This repository contains the source code for a P4 based traffic generator based on the Intel Tofino(TM) ASIC, called P4TG (Paper). The paper version corresponds to v.1.0.0.
In generation mode, P4TG is capable of generating traffic up to 1 Tb/s split across 10x 100 Gb/s ports. With Intel Tofino(TM) 2, P4TG is capable of generating traffic up to 4 Tb/s split across 10x 400 Gb/s ports. Thereby it measures rates directly in the data plane. Generated traffic may be fed back from the output to the input ports, possibly through other equipment, to record packet loss, packet reordering, IATs and sampled RTTs. In analysis mode, P4TG measures rates on the input ports, measures IATs, and forwards traffic through its output ports.
- P4TG (v2.0.0) supports VLAN (802.1Q) and QinQ (802.1ad) encapsulation
- P4TG (v2.1.0) further supports MPLS
- P4TG (v2.2.0) supports VxLAN
- P4TG (v2.3.0) supports the Intel Tofino 2 with up to 4 Tb/s
- P4TG (v2.3.2) supports IPv6, SRv6 (Tofino 2 only), many bug fixes, and a dark mode
See Changelog for a full changelog.
P4TG consists of:
- a P4 program for the Intel Tofino(TM)
- Rust control plane
- React configuration GUI
| Feature | Status |
|---|---|
| Statistics | ✅ Available (Implemented) |
| Tofino 1 support | ✅ Available (Implemented) |
| Tofino 2 support | ✅ Available (Implemented) |
| Web frontend | ✅ Available (Implemented) |
| Rust backend | ✅ Available (Implemented) |
| Ethernet | ✅ Available (Implemented) |
| IPv4 | ✅ Available (Implemented) |
| IPv6 | ✅ Available (Implemented) |
| QinQ | ✅ Available (Implemented) |
| MPLS | ✅ Available (Implemented) |
| VLAN | ✅ Available (Implemented) |
| VxLAN | ✅ Available (Implemented) |
| SRv6 | ✅ Available (Implemented) |
| ARP replies | ✅ Available (Implemented) |
| Dark mode | ✅ Available (Implemented) |
| Test profiles | |
| File reporting | |
| Auto. testing | |
| Localization | |
| NDP | ⏳ Planned (Not yet implemented) |
| NETCONF | ⏳ Planned (Not yet implemented) |
| RTT percentile reporting | ⏳ Planned (Not yet implemented) |
Go to P4-Implementation and compile p4tg via make compile.
This compiles the program and copies the resulting configs to the target directory.
Afterwards, start p4tg via make start.
- For Intel Tofino1, run
make compile TARGET=tofinoandmake start TARGET=tofino. - For Intel Tofino2, run
make compile TARGET=tofino2andmake start TARGET=tofino2.
This requires a fully setup SDE with set $SDE and $SDE_INSTALL environment variables.
Tested on:
- SDE 9.9.0 (up to v2.0.0)
- SDE 9.13.0
- SDE 9.13.1
- SDE 9.13.2
- SDE 9.13.3
- SDE 9.13.4
The controller is written in Rust and can be started via cd Controller && docker compose up. This will pull a prebuilt docker image.
The controller then starts a REST-API server at port P4TG_PORT and endpoint /api (see docker-compose.yaml) that is used to communicate with the configuration GUI.
It also serves the configuration GUI at port P4TG_PORT and endpoint /.
The configuration GUI can then be accessed at http://ip-of-tofino-controller:P4TG_PORT
Set SAMPLE=1 in docker-compose.yaml to activate IAT sampling mode instead of data plane measurement.
Data plane measurement mode (SAMPLE=0) is more accurate and the default.
Set LOOPBACK=true in docker-compose.yaml to configure all ports in loopback mode. This is only useful for testing.
The configuration file in Controller/config.json can be used to specify the ports that are used for traffic generation.
If no config file is provided, the first 10 ports are used.
Further, the MAC address can be specified that should be used to answer ARP requests when the ARP reply option is enabled in the UI.
Connect to the REST-API server through the frontend of the configuration GUI: http://ip-of-tofino-controller:P4TG_PORT/api
The configuration GUI is automatically served by the controller at http://ip-of-tofino-controller:P4TG_PORT and is included in the prebuilt docker image.
If you want to adapt the configuration GUI, or the control plane:
- Start the controller with
RUST_LOG=info cargo runfrom inside theControllerfolder. The controller communicates via the gRPC port (default: 50052) with the data plane. - Re-build the configuration GUI via
npm install && npm startwithin theConfiguration GUIfolder. - Adapt the
API_URLin theconfig.tsto run the controller and configuration GUI independently.
Afterward, re-build the controller via docker compose build within the Controller folder.
This will automatically build and copy the configuration GUI into the controller folder and create an image containing the modifications.
If you update to a newer version, it might be necessary to delete your local storage in your browser for the configuration GUI.
The documentation of the REST-API can be found here or at the /api/docs endpoint in the REST-API of the controller.
If you use P4TG in any of your publications, please cite the following papers:
- 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
- F. Ihle, E. Zink, S. Lindner, and M. Menth: Enhancements to P4TG: Protocols, Performance, and Automation, in KuVS Fachgespräch "Network Softwarization", online, Apr. 2025
@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 Fachgespräch "Network Softwarization"}}
year = 2025,
month = apr
}