The URLLC debate (pt2): Why is ultra low-latency 5G so hard? (It’s not just about the radio)
In the 5G firmament, the constellation of ultra-reliable low-latency communications (URLLC) shines brightest and newest. The other star clusters, outlining enhanced mobile broadband (eMBB) and massive machine-type communications (mMTC), do not cast the same light.
They are familiar already. In essence, they will bring go-faster versions of traditional one-to-one services and no-frills versions of newer machine-to-machine services. URLLC is different. It promises to change the world by giving life to instant, intelligent, animated systems.
The great promise of 5G networks, bringing industrial and social change, hinges on their URLLC capabilities, and not the enhanced functions mobile operators are presently deploying under 5G banners.
And yet while the promise of URLLC burns brightest, it does so only in our imagination. There is nothing to see, as yet, and nothing to grasp. It remains out of reach.
“URLLC is perhaps the most complex part of the 5G end-to-end architecture,” said Anthony Magee, principal engineer at ADVA Optical Networking, during a panel session at URLC 2018 in London earlier this month.
“There is a reason operators will do eMBB first – and that’s because it is known, and has predictable revenues. URLLC is hard. It’s not going to be easy. It’s not just about the standard from 3GPP. It is going to require an end-to-end discussion.”
Part of the challenge with 5G for network engineers is how to pin the use cases down. “It is led by the business case; the economics will really be the driving force for it,” said Magee.
The inherent fragmentation of industrial 5G use cases raises questions about its scalability, a mark of the success of the GSM and LTE standards in particular. There are at least two categories of URLLC, even, and people mix them up, said Xueli An, principal researcher at Huawei, at the London session.
One is for wide-area coverage, leveraging real-time artificial intelligence (AI) and bandwidth-heavy mixed reality technologies. Autonomous vehicles and immersive entertainment services are use cases. The other relates to operations within controlled environments, like industrial plants and factories, where the signal channel is controlled and the environment is predictable.
“The solution is quite different in these two environments, whether to consider URLLC applications in wide coverage or in a control environment,” she said.
But the engineering challenge with URLLC goes beyond the upgrade of the radio network; it requires swapping-out of the core network and transport functions as well.
The URLLC 2018 session in London discussed key technology enablers to bring the 5G system together, notably multi-access edge computing (MEC), network slicing, and latency synchronisation. The last of these receives less press. Magee at ADVA is a self-confessed “sync-fanatic”; his explanation of it was helpful.
“It is well known in mobile networks that you need to synchronise your air interface. That isn’t going to go away. If you disaggregate your base-band unit into a virtual front-haul, then you will need to synchronise that part of the network even more tightly,” he explained.
Magee broke down the discipline in the context of both functions of URLLC. To achieve low latency, and to offer assurances around latency, the network should be time aware, he said. For it to be time aware, all the devices in the URLLC slice should be synchronised to the same time-base.
“That will give you a way to build a low-latency slice across the infrastructure,” he said.
To achieve ultra reliability, as well, operators should “bound the latency” to upper and lower limits, he said. Emerging technologies like time-sensitive networking (TSN) will allow the “policers and shapers” used in traffic management to be time aware.
“That’s where SON is now becoming part of the URLLC discussion,” he said.
Mansoor Hanif, chief technology officer at UK regulator Ofcom, and chair of the London session, noted the “deep hardware and software trade-offs” at every turn. “Future URRLC networks will rely on many technologies across many fields,” he said.
Hanif asked the panel if this “toolkit of enablers”, covering also MEC and slicing, is more related to the 5G ‘new radio’ (NR) component or the ‘next-generation core’.
Mark Gilmour, director of portfolio strategy at Ciena, commented: “The sad answer is it will be quite complicated, because there won’t be one-size-fits-all. It tends to point towards something modular in terms of hardware and software, for operators to be able to pick what is needed. Because ultimately that end-to-end architecture has to be flexible and programmable.”