Network slicing and net neutrality rules in Europe

This webinar discusses key topics relating to offering network slicing and specialised services following the EU net neutrality regulation. We consider the following.

  • Network slicing in 5G networks
  • EU Regulation and the BEREC guidelines
  • Specialised services
  • Are network slices specialised services ?
  • Do we need approval ex-ante? Are there implications for “best efforts” traffic?

We highlight practical implications for operators and regulators.

This on-demand webinar is of interest to operators, service providers, policy makers and regulatory authorities involved in net neutrality and specialised services.

Transcript of Webinar Q&A

Question: Are mobile IoT services classified as specialised services (this could include data and/or voice)?

James Allen: It’s possible mobile IoT services could be specialised services. It’s also possible that mobile IoT services could be just Internet access services. They would be specialised services if they wanted to use some characteristic of the network in a way that was different to IAS. I have asked a regulator whether you could have a service that was ‘worse’ than the Internet access services – in other words, imagine a service that had very limited capacity or was designed to use up spare network capacity, say in the middle of the night when it would be very efficient to do electric meter reading and there was a big dip in Internet access traffic between, say, 4am and 9am – and whether that would be allowed under the Regulation. The regulator was quite pragmatic and they said ‘well, if it was a problem, we’d find a way of interpreting the rules such that it wasn’t a problem’. So, I think they probably could be specialised services, but they wouldn’t necessarily be specialised services, and you’d have to see what you needed.

Question: Do you think there’s an expectation of the time difference between networks slices in advanced 4G networks and in 5G networks?

James Allen: That’s a very good question – I’m afraid I don’t know the answer. I will have to ask my colleagues in the software-defined networks area of research whether they have some timing numbers. If I find an answer, I will send it to you separately.

Question: How should the BEREC guidelines be refined to provide more predictability for market players?

James Allen: I think that’s quite tricky. BEREC had a very short period to produce the guidelines. I think the guidelines do go beyond the Regulation in some ways and that might end up being litigated, although I’m not a lawyer. I think for some things, such as the notion of objective necessity in this context, you could write a BEREC document about how you test whether something is objectively necessary in this context, and you could say a network slice which is competing with Internet access is allowable if there’s commercial demand for it, as long as the other rules (e.g. about not crowding out the bandwidth) are still met, because that is the objective necessity as somebody is prepared to pay money for it. And so, you could refine what you’ve got to remove some of the bits that are unnecessary. Beyond that, I think there’s probably quite an interesting, difficult question – I know BEREC has done work on measuring Internet access performance, for example, and is working on tools for whether there can be some kind of standardised measurement. That’s the very start of some science investigating the question of: ‘how do applications on a common network interfere with each other?’. At the moment, we don’t even know how to measure it properly except possibly for single applications in single networks, which is like a test case or a toy by comparison.

Question: If you have a network slicing between two telecom operators, would this be interpreted as a ‘specialised service’?

James Allen: I think you’re probably thinking of some kind of network sharing or mutual MVNO deal. I think it isn’t necessarily a specialised service; it could be that it is providing Internet access services; and in that context, the same rules would apply. So, the slice could have more slices inside it, or there could be several slices, some of which are specialised services and some of which are not.

Question: What is the impact of network slicing on infrastructure requirements? Will there be more need for infrastructure, or will spectrum optimisation lead to lower infrastructure needs?

James Allen: I think the interesting question about all this stuff to do with virtualisation is whether the network has less active equipment in it, or the active equipment in the network becomes much more generic. So, you could imagine having white boxes that are effectively commodity processors at many of the current network locations, including base stations, and having a lot of the processing being virtualised across those. So, you could have a virtual base station and a virtual remote node controller, or even the roles of those in the processing could be further desegregated into smaller lumps which could then be located anywhere within the network. And to have that full flexibility, it’s quite likely people will be more interested in having higher bandwidth and lower latency links to the base stations, and more fibre. So, you can imagine the need to get your infrastructure onto the fibre network will be even more important than it was previously, because part of the reason for having local processing is to reduce the quantity of data you’re shipping around so you get full flexibility with more fibre in the network. In principle you might get better efficiency and that might mean you needed less spectrum, but I think it probably means more fibre than other things.

Question: Where do the BEREC guidelines allow blocking in hardware?

James Allen: It’s not necessarily blocking in hardware that the guidelines allow; maybe I misspoke there. It’s that you’re allowed to offer something that only has access to part of the Internet if it’s done in hardware. I can probably look up exactly what it refers to offline. I think what the parliamentarians were thinking of is devices such as kindles which can do downloads. A kindle is obviously using the Internet, but it’s effectively only using a VPN back to Amazon’s private servers. I’ll try and get back to you offline on that one.

Question: How do you feel about RAN-sharing? Would it be a viable alternative to network slicing?

James Allen: Well, RAN-sharing is here now, and it does something different. It doesn’t need the virtualisation and software-defined elements of network slicing. In a previous question answer, we talked about whether you could use slicing in order to do network sharing, so I guess the answer is: you could use slicing to do network sharing but you don’t have to, and it doesn’t change the way that works. RAN-sharing is a good idea; to the extent that the benefits can be passed onto consumers, it’s an excellent idea.

Question: Could slicing be used to offer different flavours of Internet access (e.g. consumer, business-grade, etc.)?

James Allen: It’s true that could have a consumer-grade product and a business-grade product, and they might have different characteristics such as different speeds, but it is difficult to work out how you would be allowed to discriminate between the packets. Once the packets are in the network, you sort of have to carry them equally.

Question: Do you think eliminating net neutrality is the best solution for the seamless introduction of 5G?

James Allen: This is a political question; it’s not really a regulatory question. Net neutrality probably wasn’t needed in the first place. Now that we have it, we can work within it and we can probably deliver the vast majority of things that people want or need within the guidelines that we have. If you want to run a political campaign to get rid of it, that’s a separate matter, I think; but we can live with what we’ve got.

Question: How does network slicing differ from the current priority mechanisms such as QoS?

James Allen: I think that’s a very good question, because we see that there are QoS mechanisms in mobile networks, and they are very rarely used. In other words, these things have been put in at great expense, but they are not used much within actual service provisioning in the networks. I’m not sure I’ve ever seen a retail service offered which actually used prioritisation; maybe some people are setting their VoLTE up that way internally within their networks. But the net effect – and this has happened several times – is that network operators and vendors have built on a number of technologies different ways of doing prioritisation inside the network, and for network engineering purposes they’re quite useful, but in terms of end users they haven’t been very successful in actually getting people to use the flexibility that’s within the network. Now, maybe we’re starting to see some of that with the more software-defined networks, but it’s still an arguable case that every time there’s an opportunity to bet on either some kind of gold-plated, quality-enabled network or the cheapest possible thing, the Internet has chosen the cheapest possible thing and been right so far. So, it’s a good question – the difference is that network slicing is not just about prioritisation; it can do things that you can’t do with just priority (it can try and affect latency, for example).

Question: Is there a legal difference between slicing in the access (e.g. in the RAN) and slicing in the core network (e.g. of a mobile operator)?

James Allen: I don’t think there should be, but I’m not a lawyer. I don’t think the Regulation discriminates between core and access. The only way the Regulation guidelines do discriminate is that they talk about some things being okay if they’re done in the end-user terminal but not in the network – which, as I’ve already said, doesn’t make much sense.

Question: Is there any expectation or anticipation as to what regional size (e.g. a campus, a city, a district, a federal state, etc.) a network slice could be implemented across?

James Allen: I think network slices could be implemented across any scale, frankly. Their scale is really restricted to what you think their usefulness might be. I would imagine the campus scale could be quite interesting for certain things. Anything to do with low-latency is intrinsically local because the speed of light limits how far you can get in so many milliseconds. So, small-scale geography can be interesting, but with the IoT case we talked about, those will presumably be national or pan-national.

Question: Is it possible that a single network slice be used for various services (for example, used for both healthcare and self-driving cars)?

James Allen: It could be, but I think the idea of slicing is that slices are cheap to set up. So, there would be no actual necessity to share a slice, because you could just say ‘I’ll have a slice that looks very much like his slice, please’, and then you could have your own one and manage it as its own thing. I guess the only case where you might say ‘oh, I’ve already got a low-latency slice’ is for test purposes, or because you find it difficult to convince your regulator to approve your setting up multiple network slices (which we don’t want it to be). But I’m guessing that’s not how people are thinking of this. I have a colleague who thinks that, in some sense, there won’t be a common view of the network, and thinks that every end user will have their own view of the network. So, in their view, slices will be one each or multiple slices each, which just shows you the difference. We’re talking about a few MVNOs on the network, and they’re talking about – potentially – multiple slices per customer; so, ‘let’s have a few million of them’. That’s completely different.

Question: Why would broadcast IPTV benefit from network slicing?

James Allen: I think the idea here is to think about what’s happened in the fixed network. If you’re doing managed television services, maybe you want to make sure they get the bandwidth in order that the streams don’t have to buffer to deal with jitter… especially for live services, if it’s downloaded it doesn’t really matter, and if it’s not live, then you can have a big enough buffer such that if there is jitter, you can deal with it. So, effectively, Netflix works beautifully with a ten-second delay. Once you’ve started the film, you don’t notice the ten-second delay, and the ten-second delay can cope with the fact that the network may be up and down a bit within the ten seconds, but on average it can still keep the pipe full. That’s why broadcast IPTV is so fussy about getting its bandwidth all the time. There’s a scientific term about bandwidth averaged over a time period; if I remember it I’ll send it to you. I was referring to “effective bandwidth”.

Question: Would it be possible to run different versions of the mobile standards on the RAN using the slicing mechanism (e.g. for interoperability or backward compatibility needs)?

James Allen: I don’t know. I suspect that’s not what people are thinking of. I think, although it is possible to do things like run different modulation schemes in different time slots, you need the end devices to be compatible with that. And so, until you go to a completely software-defined terminal, I think the RAN interfaces are going to be these major generational things with software upgrades, rather than having something that is completely up in the air, at least in the medium term.

Question: How many slices could possibly be introduced – is it five or five hundred?

James Allen: I think it’s at least five hundred, and it might be five hundred thousand. It could be a lot.

Question: Do you think classifying specialised services based on commercial demand is kind of tricky, because use cases such as autonomous driving will have significant commercial demand?

James Allen: Well, to my mind, it’s objectively necessary if there’s commercial demand for it. Maybe autonomous driving has some kind of commercial demand – that’s true – but it doesn’t necessarily mean that autonomous driving wants to use a slice. It might be that it uses normal Internet access mechanisms and makes its own decisions about what kind of communications it has, because a car still needs to work even when there are no communications (in a tunnel; underground; in your garage). So, if the autonomous driving service providers say to the network operators ‘we really need this service, and we’re prepared to pay slightly more to have these particular features’, then I would say the objective necessity criterion is passed, and then what the regulators would have to check is that by offering those features to some fraction of the users, the Internet access service still gets its capacity. I hope that answers your question.

Question: How will the USA’s plan to roll back the 2015 Open Internet Order impact network slicing?

James Allen: I think the answer is: if European regulators do this right, then network slicing will be just as usable in Europe as elsewhere, because the rules for specialised services seem to allow what people would like to do. If we get it wrong in Europe, and other people without these rules manage to use slicing in more innovative ways, then it will become obvious to us that we’ve got it wrong, because other parts of the world will use network slicing more effectively than we will. At which point, I guess, there will be pressure to change our rules, or how we interpret our rules, to free that up. So, it will be interesting to see whether slicing takes off faster in countries with no net neutrality rules. That will be an interesting question.


James Allen, Analysys Mason
Partner, Consulting james.allen@analysysmason.com +44 1223 460 600