Your browser is not supported. Please update it.

31 January 2018

Nokia comes full circle, differentiating its 5G offering with its own silicon

It has come full circle for the big mobile vendors. They used to control every aspect of the design of their platforms, from antennas to processors to devices, but increasingly they looked to reduce cost by focusing on a few key elements and delegating the rest to partners. In the age of virtualized networks, it seemed that trend would intensify, as the game became all about software on commodity boxes.

Except that it didn’t. In the access and backhaul networks, the remaining physical components, such as antennas, have to perform at unprecedented levels to meet the demands of the massive data loads the virtualized architectures promise to support. There is differentiation, once again, to make in the chips and the hardware, even if the digital processing and intelligence is embedded in virtual network functions in the cloud.

This return to the high-stakes game of differentiating via hardware is clear in Nokia’s pre-Mobile World Congress announcements, which set out an all-encompassing architecture called Future X, but center on an infrastructure processor called ReefShark. Given that Nokia recently announced commercial availability of another heavyweight chip, the FP4, which powers its ‘petabit-class’ router, it is clear that the Finnish vendor has high ambitions to drive the massive hardware that will be needed to run the software-driven carrier network – as well as that software too.

Just as the FP4 underpins a broader architecture, and a portfolio of hardware products, so ReefShark will initially be at the heart of the latest AirScale 5G-ready base stations, but will evolve to run a wider range of offerings. By investing in these processors, Nokia is creating a broad integrated platform which can be targeted at webscale and data center giants, as well as telcos. It is also setting its cap against the merchant chip vendors, most notably its close ally Intel, which also has visions of an architecture to power the virtualized carrier and webscale networks of the future.

The Future X architecture as a whole is perhaps the most coherent end-to-end blueprint  which a vendor has produced for 5G as a software-defined network (SDN) enabled by full virtualization and automation (though Huawei arguably has more of the pieces in place, and will no doubt set out its own updated vision when it stages its pre-MWC event next week).

What both these vendors are doing – but Ericsson rather less successfully – is provide operators on the verge of 5G trials or deployment with a broader roadmap which smaller challengers will not be able to match. They are highlighting their differentiation in the highest value areas of the network, and those with the highest barriers to entry for others – the baseband, the high end infrastructure processors, the orchestrators which will tie it all together and enable next generation activities like network slicing.

This, they hope, will raise those barriers, not just against emerging new software-driven start-ups in the open source world, which the operators are championing to reduce their costs dramatically in the 5G era (see separate item on Orange); but also against giants trying to enter their territory from another market, whether the webscalers themselves, with their cloud platforms and hosted services; or the enterprise integrators like HPE, bringing their knowledge of data center virtualization to the party.

Future X has been Nokia Bell Labs’ label for its 5G meta-architecture for some time, but is now coming a step closer to the real world. Last year, Nokia set up a joint venture in China with industrial investment organization China Huaxin to engage in long term research projects aligned to Future X Network, and so seed the ground in that country. Now it is opening up Future X to the world, outlining an end-to-end architecture encompassing the data center, core, RAN and network edge.

It includes some existing commercial elements but is firmly focused on 5G. Based around SDN orchestration, it automatically scales and adjusts in response to different applications, traffic levels and service requirements. As such it can support network slicing, in which virtual slices are created – sometimes dynamically and temporarily – to provide optimized connectivity for a particular service on-demand. This is only possible because, in a virtualized network, different network resources can be dialled up and down as required.

The eight elements of Future X are:

  • Nokia’s 5G New Radio implementation, based on the newly finalized 3GPP specifications for 5G NR Non-Standalone (still requires a 4G anchor network and core).
  • The modular AirScale Radio Access, which will be the first to take advantage of ReefShark.
  • 5G Small Cells.
  • AirScale Active Antennas – ReefShark will not just boost AirScale, but will enable far smaller, more integrated 5G antennas, which will have significant potential impact on network efficiency (see below).
  • Massive Scale Access – other access technologies are supported to complement 5G NR, including FTTx (fiber-to-the-x), cable broadband and wireless PON (passive optical network).
  • 5G Anyhaul – the transport category includes SDN-enabled fronthaul, backhaul and midhaul including a range of technologies which can be mixed and matched according to requirements (NGPON fiber, microwave, optical transport).
  • The 5G core platform uses the FP4 processor and supports fixed and wireless services, and slicing, in a cloud-native environment.
  • And of course, there is a range of ‘5G Acceleration Services’ which try to boost the value of any contracts by adding consultancy and professional services.

Automation, including the use of machine learning to support intelligent, flexible network optimization, is a big theme too. Nokia has been one of the more active large OEMs in harnessing artificial intelligence to improve network performance, and clearly sees this as a differentiator for its 5G offerings. “Automation is key to realizing the full operating potential of 5G and dealing with the massive number of network connections and their different requirements,” it said in its release.

“With our 5G Future X portfolio we are opening up network data and network intelligence to our customers to jointly program and tailor machine learning and automation that runs on our new silicon,” Marc Rouanne, president of mobile networks said. “The Future X architecture invented by our Nokia Bell Labs research has made it possible to mix the knowledge across Nokia, between IP, Optics, RF, software and innovative inhouse silicon.”

As for ReefShark, Nokia claims it can increase the throughput of a cell site by three times, delivering capacity of 84Gbps. It will initially be offered as a plug-in unit for the AirScale baseband module, which promises to be software-upgradeable to 5G. By boosting the baseband unit’s throughput with ReefShark, multiple AirScale modules can support throughputs of up to 6Tbps.

The chip will also significantly reducethe size of a Massive MIMO antenna (this has been a barrier to deploying this core enabler of 5G performance on a widespread basis). It will allow new antennas to be shrunk from current heights of between one meter and 2.6 meters, to about 45cm; and the chipsets support automated beamforming, all of this helping  to boost cell range and capacity, and give greater flexibility about where antennas are installed.

ReefShark also claims to cut power consumption of a cell site by 85%, another critical concern for 5G, and to use embedded AI to optimize radio resource usage.

Nokia will ship the chip in volume in the third quarter but says it is already engaged with about 30 operators, its first named supporter being the UK’s BT, living up to its long-held reputation as the most willing of operators to try a new technology.

The risk for Nokia is that its approach is out of touch with the times. It is trying to prolong the age when it could provide operators with a single-vendor system, controlling all the elements from backhaul to access to silicon. This seems to fly in the face of what operators claim to want – an open architecture into which they can slot commoditized hardware and innovative VNFs from many suppliers.

Six years ago, Rouanne was heading up another launch, of the Liquid Net architecture. This marked Nokia out among the large network OEMs as the one which had adjusted its thinking to the looming spectre of virtualization and SDN – which were poised to undermine the economics, margins and entrenched positions that had delivered such success.

With Liquid, Nokia invested heavily in a flexible software-driven approach. It acquired companies like Eden Rock, for its multivendor SON offering – and not only kept it multivendor, but increased its flexibility with open interfaces. It cooperated closely with Intel to meet the convergence of the IT/compute and telecoms worlds – the partners’ Liquid Apps platform, which installed compute and storage resource at the base station, was an early step towards the edge computing models of MEC and fog.

So is Future X, for all the technical wizardry of ReefShark, a step back to the old walled gardens? To some extent it is, but it may prove to be an essential one. The hallmarks of 5G RAN – virtualization, Cloud-RAN, Massive MIMO, ultra-density, ultra-low latency and what Deutsche Telekom calls “brutal automation” – have one thing in common. They all require massive compute power.

That means the processors will remain a source of high value and differentiation in the network, and far harder for others to challenge, or operators to replace, than a clever piece of software. If Nokia does not design its silicon inhouse, it will rely on Intel or other partners for the most valuable and strategic part of its platform. It may be happy to stick with Intel chips for the server elements of its offering, including AirFrame, but in the bowels of the network where its power lies, it needs to control performance.

This is why Google, Amazon, Facebook and Microsoft all work on inhouse processor projects, because they are similarly reliant on driving ever-higher levels of compute performance, at lower cost and power consumption, into their platforms. Nokia’s telco customers will face the same challenge, and if Nokia cannot deliver those capabilities, there will be one more reason to move one’s virtualized network to run in the AWS cloud.

Even if that proves to be the end game, Nokia will still have a strong role in the value chain if it can continue to drive the physical elements, like the antennas, and even provide the processors to the webscale giants, as it hopes to do with the FP4 and the petabit routers.