Mobile networks will transform from now through 2020, more than since the inception of 2G. New 4G capabilities will trigger some of that, however, 5G both enables and encourages more fundamental change.
In 4G, the support of higher orders of modulation (256 QAM), additional carrier aggregation combinations (4CC and upward), and more transmission layers will enable gigabit services. You have likely read about them coming already. In some cases, those rates will be based solely on licensed spectrum; in others, through combinations with unlicensed (WiFi) or even shared spectrum. While not pervasive in coverage, gigabit rates will give users an experience never imagined on wireless. Peak rates and capacity don't come for free, though. In addition to antenna, radio and basestation changes, focused transport investments will also be required to carry the increased traffic loads.
Happening almost coincidentally, is a move towards access centralization and virtualization on the premise of simplification (and thus opex reduction), and performance improvements.
In centralization, collections of baseband processing units, typically located near antennas, are moved 15 to 24km away into environmentally controlled rooms or buildings, sometimes called hotels. Visualize racks of baseband hardware with dark fiber connections to the radio sites where they were previously located.
Virtualization, which builds on centralization or can be implemented independently, moves some of the software previously implemented on proprietary hardware or bare metal onto off-the-shelf IT hardware, with cloud-based VNF software structures. The visualization is the same as hotels, except now the racks contain IT hardware. Let's call this collection of IT hardware and software an edge cloud. To implement an edge cloud, basestation protocol stacks need to be split with some layers remaining near or embedded in the antenna and some in the edge cloud. Distances between them vary depending on where the split is done.
Now for 5G. While changes in 4G are optional, though likely, 5G encourages and almost demands change. Its primary sales pitch is the introduction of very high speed and low latency services. The latter are enabled but not enforced by 5G, however, as they create new sales opportunities, we expect many operators to take advantage of that capability.
Core network changes are dominated by moving towards a Service Based Architecture (SBA) and different connectivity options to the access network. In fact, there are ten possible options for a BTS to connect to the core and eight options for how access protocol layers will be split (to enable virtualization). Combined, I suppose you could say there are theoretically 80 possible combinations. Realistically, current plans point towards perhaps three core splits and two access splits being implemented.
The item having the greatest impact on transformation, comes from those low latency services, called ultra-reliable low latency communications (URLLC) -- quite a mouthful. URLCC services have an iconic 1ms access-network latency target. Stating the obvious, a low latency service requires a low latency network, implemented as extensions to the hotels and edge clouds noted previously. The extensions would allow the introduction of other services at the edge, such as core network user plane forwarding or local caching. The principal idea here is to not backhaul all traffic to a remotely located, centralized operations center, as is typically done today. Doing so over thousands of kilometers won't work. Even the speed of light (at 100 microseconds per 10km) can't help to deliver URLLC targets.
The challenge in this transformation lies in the engineering of edge clouds. It requires selecting a latency target, knowledge of existing candidate locations (like central office buildings or other fiber centers), forecasting macro and small cell deployments and their capacities, and deciding on basic pooling sizes (number of basebands per edge cloud). The possible permutations and impacts are daunting. For example, a change of 100 microseconds in the latency target could change a design plan from hundreds to thousands of edge clouds.
Clearly, all of this represents a complex CTO technical task. But for CFOs, it may be even more difficult -- because it needs faith! The fact is, the proceeds from low latency or very high-capacity services is far from clear. However, to not start the journey now would be catastrophic. These things take time and a misstep now could take years to recover.
So how do all these things relate? I like to say the best 5G network is a great 4G network. That's because for quite some time, users will be moving back and forth between 5G and 4G significantly, whether at low bands or mmWave. This experience will be good or bad depending on how similar the networks are in speeds and latency. A great 4G network will make the experience better, and thus Gigabit LTE can help. It is the foundation of a great 5G network. The same applies to edge clouds. While they might be more easily justified by 5G, they will also be required by 4G to achieve lower latencies and launch the virtualization of access networks to reap the perceived opex and performance gains they imply.
Finally, what about that fog comment in the title? The IT industry gave the telecom industry clouds. Some fundamental work had been done; it just needed tuning. Then along came fog computing, a term created to indicate how clouds would extend to the extreme edges, in IoT devices. However, have you ever looked up? There are large core clouds. At times, there is fog near the ground, that's true. But there are also a lot of small clouds most of the time. The edges. So, core clouds, edge clouds and fog complete the visual image in the sky and on the ground, and summarize nicely mobile network transformation. Let's just hope there are no thunderstorms along the way.
— Michael Murphy, CTO for North America, Nokia