At RANWorld, held in Barcelona last week, there was a clear focus on Cloud and Virtual Radio Access Networks, and specifically how to do it. Can the vendors “open up” to enable market innovation on the radio side. How will a disaggregated RAN be architected? How will it be managed?
So here are the things that TMN noticed as, er, noteworthy in that area.
TELEFONICA: Juan Carlos García, Director of Access Radio Network & Transport Optimization – Telefónica
We’re IoT ready, we’re switching off 2G as much as possible. The Spanish giant wants to have only a very narrow bandwdth left by 2020 – fewer than 5MHz bandwidth to support remaining 2G connected devices.
In three years we’ll have new features for 4G macro layer – such as 5x carrier aggregation and new features in antennas will be 4×4 MIMO, 256QAM.
Also from 2020 we will consider RAN sharing scenarios, deploying/adding unlicensed spectrum for small cells. This is all running in preparation for 5G – as we clean and simplify 2G we’re making room in sites for new antennas and tech.
On NFV and the Telco Cloud Architecture, we are developing a single environment to connect people and things at the same time. Our Core READY programme is building the enablers. UNICA is the machine across the network that will support the telco cloud. A software defined transport network will have the ability to deliver services across the top.
On RAN virtualisation there’s work to do. We need agreement and an industry standard on how to do it and where. The vision is a distributed vRan with RAN functions on general purpose machines and we still haven’t seen it.
Our core is already 30-40% virtualised. On RAN virtualisation there’s work to do. We need agreement and an industry standard on how to do it and where. The vision is a distributed vRan with RAN functions on general purpose machines and we still haven’t seen it.
Our need for this is because we face an order magnitude density increase and so need to increas efficiency significantly on radio stations, on sites, and we need to be able to scale very much. Can only do this with a virtualised solution. For example, a virtual environment will enable us to assign idle Radio resource where it is needed, instead of building to capacity. Work is needed in the industry on agreeing how to to do this in a standardised way.
Finally – we will support all this with an orchestration mechanism to auto configure and manage the telco cloud.
It is the OpenSource environment that will take us to this new paradigm.That’s because the most important part is creating an ecosystem to be able to deliver new services.
However we will not run to deploy MMwave radio. If we want to accelerate mmwave need to work a lot on use cases and finding the services that may trigger the invesment.
We can get a long way with LTE-A PRO. In R13 – what Nokia terms 4.5G PRO – it can support 1Gbps and have radio platforms “5G-ready”. Nokia’s 4.9G will see bit rates beyond 1Gbps and latency sub 2ms, supporting massive mimo with LTE. The vendor expects to see Cat18 devices by next year, based on Qualcomm X20 modems.
5G is needed as a step change where even this 4G performance is not enough
Moving forward it says that R14/15 will provide theoretical round trip time 6ms. By playing with OFDM numerology, 2 OFDM symbols and TTI of 0.14ms and will get less than 2ms latency in 4.9G.
So this means that 5G is needed as a step change where even this 4G performance is not enough. That means use cases such as stadiums (lots of people, extreme overall capacity requirement, low latency), with security & emergency services and consumers all on the same network. Another example would be manufacturing where robots and humans co-exist, with robots wirelessly connected instead of over fixed line connections.
BT/ EE: Mansoor Hanif, Director, Converged Networks & Innovation
We are looking for tech that can get us to 100% coverage indoor and outdoor, and within that prioritising railways, motorways and rural. We’ve talked before about all the potential coverage solutions. One new thought – we’ll be putting mmWave into tunnels as bounces off walls a lot better than sub6GHz, which means we don’t have to get equipment inside tunnels.
We are also preparing for ESN (Emergency Services Network) over 4G on shared frequencies, that’s like a network slice on a big scale: partners are getting really busy rolling that out.
Virtualisation: vEPC is low hanging fruit – relatively simple compared to vRAN. There’s a crossover in direction – the core, once virtualisable is headed to the RAN while the vRAN is going towards the core.
Multi access small cell evolution is driven by vRAN. vRAN is what we are calling it (not CloudRAN). Disaggregation of the vRAN depends on low cost pluggable wideband RRHs with PoE – but it cannot lead to a total loss of control.
We also have innovation on backhaul and fronthaul – XHAUL – where the dimensions of transmission are increasing. There’s a debate on the future of CPRI, eCPRI is not locked in, and that could be a situation where it will not work out for operators so we need to make a big push to make sure not happen again.
Edge computing – accepted as fundamental bedrock of 5G. The cloud moving to the edge and edge moving to the cloud and they are merging into a single platform.
Standards are allowing challengers to come to our space with MulteFire and we need to compete with that solution. We cannot be arrogant enough to not look at technologies. (If that’s what our customers want we have to offer and be expert as well but need to show added value in orchestration and management of that network)
Converged user experience: there’s a set of capabilities we can configure for customers either from MEC or a converged core and we should set these ourselves rather than interwork with devices. Things like traffic management, ids, network selection on bandwidth, battery life, latency, mobility between access types.
vRAN fronthaul project overview
For virtualisation, the RAN business model less clear. People are claiming no cost saving except for very simple business cases. Needs be multi vendor for a safer and stable future
First endeavour in vRAN was with Cavium (on ARM). Will put infra from Intel and ARM into BT labs – ARM looking at VNFs on ARM infra, putting OPNFV on top and bringing as many VNFs as we can.
Onboarding VNFs for ARM – not many companies will do that so need show compatibility.
Infrastructure market has Intel moving to edge, and we have ARM moving the other way from mobile to data centre and so there will be will overlap. We need to be clear about what we virtualise first and where, so we can move functions back and forwards over ARM and Intel. Needs a lot of work to get there.
It’s not that we don’t like Intel but we think we need at least two architectures.
vRAN focussed where there is no ideal fronthaul, so use as much underlying infra as you can.
CableLabs are using Docsis as underlying infra. Requires an open API between the RRU and BBU, with the RRU and vBBU reconfigurable based on software on general purpose hardware. SIs partnering with are Radisys and Mavenir,
Mavenir at the moment working with a proprietary split for fronthaul – compressing the fronthaul -but are working to open that and come to a split that’s viable as an open interface. Can also test other interfaces, eCPRI, xRAN. Still feel eCPRI too much a grey area between what has and has not been specified and will result in our view that it will not be open enough, and will be too vendor specific.
BT use cases street coverage with neutral host. Campus and temporary coverage. TIM on vRAN cluster in a HetNet
Still away from a proper licensing model. Need to sit down and talk about it, when virtualise cannot just apply software licensing to virtual model. Where we make revenue the vendor gets a chunk of that. Where we cannot charge for it, we cannot pay you for it. We need understand that it’s not a free for all, need to be pragmatic.
Open Air Interface very interesting for virtualisation: OpenSource software and Open hardware – Hardware having programmability as an option is very useful (FPGA) when you go to ASIC, sinking 300m dollars sunk into one ASIC, you are only going to do that when you have big market.
Containers has to happen. There’s a misunderstanding of the complexity involved. The complexity is the flexibility – and we can solve that with containers
Flexibility and management: Containers has to happen. There’s a misunderstanding of the complexity involved. The complexity is the flexibility – and we can solve that with containers. Containerisation is a means to get to that goal and we are driving hard to get that containerisation operation.
QUALCOMM – Lorenzo Casaccia, VP, Technical Standards
5GNR will be OFDM technology with scalable numerology and TTI, but structure modified compared to LTE. Scale a bit to fit different bands and forward compatibility.
First targets are at 3.5GHz spectrum or 4.5GHz in other parts – 3D beamforming with 256 antenna elements. 200m intersite distance, 48dbM transmit power. Median user perceived throughput goes 3.8x from 52Mbs to 194Mbps. And cell edge perceived throughput from 27Mbps with 4G 4×4 MIMO to 79Mbps with 5GNR Massive MIMO.
In 5GNR beamforming is much more accurate than in LTE because the updated signals (SRS signals) are 10x more frequent in 5GNR than LTE. That gives a BTS much more accurate understanding of the position of a user device on a millisecond basis. If you can have more accurate beams then overall capacity naturally increases.
Just by using existing urban cell sites mmWave can offer pretty good performance for outdoor users
NR is designed to allow 3.5GHz to be a macro type signal. That makes 3.5GHz not just a local deployment but more a macro type deployment band – thanks to the beamforming algorithm.
We have been mapping mmWave prototyping to real city maps, for 15 cities, and we have simulated that just by using existing urban cell sites mmWave can offer pretty good performance for outdoor users. That’s by co-siting 28GHz radios with LTE. (Editor’s note: the significance of this is that current industry orthodoxy is that mmWave will require a much denser grid than currently, necessitating a new process for site acquisition, design and approvals. Note, though, that Qualcomm’s models are on already quite dense deployments, not on the widest macro cells)
That would also free up sub 6GHz resources for out-to-indoor capacity. Co-siting 28GHz mmwave with LTE.
MAVENIR: John Baker: SVP, Business Development
We’ve gone back the drawing board to rebuild the RAN from the ground up.
New vendors have a chance to open up the market. We’re a firm believer in pushing open the interfaces in the RAN – market has been trimmed down to three vendors because of these closed interfaces.
Virtualisation opens door to more places in market where can deploy LTE networks.
Can get together to work out open interface between BBU and RRH. eCPRI specs from a week ago – if anyone can interpret come and tell me because is a cloud with 250 odd vendor specs sitting in there. Will end up being interpreted and theefore be a proprietary tech.
We’re a firm believer in pushing open the interfaces in the RAN – market has been trimmed down to three vendors because of these closed interfaces.
RF over Ethernet in another forum, just by moving MACH layers around can get 10-1 bandwidth requirement difference = 1/10th bandwidth.
xRAN is driving that – product of an open interface based on the Split 7 architecture that allows you to upgrade the radio easily. There will be a public spec at the end of October. That will allow manufacturers to make BBUs or RRHs and then get into interoperability testing. Lot of effort to disaggregate radio function from the software.
In this CUPS (Control User Plane Separation), architecture means you can get to disaggregated and functional based platform that is all software and runs on a low number of cores. Have done already with full EPC, have it ready today where can split out control and user plane as logical elements. Allows you to think about how to place entities in the network from core to edge, but splitting functionality apart today you can start deploying 5G advantages in 4G networks.
Could see applications like a cell site as an enterprise white box – running on 8 cores complete – a lot of zero touch on enterprise side. This network at edge simplifies micro services, run straight off the EPC: APN routing straight through EPC using distributed EPC gets close to end user – could be all traffic dropped off to the enterprise.
See “build your own network” – white box to a premise, download firewalls, wireless network, then through agreements setting up enterprise.