6G RAN's "Linux Kernel"

The following is the transcript of my talk delivered at OpenGovCon: "6G Innovation Day", hosted by the Linux Foundation ONE Summit on May 1st, 2024 and supported by NIST, NTIA, NSF, DHS and OUSD R&E:

5G is an amazing technology. It's authors took 4G, expanded it and made everything configurable. It's the ultimate software defined waveform. We can mould and shape it to support almost any wireless comms use-case. I'm not going to pretend I know what 6G RAN will look like but I expect that it will build on the flexibility of the 5G waveform.

The Linux Kernel is one of the most successful open source projects in the world. Under development since 1991, it now runs to 29 million lines of code and is deployed on billions of devices from smartphones to supercomputers. It has unlocked incredible innovation and driven increased accessibility, transparency and security in computer operating systems. As the word Kernel suggests, it's at the heart of the system - everything else builds upon and relies upon it.

The RAN stack is the kernel of 5G and FutureG. This is the CU and DU software that sits between the radio and the core. It sits at the heart of the system and is by far the most important software part of the network. It's also the most complex. The basic technical specification for the 5G RAN runs to over 2500 pages across more than 16 documents. And they're not exactly a light read! RAN stacks deployed in production networks are highly proprietary. They're usually tied to specific hardware and they're very difficult for innovators to get their hands on and build upon.

I firmly believe that in the same way the Linux Kernel transformed computer operating systems, an open-source RAN kernel can transform wireless communications systems. An open-source RAN kernel will unlock incredible innovation and drive increased transparency and security in mobile networks. It will address supply chain security for critical infrastructure and will do so in a way that reduces costs by removing huge inefficiencies that currently exist for integration, testing and validation. It will transform workforce development.

I firmly believe this because I've already seen many of these effects with the srsRAN Project.

At SRS, we've been building open-source RAN for a decade. By open-source, I mean publicly releasing software under an OSI approved open-source license. By RAN, I mean the CU and DU, complete from radio interface to core network interface. Right now, anyone can plug a radio into their computer, download our software and connect with the handset in their pocket.

We've publicly released 2 million lines of code across our 4G and 5G codebases and our aim has always been to provide the most performant yet most user-friendly, accessible open-source RAN code in the world. We have thousands of active users each month across our code repos, documentation, discussion forum and mailing list.

Commercially, our software has enabled some major innovations. With Smartsky networks we have built a coast-to-coast air-to-ground network for private jets and commercial aircraft. That network runs over 700 instances of our RAN software in production. With AST Spacemobile, we developed ground-breaking IP for direct-to-mobile satellite broadband and deployed it with Bluewalker 3, the largest ever deployed LEO communications satellite. It has been deployed on drones for search-and-rescue and in secure facilities for wireless threat analysis. Mobile wireless test equipment, passive cell sniffers, drone-based cell-tower testing, disaster zone emergency broadcasts, the list goes on.

This is real innovation, making it all the way from concept to products deployed in the real world. And this real-world deployment is critical. Paying customers demand performance and real deployments will test all the corner cases of your software. This means maturing your code, not just to support flash-in-the-pan demos or testbeds but to provide a solid platform for building real products and deploying real services.

So, from our decade of experience, what do I think are the necessary ingredients for this vision of a RAN "kernel"?

I think there are four - it must be:

• open
• complete
• performant
• deployed.

By open, it needs to be available under an OSI-approved open-source license. It also needs to be accessible - well documented, well-structured and with a clean consistent code style.

By complete, it needs to be self-contained with minimal third-party dependencies. It can't be tied to a specific brand or type of hardware and it must be scalable across architectures small and large.

By perfomant, I mean it must be stable, efficient and compliant with specifications.

Finally, by deployed, I mean it must have dependent consumers, running it in production and demanding performance. Dependent consumers will ultimately invest in and sustain the project.

At SRS, we have built the srsRAN Project specifically with this vision of the RAN Kernel in mind. At MWC earlier this year, we showcased our CU+DU running on the latest processors from Intel, AMD and ARM and being exercised by the world's leading network test equipment in end-to-end multi-cell, multi-UE configurations.

We've invested millions over the past four years and we've built a rock-solid foundation in srsRAN. To fully realize the RAN Kernel vision, we look forward to working with partners across government, industry and academia who are ready to invest in that vision. The time for that investment is now.