Recently it seems like you can’t move for outrageous superlatives about the impact of 5G. I could quote a number, but I think this one from Steve Mollenkopf, CEO of Qualcomm, sums it up:
“5G will have an impact similar to the introduction of electricity or the car, affecting entire economies and benefiting entire societies”.
Really? Than electricity? Electricity without which we couldn’t have 5G networks, or indeed just about any facet of modern life? Of course, I don’t believe that. You don’t believe that. I don’t even think Steve Mollenkopf believes it. It is just the Defcon 1 of the 5G claims-race, which has seen execs trying to outdo each other in claims of how much of an impact 5G will have. We’ve heard about it being a key trigger for the 4th Industrial Revolution, as the catalyst for IoT adoption and so forth. An amazing amount of hyperbole. There have also been some *ahem* ambitious predictions for the dollar value of the economic impact of 5G.
All this talking up of 5G seems a little ridiculous compared to the silence from would-be technology adopters. In discussions with enterprises the subject rarely if ever comes up. A little during discussions about industrial use cases, but that’s about it. Even on the vendor side, it’s only those organisations directly affected by 5G that seem to have much faith in it, and specifically the impact it will have on IoT. About 3 years ago I said that 5G talks about IoT a lot. IoT talks about 5G not at jot. That has changed slightly, but only slightly.
But I don’t want this to come across as another 5G-sceptic article. It’s not. Actually, 5G represents a paradigm shift in the telecoms world which could, and should, shake up all sorts of elements in how ICT services are delivered.
Usually the benefits of 5G are discussed in terms of three areas: higher bandwidth, the ability to support more devices and lower latency. The additional bandwidth and ability to support additional devices naturally provides a superior experience. That alone will guarantee demand. There is perhaps not immediate huge pent up demand for bandwidth, but with HD video streaming and real-time video-gaming realistically beyond the capabilities of 4G there are obvious services for 5G to support. Put bluntly, if 5G provides higher bandwidth there is bound to be demand. And, furthermore, if it provides additional capacity the mobile network operators will naturally want to use it.
As for supporting massive machine-type deployments, the existing networks aren’t anywhere near bumping up against capacity yet, although that may come. All of this comes with the caveat that the higher frequencies that will typically be used for 5G mean worse in-building coverage. But hey, maybe you’ll get a repeater. Things are complicated a bit by the fact that NB-IoT will be dragged kicking and screaming into 5G in the near future.
It’s the third leg of the stool that represents the real paradigm shift: lower latency. Historically the radio access network represented the pinch point for any application. It didn’t matter how fast the core network was, cellular networks never had any better latency than about 50ms. 5G promises latency as low as 1ms, although for high bandwidth applications it is more like 10ms. This changes the dynamic when put up against fibre, which has a theoretical latency of 5ms on a 1,000km round trip, and realistically 3-5x that figure, so something around 20ms .
The interested thing is that it then becomes the core rather than the edge of the network which is the limiting factor. If a RAN adds 50ms delay it doesn’t matter much if the core adds a 20ms delay while it indulges in a 1,000km round trip to a central server (notwithstanding processing and routing delays). A centralised architecture is fine. In contrast, if the RAN latency is 10ms the greater delaying factor on applications may be the core network. The net result in the example I quoted is a reduction from 70ms to 30ms. Great. But in some applications having a delay of two-to-three times will make a significant difference.
The natural response must then surely be to put more processing and storage at the edge in a much more distributed model. One model sees 5G base-stations doubling up as mini data centres. Another takes the compute power for IoT applications from edge devices and puts it into the base-station. Why have smart edge devices when the network latency back to a more cost-effective semi-centralised compute function is only 10ms? Take video image processing, for instance. With 4G you couldn’t realistically stream all data, so some processing on the edge device was inevitable, and then sending outliers to a central processing function made sense. With 5G that is all turned on its head. Neither the edge device nor the central server make sense as the correct place to do processing.
The big limitation of this transformation is practical. When do we expect that 5G coverage will be near enough to universal to take advantage. Other than in controlled environments such as campuses and factories, almost never, sadly. The aforementioned challenges of getting sufficient geographical coverage mean that there won’t be 5G-only devices for a long time to come. If devices need to have a fall-back then that needs to be supported by edge processing and/or a centralised architecture. Once the investment has been made in that capability, there is little to be gained from introducing a third cost into the equation in the form of base-station compute. Plus it means managing two solution architectures (5G and non-5G) with all the headaches that entails. For narrowly bound use-cases focused on urban areas or campuses that are more easily controlled, it is potentially very useful. Or if carriers were prepared to commit to having 100% 5G New Radio (NR) it might make some sense. The latter is particularly interesting because it puts the spend by the client into an area controlled by the carrier, unlike the device or the cloud (in most cases). A cost benefit analysis will be needed, clearly, but I have my doubts that the revenue opportunity outweighs the additional cost of network deployment.
5G in theory offers a very interesting shift in dynamics in cloud vs edge, and an opportunity for carriers to generate additional revenue. However the hybrid nature of most carriers’ networks, and end users’ need for ubiquity and homogeneity of service, will make it hard to realise.