Democratizing 5G
Bringing technology to a wider audience
Over the last several years, we've been involved in multiple efforts to bring 5G to a wider audience, through both open source efforts (Aether, Magma, etc.) and our books. These efforts have served to remind us how much the mobile wireless architecture, in sharp contrast to the Internet, was designed with the needs of telco operators in mind. That said, 5G bases much of its architecture on standard cloud technologies, and it is now feasible for enterprise users–and the occasional book author–to deploy 5G. This places 5G in a long line of efforts to democratize access to technology as we discuss below.
I recently attended SIGCOMM for the first time in many years, and was immediately reminded of the standard salutation when you run into someone at an academic conference: “What are you working on?” I wasn’t prepared with a succinct-yet-meaningful response, but quickly settled on “Democratizing 5G”. It wasn’t the expected research-focused answer, but it did have the quaint advantage of not directly involving ML. It seemed to do the trick, but also got me thinking about how common the word “democratizing” has become in our technical jargon.
I believe the first time I heard the word being used in a technical setting was in 2004, when Mike Freedman and colleagues published Democratizing Content Publication with Coral at the first NSDI. And if the meaning wasn’t already intuitive, the paper gave a definition that nicely captures the spirit of what it means to democratize something technology-related:
CoralCDN replicates content in proportion to the content’s popularity, regardless of the publisher’s resources, in effect democratizing publication.
The Internet (like the printing press centuries before) is widely regarded as playing a central role in the democratization of knowledge—making information readily available among the wider population. CoralCDN focused specifically on the publishing side of the equation, making it easier for the wider population to also disseminate information. That’s now taken for granted with the explosion of various platforms (e.g., blogs, social media) that include built-in content distribution mechanisms, but was novel at the time. It also highlights the idea that there are multiple barriers to lower; in this case, both acquiring and disseminating content. Less obviously, it points to the role “access to resources” plays in the equation. CoralCDN was only able to democratize publication because it ran on a publicly-funded infrastructure, in this case PlanetLab.
My use of the word democratization in the context of 5G focuses more on the underlying network infrastructure—with the intended audience being people who build and operate that infrastructure—than on end-users that benefit from that infrastructure. I tend to think of it as lowering the barrier to innovation, but it also shares much with the idea of Freedom to Tinker, where know-how is essential to crafting good policy. Either way, it’s about broadening the set of people able to participate in technologies that impact our lives.
For 5G, democratization turns out to be a multi-faceted challenge. A necessary condition is access to open source implementations of both the RAN and the Mobile Core, which, thanks to various open source organizations (ONF’s Aether and OAI for example), now exist. But a quick perusal of the Git repos for those and similar projects will immediately convince you that the mere existence of open source software is not sufficient; users also need the wherewithal to deploy and operate the code if they have any ambition to take advantage of it. My experience is that “wherewithal” maps onto a combination of tooling and documentation, with the latter being the “long pole”. The result—our Private 5G book, which includes a tutorial guiding the reader through the OnRamp deployment toolset—is now available.
I’ve written about that topic before, so won’t rehash the challenges here, except to make two observations. The first is that in our haste to implement new functionality, we have created a nearly impenetrable mountain of configuration variables needed to manage that functionality; YAML files layered on top of Jinja2 templates overlaid on still other YAML files layered on top of JSON… I’m sure it’s not the intention to obfuscate the underlying code, but that is the practical effect. It’s almost as though we’ve created a problem only AI will be able to help us solve. The other option is to be a well-resourced company with a team of experienced engineers, but of course that flies in the face of our goal, which is to democratize access to that know-how.
The second observation, which follows from my experience trying to bring up a 5G small cell, is that the mobile cellular technology has operational complexity baked into its design. This makes sense (in a perverse way) when you consider that the technology was defined by MNOs that built businesses around their ability to operate the network on behalf of their subscribers. But this creates additional barriers that need to be lowered if you want to broaden participation. For example, programming a SIM is a required step to establishing a secure 5G connection, but being able to do that in turn requires having the necessary credentials (plus the know-how to correctly specify another few hundred lines of YAML). Fortunately, the broader ecosystem includes players that help on that front, but the main takeaway is that there is much more to the democratization of technology than initially meets the eye.
Why are we doing this?
But that all raises the question—typically following immediately after “What are you working on?”—which is: “Why should I care?” It’s a good question. If you’re going to put the effort into lowering the barrier-to-entry, there ought to be something important on the other side of those barriers. Again, there are a couple of parts to the answer.
The first part is to simply acknowledge that Internet access is going to be dominated by mobile wireless connectivity. Nearly 60% of all web traffic already comes from mobile devices, and that doesn’t yet take into account the tens of billions of IoT devices that are expected to connect to the Internet over the next few years. We’ve spent 40 years building the Internet out of open and accessible technologies, but going forward, democratizing Internet infrastructure is meaningless without also democratizing wireless access. That the mobile cellular industry has been so closed and proprietary for so long makes that goal all the more relevant.
The second part is to zero in on 5G vs other wireless technologies, most obviously Wi-Fi. At the coding and modulation level, Wi-Fi 6 and 5G’s New Radio (NR) are converging on OFDMA. The difference is how the available spectrum is allocated by the two systems, which Bruce and his Magma collaborators discuss in depth in their NSDI paper on Magma. Digging deeper, 5G scheduling includes the ability to dynamically change the size and number of schedulable resource units, including scheduling intervals as short 0.125ms. This opens the door to making fine-grained scheduling decisions that are critical to predictable, low-latency communication. The 5G scheduler also allocates some of the available spectrum to a light-weight over-the-air-interface that is simple enough for IoT devices to implement. These devices are not particularly latency-sensitive or bandwidth-hungry, but they often require long battery lifetimes, and hence, reduced hardware complexity that draws less power.
There’s reason to be skeptical, but as Bruce discusses in a previous post, the application of cloud best-practices to 5G is a game changer. The hype around 5G has certainly gotten out in front of the reality, but it’s only just now the case that 5G is becoming viable in the enterprise (aka Private 5G). For example, the Aether project has only recently been able to certify a commercially available small cell radio, with ubiquitous 5G devices (other than smartphones) still to follow. Until these components become ubiquitous, many of these advantages outlined above will not be realized.
Bringing this discussion back to the question of what democratization means in the technical world, a personal takeaway is that I now see a direct line between “Democratizing X” and “X: A Systems Approach”. The Systems Approach has always looked at technology through the lens of deployed systems, using real implementations to explain the design decisions. Democratizing access to technologies is an almost inevitable consequence of how we help readers to understand them—by first deconstructing them into their elemental components and then showing how all the pieces are assembled into a coherent whole.
Aether OnRamp is now available, providing what aims to be a relatively painless way to stand up a private 5G deployment. Watch the Techniar Video and learn about the commercial 5G small cell radio that can be deployed with Aether. And for a different approach to democratizing 5G, see the Helium Network.