P4TG: Traffic Generation for Ethernet/IP Networks

• 📑 Table of Contents
• 📝 Overview
  • Traffic Generation
  • Statistics
  • Version Highlights
  • Features
• 🚀 Installation & Quick Start
  • Quick Start
  • Manual Installation
    • Data Plane
    • Control Plane
  • Configuration
    • Configuration Options
    • 64-port Tofino
• 🤖 Test Automation
• 🔄 Update Guide
• 📚 Documentation
• 🛠️ Development
• 🖼️ Preview
• 📖 Cite

P4TG (P4-based Traffic Generator) is a programmable traffic generator built on the Intel Tofino™ ASIC.
It enables reproducible experiments with Ethernet/IP traffic at high data rates while providing precise in-data-plane measurements.
P4TG combines a P4 data plane program, a Rust-based control plane, and a React-based configuration GUI.

📖 Related publications:

• P4TG: 1 Tb/s Traffic Generation for Ethernet/IP Networks (v1.0.0)
• Enhancements to P4TG: Protocols, Performance, and Automation (v2.3.2)
• Enhancements to P4TG: Histogram-Based RTT Monitoring in the Data Plane (v2.4.0)

• Tofino 1: Up to 1 Tb/s across 10 × 100 Gb/s ports, or 40 × 10/25 Gb/s
• Tofino 2: Up to 4 Tb/s across 10 × 400 Gb/s ports, or 40 × 10/25 Gb/s

Generated packet headers can be fully customized, including:

• Ethernet
• IPv4 (full address randomization)
• IPv6 (address randomization of 48 LSB)
• MPLS (up to 15 LSEs)
• VLAN
• QinQ
• VxLAN
• SRv6 (up to 3 SIDs, Tofino2 only)

Traffic may be looped back (directly or through other equipment) to measure:

• TX/RX rates
• Packet loss
• Packet reordering
• Inter-arrival times (IATs)
• Round-trip times (RTTs)
• Frame types
• Frame sizes

directly in the data plane to ensure accurate results.

See the full Changelog.

P4TG requires a fully set up SDE with $SDE and $SDE_INSTALL environment variables set. A detailed installation guide for open-p4studio and P4TG can be found here.

The provided p4tg.sh script automates the installation of the data and control plane on Debian- / Ubuntu-based systems, provided that the SDE is installed correctly. Make sure that the environment variables $SDE and $SDE_INSTALL are set. Run with sudo -E to pass environment variables.

Usage: ./p4tg.sh [install|update|start|stop|restart|status]

Clone P4TG into /opt/P4TG and simply run sudo -E ./p4tg.sh install (tested on Debian-based systems). Change the paths at the top of p4tg.sh if needed.

The install command will:

• Compile the data plane and copy it in place.
• Pull the docker image and start it.
• Copy the p4tg.sh script to /usr/local/bin.
• Copy the p4tg.service file to /etc/systemd/system/p4tg.service. This service file can be used to autostart P4TG on boot.

The start command will:

• Load all required kernel modules to operate the Tofino.
• Start the data plane and wait for it to become ready.
• Start the control plane docker image.

cd P4-Implementation

• Tofino 1:

  make compile TARGET=tofino
  make start TARGET=tofino

• Tofino 2:

  make compile TARGET=tofino2
  make start TARGET=tofino2

Tested on:

• SDE 9.9.0 (up to v2.0.0)
• SDE 9.13.{0,...,4}

cd Controller
docker compose up

• Starts a REST-API server on port P4TG_PORT (default: 8000) at /api
• Serves the React GUI at /
• Access GUI: http://<tofino-controller-ip>:P4TG_PORT
• Access API: http://<tofino-controller-ip>:P4TG_PORT/api

Docker compose file Controller/docker-compose.yaml

• SAMPLE=1 → IAT sampling mode (default: 0, data plane measurement)
• LOOPBACK=true → enable loopback testing mode
• P4TG_PORT=8000 → changes the controller port
• NUM_PORTS=32 → set number of front panel ports of your device

Config file: Controller/config.json

• Specify ports for traffic generation. Per default, front panel ports 1 - 10 are configured automatically.
• Configure MAC address for ARP replies, breakout mode, and port settings. Port settings can further be changed during runtime through the API or the GUI.

Example:

{
  "tg_ports": [
    {
      "port": 1,
      "mac": "00:d0:67:a2:a9:42",
      "speed": "BF_SPEED_100G"
    },    
    {
      "port": 2,
      "mac": "fa:a6:68:e0:3d:70",
      "breakout_mode": true,
      "speed": "BF_SPEED_100G"
    },
    {
      "port": 3,
      "mac": "00:d0:67:a2:a9:42",
      "speed": "BF_SPEED_100G",
      "fec": "BF_FEC_TYP_NONE",
      "auto_negotiation": "PM_AN_FORCE_DISABLE"
    },
  ]
}

Each front panel port requires two recirculation ports. They are configured automatically. However, on the 64-port Tofino devices, you may have to manually set internal TX/RX recirculation ports to match your device. Below is an example:

{
  "tg_ports": [
    {
      "port": 49,
      "mac": "fa:a6:68:e0:3d:70",
      "speed": "BF_SPEED_100G",
      "recirculation_ports": {
        "tx_port": 50,
        "rx_port": 51
      }
    }
  ]
}

This uses front panel port 49 as a port for traffic generation and port 50 and 51 for internal recirculation. Ensure that the recirculation ports support the same line rate as the TG port.

Automation scripts in p4tg_test_automation:

• Python script for starting/stopping tests, fetching stats, and rendering plots
• Quick start:

  python run.py --payload payloads/your_test.json

• Example payloads included

See README for details.

1. Rebuild the data plane as described above.
2. Update controller:

   docker compose pull
   docker compose up -d

3. If GUI issues occur, clear browser storage.

• REST-API docs: P4TG API Docs
• Or via controller: /api/docs

For development instructions, please see here

If you use P4TG in any of your publications, please cite the following papers:

1. S. Lindner, Marco Häberle, and M. Menth: P4TG: 1 Tb/s Traffic Generation for Ethernet/IP Networks, in IEEE Access, vol. 11, p. 17525 – 17535, Feb. 2023, IEEE
2. F. Ihle, E. Zink, S. Lindner, and M. Menth: Enhancements to P4TG: Protocols, Performance, and Automation, in KuVS Workshop on Network Softwarization (KuVS NetSoft), online, Apr. 2025
3. F. Ihle, E. Zink, M. Menth: Enhancements to P4TG: Histogram-Based RTT Monitoring in the Data Plane, in Workshop on Resilient Networks and Systems (ReNeSys), Jul. 2025, Ilmenau, Germany

@article{LiHae23,
  title   = {{P4TG: 1 Tb/s Traffic Generation for Ethernet/IP Networks}},
  author  = {Steffen Lindner and Marco Häberle and Michael Menth},
  journal = {{IEEE Access}},
  year    = 2023,
  month   = feb,
  volume  = 11,
  pages   = {17525--17535}
}

@article{IhZi25,
  title  = {{Enhancements to P4TG: Protocols, Performance, and Automation}},
  author = {Fabian Ihle and Etienne Zink and Steffen Lindner and Michael Menth},
  journal = {{KuVS Workshop on Network Softwarization (KuVS NetSoft)}},
  year   = 2025,
  month  = apr
}

@article{IhZi25_2,
  title  = {{Enhancements to P4TG: Histogram-Based RTT Monitoring in the Data Plane}},
  author = {Fabian Ihle and Etienne Zink and Michael Menth},
  journal = {{Workshop on Resilient Networks and Systems (ReNeSys)}},
  year   = 2025,
  month  = sep
}