Working on the future of networks: P4 training for GÉANT Community*

One of the hot new areas of interest in the programmable networking domain is the potential for new services and new ways of delivering networks using the P4 language (Programming Protocol-independent Packet Processors). P4 allows dynamic programming of a network forwarding plane in network devices such as switches and routers and is designed to be implementation-independent, meaning P4 programs can in principle be compiled on a wide range of platforms.  As an enabler, P4 can be coupled with white box networking devices to potentially offer new services such as network telemetry and DDoS detection and mitigation running on open, flexible platforms.

In GN4-3, Network Technologies and Services Development work package (WP6) has a task dedicated to evaluating new network technologies (Task 1), to see whether such technologies can add value in use cases of interest to the GÉANT and NREN community. Within that task, there are two subtasks where P4 is of specific interest: Router for Academia, Research and Education (RARE), which aims to provide a fully open-source software (control and data plane) for research and education community, and Data Plane Programming (DPP), which is exploring advanced network features that can be created thanks to P4, such as the new telemetry and new DDoS detection and mitigation capabilities mentioned above.

To ensure and develop a P4 capability in-GÉANT-house WP6 teamed up with GLAD (WP1 T5) and the partnership has proven to be very valuable. The training has allowed the WP6 team of engineers to master their developments based on P4 and therefore achieve their objectives much more quickly than by self-training. It has also helped the RARE project go beyond the initial project objectives. The engineer testimonies below highlight the particular value of the GLAD training:

Frédéric Loui (RARE team leader) (RENATER) – “With my RARE technical leader cap on, starting this project from a clean slate was a real feat. The project is not only ambitious technically, but it also forces us to rethink how networking is being implemented from decades. The GLAD team played a major role in enabling this training for the whole team in such a short amount of time. For anyone who wants to learn p4, this training must be followed as it is a real game-changer. It enabled each participant: 

  • to acquire technical knowledge
  • to extend RARE team networking reach. We met people working on various projects having significant different objectives as RARE’s
  • most importantly it enabled us to achieve the tremendous work done during Year 1. Indeed without this training, we would have never been able to reach such a result in only one year 

Last but not least, we are now able to share this knowledge with any organisation in the R&E community. This knowledge transfer can take various forms:

  • By preparing self-paced training materials:
  • By preparing a generic p4 workshop with the help of GLAD
  • By preparing a specific tailored workshop addressing R&E community member use cases that demonstrate a tangible interest in P4 …”

Jordi Ortiz (REDIris) – “The P4 course provided me with a deeper knowledge on how real boxes work and now that they are programmable how to properly do it. The support from the GLAD team was the icing on the cake to get involved in a technology that is probably going to be disruptive in network provisioning and academia.

Damu Ding (GARR) – “This course helped me to understand from scratch how P4_16 can be used to compile code that can run in Tofino switches. The 4-days class covered broad aspects of topics and the material was well prepared. During the course, there were also some useful hands-on labs that helped me better understand what I was learning. Based on the knowledge obtained from the course, I am now able to quickly set up and configure the P4-programmable switches in our physical testbed. In addition to this, I have successfully implemented the P4 code for our DDoS detection use case, which is now fully executable in the Tofino switches. This would have been very difficult for me without attending the course.

Maxime Wisslé (RENATER) – “The course was very interesting, especially on the low level of the hardware. It provides the knowledge on the running of each piece of hardware and how P4 work with all of this component to perform a powerful forwarding.

Xavier Jeannin (RENATER):As WP6 T1 task leader, it was necessary that I would be able to understand this technology to coordinate the very innovative work of these two teams. The training matches this goal perfectly.” 

Alexander GALL (SWITCH) – “The course provided two essential insights for me. One is a deeper understanding of the P4 language itself by analyzing programs that cover various techniques for packet processing relevant to real-world applications. The other is a coverage of the architecture-specific features required to make these programs work on a specific ASIC. Without this knowledge, it is very challenging to write non-trivial programs on actual hardware. Both aspects were covered in a manner that allows me to apply them directly to the problems we have set out to solve in the RARE project.

On behalf of GLAD, we can only say ‘wow’, we are so pleased be a part of such a rich learning expedition.

*Writted by Xavier Jeannin (RENATER) with a few additions from GLAD.