To Infinity and Beyond
We now have a complete draft of our Private 5G book. It’s not ready to publish just yet, but we do invite you to take a look on GitHub and give us your feedback. This week’s post considers one of the questions that kept nagging us as we worked on the project: How might the new technology introduced by 5G disrupt the underlying mobile cellular industry? And what comes after 5G?
In late 2022 I attended a Beyond 5G International Conference hosted by Japan’s Ministry of Communications, where considerable energy was put into international alignment over how standards will evolve (or so it seemed to a technical person like myself). This got me thinking about the interplay between standards and technology—and the more I thought about it, also marketing—all of which are threads I seem to keep pulling on as we bring our new book on Private 5G across the finish line.
From the moment MNOs started rolling out 5G base stations in 2019, people started talking about what comes next. The obvious answer is 6G, but it's not at all clear that the decadal generations of the past 40 years will continue into the future, especially since the “Gs” are as much a marketing construct as anything else. Today you often hear alternatives like “NextG” and “Beyond 5G” more often than 6G, which could be read as a sign that the rules of the game are changing. From my technical vantage point, I would certainly argue that we’re in the midst of a sea change that will render the generational distinction largely meaningless (except, perhaps, as click bait).
First a little history. One way to talk about mobile cellular generations is with respect to the radio technology: 2G used Time Division Multiple Access (TDMA), 3G used Code Division Multiple Access (CDMA), 4G introduced Orthogonal Frequency-Division Multiple Access (OFDMA), and 5G is now adding more degrees-of-freedom to OFDMA in support of a broader set of service classes. A second way to talk about generations is in terms of service offerings: 2G supported voice and text, 3G introduced support for broadband (but was still fundamentally circuit-based), 4G made the broadband “IP-native”, and 5G is now trying to add additional services (beyond broadband) by becoming “Cloud-native”.
I’m out of my depth when it comes to what’s next in coding and modulation, but it’s hard for me to imagine how it’s possible to top the stated goals of 5G’s New Radio (NR), which include ultra-high density (1 million nodes per square kilometer); ultra-high availability (greater than 99.999%); ultra-low latency (as low as 1 ms); extreme mobility (up to 100 km/h); extreme data rates (multi-Gbps peak); and extreme capacity (10 Tbps of aggregate throughput per square kilometer). Maybe transformational technology will come along, and if it does, it’s quite reasonable for it to claim the generational mantle, but someone’s also going to need to come up with marketing terms that surpass “ultra” and “extreme”.
What I find more interesting is the services that will be supported, and on this front, I see two complementary reasons for 6G becoming a hollow term. Both are at the heart of what’s important about Private 5G. The first factor is that by adopting cloud technologies—including cloud native microservices, Software-Defined Networking, and cloud management practices—the mobile cellular network is hoping to cash in on the promise of feature velocity. This "agility" story was always included in the early 5G promotional material, as part of the case for why a 5G upgrade would be a worthwhile investment, but the consequence of those technologies now finding their way into the mainstream is that new features can be introduced rapidly and deployed continuously. At some point, the frequency of continual improvements render generational distinctions irrelevant.
The second factor is that agility isn’t only about cadence; it’s also about customization. That is, these changes can be introduced bottom-up—for example by enterprises and their edge cloud partners in the case of Private 5G—without necessarily depending on (or waiting for) a global standardization effort. If an enterprise finds a new use case that requires a specialized Mobile Core, only its Private 5G deployment needs to adopt the necessary changes. Similarly, if there is an xApp that can be customized for a particular use, this can potentially happen in a purely local way (assuming the RAN elements support the necessary control interfaces and performance metrics). Reaching agreement with all the incumbent stakeholders will no longer be a requirement.
It's anyone's guess where this will take us, but it will be interesting to see how this dynamic impacts the role of standardization: what aspects of the mobile network require global agreement and what aspects do not (because they can evolve on a case-by-case basis). For example, while it is likely important that the over-the-air interface remains universal so devices can connect anywhere (and surely this interface will evolve over time), the internal interfaces of the control plane—or even how the control plane is factored into individual components for that matter—need not adhere to the 3GPP specification (as Magma has already demonstrated). This would seem to be especially true for IoT devices that (a) never make or accept voice calls, and (b) never move off-site. Much of the value that the 3GPP-defined Mobile Core provides is in supporting universal connectivity for phone calls (see our discussion of this point in a recent post), which could very well become a niche application, and even then, supported as some variant of Voice-over-IP.
As for how the RAN evolves, at this point it’s not clear that there will be universal agreement in any meaningful way, even as the O-RAN Alliance works towards standardizing a control interface in the name of eliminating vendor lock-in. This is because the incumbent vendors have already carved out device-specific exceptions, reminiscent of how OpenFlow became overly coupled to the underlying switching hardware being controlled. Adopting a P4-like strategy for the interface to RAN elements would help address this issue, but whether that actually happens is far from assured. Should this actually happen, the RAN will be put on the same footing as the Core: It will be entirely software-defined, with all comers free to innovate as they please. And once again, we can expect the diversity of application domains that Private 5G enables to provide the market pull for those innovations.
All of which is to say, while standards often spur innovation (TCP and HTTP are two great examples from the Internet experience), sometimes standards primarily serve as a barrier to competition, and hence, innovation. Now that software is finally eating the mobile cellular network (with Private 5G deployed in enterprises likely setting the pace), we will quickly learn which standards are which. Meanwhile, marketing teams can return to their wheelhouse—differentiation—instead of looking for creative ways to promote lock-step generational upgrades.
We have a long-standing interest in AI (Bruce narrowly avoided becoming an AI student just before the AI winter of the 1980s) so we have been watching with interest the impact that AI might have on software development. We particularly enjoyed this post from Forrest Brazeal (also known for his musical videos) which responded rather thoughtfully to “The End of Programming” and the debate it generated. In related news, Nick Cave did not hold back when someone asked him for an opinion about a song written “in the style of Nick Cave” by ChatGPT. We haven’t been able to get ChatGPT to write newsletters to our satisfaction either, so we’re sticking to artisanal prose for now. And speaking of artisanal prose, our ebooks are on sale for a few more weeks.