrea.loiacono@studenti.unitn.it)

• Jago Revrenna (jago.revrenna@studenti.unitn.it)

Acknowledgments

---

Networking Module 2 Course - Professor: Fabrizio Granelli

License

---

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

---

Back to Top

---