I came back yesterday from NGMN’s 5G themed conference. I was covering it live so if you want to look back at what each speaker was saying, in more or less detail, then here is Day One and here is Day Two.
Those are quite unmediated reports, though. Not much reflection. I wanted to put a few thoughts down that I think are of note regarding the current “state of 5G”. I have a lot of thoughts. This is Part One, where I ask some questions. Part Two will attempt some answers, in the sense that it will look at what leading companies and research entities have been proposing so far in terms of actual technology directions.
WHAT IS 5G GOING TO BE?
The first thing, the whole question, is what is 5G going to be? You have to start and end with this question , really. All the rest is just noise. Yet what 5G will be may be as much defined by the journey as by the desired destination. So…
DO YOU START WITH THE USE CASE?
What NGMN has done, and it was its conference this week so let’s begin here, is start with the use cases that it thinks go beyond the capabilities of LTE and LTE-A. The 25 operators involved have put their heads together and come up with things like ultra-dense machine comms, ultra-reliable comms for critical use cases, highly mobile comms, and very high throughput mobile broadband in dense areas or indoors.
It has also put specific technology requirements against those use cases. So just to take one example: Broadband Access in a Crowd. The specific use case is sharing of photography and HD video. The KPIs are a 25Mbps downlink and 50Mbps uplink, end to end latency of 10ms, pedestrian speed mobility, but a very high connection density – 150,000 per square kilometre. What the industry has to do is to deliver technology that can meet those KPIs. That would, as NGMN defines it, be 5G.
Yet there are other examples with radically different KPIs. Take Number 12 on NGMN’s list: “Ultra-high reliability & Ultra-low latency” where the use case is something like vehicle-to-vehicle comms. Here end-to-end latency is a mind-bending 1ms while data rates might be a few bits per second. Yet mobility is very demanding (speed of a fast car) and device density might also be pretty high (imagine a busy traffic period I a city centre). This is also 5G.
What “5G” has to do, as we currently understand it, is meet both these use cases, and many more. So the big question is: can one “network G” do that. Can the same radio access network deliver, say, a very few bits at massively reliability with near-to-zero latency AND very low mobility, high throughput streams of live (say) HD video content. Can there be a control plane which can manage mobile broadband sessions in a moving train and a sensor lying idle by the trackside that must wake up as instantly as possible in the event of some imminent danger?
QUESTION: Why would a very low power, narrow band IoT-dedicated network need to be defined within the same “G” as something like a highly mobile, wideband nomadic node? Can it be part of the same G?
WHAT IF YOU DON’T KNOW THE USE CASES?
So while this use case before technology requirement approach looks a sensible route to follow – NGMN has also admitted it doesn’t know what all potential 5G use cases are going to be. It’s the first line of the conclusion of its own white paper: “5G is expected to have countless use cases, many unimagined today.”
So this creates an obvious and apparent contradiction. On the one hand you say we are going to define the use cases first and then the technology requirements to fulfil those. On the other hand you say you don’t know what all the use cases are going to be. So in this instance you say what you need is a flexible, open platform upon which you can build whatever use case you like.
QUESTION: Can you do both? Design networks to specific use cases and also have flexible enough technology to meet as yet unimagined ones?
TECHNOLOGY WASH, OR BUILDING BLOCKS?
This – rather crucial question – aside, NGMN has put out its 25 currently defined use cases and the KPIs (requirements) associated with each use case, and it has also outlined what it thinks will be key technology building blocks.
Yet here things get muddy. NGMN has completed what looks like a very useful technology gap analysis and produced a long list of technology building blocks, lots of which are new things. But that building block list also has in it just about everything in it currently going on in networks.
It has SDN, NFV, RAN co-ordination like CoMP, dual connectivity, licensed-unlicensed spectrum aggregation. These are “5G” technologies in the sense that they may fulfil some or pieces of some use cases. But they don’t diverge from the currently chosen path itself. Having a section on “network flexibility” is a crucial element, of course, but do we think that operators are likely to head this way in any case?
It’s more of a technology shopping list and an evaluation of the current state of play, than anything that says – this is 5G radio, this is 5G control plane and so on. It’s a bit of a technology wash.
YOU ARE MAKING IT MORE DIFFICULT WHEN YOU CALL IN EVERYTHING AS A PART OF 5G . Many of the systems building, SDN, NFV, carrier aggregation, clever coordination, analytics-driven network operations, horizontal architectures, distributed cloud etc etc – these are happening anyway. You could argue that it makes sense to roll them forward into 5G. That may be a valid stance to take – anything that enables 5G is 5G. But it’s pretty muddy. This is exactly why, of course, 3GPP is sticking to IMT2020 as its definition, and currently thinks 5G is a marketing term.
QUESTION: Who has to unpick all that into something that is 5G? Can NGMN present its requirements and also somehow make sure these are processed within the standards organisations as deliverable standards that will, as deployed technology, deliver on those requirements?
INCONSISTENCIES AND CONTRADICTIONS
Because of this “everything is 5G until told otherwise” and also this “let’s define all our technology requirements at once” approach, there’s still an awful lot in play here. That leads to an approach where you might end up with competing requirements. Let’s take an example, One requirement is for “foundational shifts in cost and energy efficiency.” One way to do this was proposed by Ericsson at the Conference – networks that are asleep when not required, and use energy only when transmitting traffic. But then, how does this square with networks of sensors that can wake up instantly with very short or even no signalling frames? If the network is asleep will these sensors have to wake the cell up? How? Or will we have to sacrifice some energy efficiency for the reliable IoT that we want?
This sort of inconsistency is, I think, to be expected. This is because we are still at a stage where 5G can be what you want it to be. The process currently taking place should look to resolve these, and it will do as long as those involved are willing to admit to and recognise them, and not try and brush them under the carpet.
Are those involved mature enough to admit that inconsistencies and contradictions exist and to do something about that?