Generating P4 Code for the Application Layer
Throughput and latency-critical applications (e.g. processing sensor data, robot control, or monitoring stock market streams) can often benefit if computations are performed close to the client. Performing these computations in the data plane can help us take it to the next level.
P4 is excellent data plane programming language, and we all love it. However, it wasn't meant to implement application-layer tasks. Thus, offloading server functionality can be challenging.
P4RROT is a code generator that helps programmers overcome certain limitations and write shorter and easier-to-read code.
Code generation can greatly simplify implementing application-layer tasks if we narrow down the scope of features.
Based on a Python script and a P4 template, P4RROT generates P4 code that can be compiled to the desired target.
P4RROT is a very young project. The current code base supports BMv2,Netronome NFP and Tofino.
A more detailed description of P4RROT can be found in our preprint paper.
A hands-on Hello World tutorial is also available at examples/hello-world.
Any kind of contribution is appreciated ranging from tutorial writing to bug fixes and implementing new features. In short, we use GitHub to host code, track issues and feature requests, as well as accept pull requests.
Read our Contributing Guidelines for more details.
The library is released under MIT license but please note that some of the tests, examples and templates might have different licensing.