Facebook’s ambition to turn the telco’s network supply chain upside-down is unbridled. At last week’s annual summit for its Telecom Infra Project (TIP), operators, start-ups and other partners were discussing radical ways to build a cellular network that would be as affordable and simple as WiFi. The projects are ticking many of the MNOs’ current hot boxes – sub-$1,000 base stations, open and interoperable networks, artificial intelligence to support extreme automation.
The TIP was launched in February 2016 and now has over 500 members. Axel Clauberg, the TIP chairman, who is also VP of innovation at Deutsche Telekom, said: “I strongly believe we need some exponential innovation to master the exponential growth challenge.”
Santiago Tenorio, group head of networks strategy and architecture at Vodafone, echoed the sentiments, saying: “Greater engagement in TIP will help us to implement faster the open and flexible network infrastructure that enables the industry as a whole to reduce the cost of providing telecoms services to our customers.”
The areas of focus in Clauberg’s keynote and other opening speeches were three key enabling technologies – OpenRAN, millimeter wave, and the Voyager optical platform, plus artificial intelligence/machine learning for telco networks.
All these could slash the cost base for cloud platforms and communications networks, and so unlock massive new levels of usage for web and connectivity services, by making it cost-effective for operators to make data “ubiquitous and abundant”. These are the words of India’s Reliance Jio, a supporter of the TIP.
The mutual interest is clear – as a new entrant, Jio can only challenge incumbents if it can offer its users more data at lower prices, and that means building a high capacity network at very low cost. For Facebook, if operators really can afford to bring fast connectivity to every corner of the globe, this will greatly expand its addressable market, and the ability of users to upgrade from WhatsApp to its more lucrative services. This is why Facebook announced that it is working with operators to get fiber optic cable laid to Uganda, because that will speed the opening-up of a potentially lucrative growth market.
This commonality of interest threatens to sideline incumbent vendors like Ericsson, whose own event, its annual Capital Markets Day, also took place last week, and offered a far more gloomy view of the telco networks world (see separate item). The large vendors are struggling to find a way to work within the operators’ new rules while they are also relying so heavily on 5G to restore their finances. Yet they will have to find a way if they really are to remain central in the 5G era. Facebook’s initiatives are bold and some will fail, and few of the clever projects on show at the Summit would be trusted, as yet, to replace the decades of engineering that lie behind an Ericsson, Nokia or Huawei platform.
But they are holding out the tantalizing prospect that operators will be able to achieve an entirely different cost base in the 5G era, even if the actual products come from somewhere else. Some of them will come from established suppliers which adapt – Ciena comes to mind on the optical side. Others will come from start-ups, and those challengers may well end up being acquired by the majors. Whatever the source, operators are looking to a combination of open standards, open source, virtualization and automation to change their economics forever, and all these approaches were on show at the Summit.
At the heart of Facebook’s vision of a new network is a fully open RAN. The idea is scarcely new. Back in 2002, the IEEE standards body was using the OpenRAN label to describe a distributed processing architecture which applied data communications network principles to the wireless network, using routed IP as the underlying transport.
If the 4G era saw the economics of IP being applied to cellular and WiFi data systems, in 5G, operators and web giants believe they can go even further, and the OpenRAN name has been commandeered by Vodafone for a project which is studying virtual and commoditized network options, including trials with the UK’s Lime Microsystems, which has designed an open source software-defined radio (SDR) complete with an ‘app store’ from which developers can download software.
The OpenRAN concept was front of stage at the TIP Summit. TIP announced a new project group called OpenRAN, which will be co-chaired by Vodafone and Intel. Vodafone said the initiative ”aimed at reducing the costs associated with building mobile networks and enabling easier market entry for smaller vendors.”
The focus will be RAN solutions which can be implemented in software on general purpose processing platforms such as x86-based servers or white box switches.
This work will complement that of other TIP projects and groups as it moves towards an end-to-end definition for a virtualized, automated RAN. At the cell site, the aim will be to have extremely low cost, low power antenna/radio units and Facebook has already contributed its own designs to the OpenCellular initiative.
Telefonica, Vodafone, South Africa’s MTN, and Australia’s NBN are all “at various stages” of lab or field trials of the open source OpenCellular box, which is targeted in particular at rural areas and is based on a low cost, ruggedized small cell design with a software-defined radio. The ecosystem is broadening – for instance, a vendor specializing in off-grid solutions, Clear Blue, will provide its Smart Controller Management API for OpenCellular’s power solution. This will enable operators to monitor power usage, control battery settings and so on, remotely.
In particular, Telefonica has an Open Cellular trial running in Peru and Vodafone, having tested the waters in India, will cooperate with MTN in South Africa and then expand to other African markets. The latest version of OpenCellular’s hardware runs at 1.8 GHz with MIMO antenna arrays and can support up to 128 active users. The OpenCellular group is working on versions for 700 MMHz, 800 MHz and 850 MHz, which would take advantage of low frequency spectrum to reduce the cost of reaching sparsely populated areas.
The OpenCellular group has also built a newer low power version of the technology, which is currently in trials.
At the Summit, Facebook went a step further than it has before in terms of putting a price tag on new-look base stations, showing a concept design for an open source base station which would cost less than $1,000 for a macrocell. Like Lime’s crowdsourced activity, this is based on an open source RF board with a software-defined radio to support different radio access technologies. Facebook expects trials of a GSM version to start in the first quarter of next year and an LTE model by mid-2018.
For the fronthaul links which will connect the cell sites to the virtualized basebands, TIP has its six-month-old vRAN Fronthaul Group, which now has over 200 members. This group is working on four use cases, sponsored by three operators (BT, Bharti Airtel and TIM) plus the US cable industry’s R&D arm, CableLabs. This will be an important activity to identify ways to reduce the main cost barrier to vRAN, the need for high quality, low latency fiber for most fronthaul scenarios.
BT and Vodafone, which are co-chairing the group, said vRAN was currently hampered by a shortage of low cost remote radio units and multivendor solutions. They said: “Open interfaces for the link between virtualized baseband units (vBBU) and the remote radio units (RRU) will be key for enabling multivendor vRAN solutions.” The group has produced an open fronthaul interface that is currently being shared among operators, RRU vendors, integrators and manufacturers.
Also in the RAN area, another TIP project is called 4G Unbundled. This is developing an OpenRAN architecture which splits the LTE base station into functional components, most of which can be virtualized, and then defines open APIs to ensure that components from different vendors can be mixed and matched. The first draft of its OpenAPI was published at the Summit and supporters believe this will greatly accelerate the progress towards a fully unbundled cellular radio and a truly multivendor network.
Although 3GPP work on functional splits and APIs for vRAN appears to overlap with this project, some operators are already expressing doubts that the most powerful players in 3GPP will really produce a fully open architecture. Instead, they fear that the flexibility of 3GPP standards will enable each vendor to implement the standards in its own incompatible way, as has happened in the past with specifications from various industry groups, such as CPRI (whose lack of uniformity, and consequent vendor lock-in, is a major barrier to affordable vRAN).
Facebook believes that millimeter wave spectrum, particularly the unlicensed 60 GHz band, will reduce the cost of high capacity backhaul for small cells and rural networks. It has contributed its own inhouse design, Terragraph, to an mmWave group being led within TIP by Deutsche Telekom. Although this has focused heavily on rural areas, Clauberg said: “We have a lot of interest from city governments. The challenge with fiber deployment is dominant in developed countries as well. Having a meshed solution in unlicensed spectrum, with fiber deployment, is the ideal combination.”
Another project built initially around a Facebook R&D project is Voyager, a packet optical transport platform which now has “host of operators”, from Latin America to Africa to Europe, testing the system, according to TIP. The original triallists were MTN and Equinix, followed by Telia.
ADVA was the first supporting vendor, followed by Acacia Communications, Lumentum Holdings, Celestica and Snaproute. Last week, they were joined by Cumulus Networks, which is contributing its own Linux-based operating system to Voyager; and by white box vendor Edgecore Networks, which is contributing its Cassini packet transponder design.
Voyager is a DWDM (Dense Wavelength Division Multiplexing) transponder for fiber providers and Telia has completed a trial of 100G and 200G using the open source equipment on its 1,089-kilometer fiber route from Stockholm to Hamburg.
Facebook unveiled Voyager a year ago, hailing it as the “first white box transponder and routing solution” and promising that it would open up packet DWDM transport networks just as the firm hopes OpenCellular and Terragraph, and its Project Aries Massive MIMO array, will open up the wireless network. Voyager is the basis of the Open DWDM platform, itself part of the Open Packet Transport project, which is designed to drive down costs in the same way as Facebook’s Wedge 100 white box switch did in the data center. Voyager uses the same Broadcom Tomahawk switch ASIC chip as Wedge 100.
Facebook has opened up the design to the TIP community via the Backhaul Open Optical Packet Transport project group and has been working with partners to test the solution. Orange also says it plans to trial the hardware.
“The Open Optical Packet Transport project group in TIP pioneered the white box packet transponder concept with its introduction last year of the Voyager design which has gained significant interest in the community and from network operators,” said George Tchaparian, CEO of Edgecore. “We have long supported open white box hardware with our switch, PON OLT and WiFi design contributions to the Open Compute Project, and are now pleased to contribute the Cassini design to TIP as an open platform to provide more flexible white box choices to operators and encourage further collaboration between the packet switching and optical technology worlds.”
One of the keys to reducing the cost of deploying and running next generation networks will be full automation, extending current self-optimizing network (SON), SDN/orchestration and AI/ML efforts. For automating optical network management, Facebook principal architect Sri Bala recently presented research which is being done inhouse, but could be contributed to open source once it stabilizes.
“At Facebook, we believe in engineers building robots. Robots manage then network,” Bala said. This is enabled by a common abstraction layer based on APIs, the NMS Adaptive Layer, which can provide the interface between equipment from any vendor, and the decision maker (whether robots, manual tools or an SDN controller). There are two challenges this aims to address – the largely manual nature of current optical network optimization, and the tendency to stick to a single vendor solution, because each supplier handles issues such as wavelength assignments differently.
Facebook has also designed a Spectrum Assignment Engine, which would make wavelength assignments, taking account of forecast demand, and feeding the information in near-real time into the Adaptive Layer. That information would be fed through an orchestrator down to the NMS Adaptive Layer.
Facebook’s influential director of engineering, Jay Parikh, summed up the near term goals which remain crucial, even while some of the work groups start to look a little more blue sky. He told the Summit: “We really need to get stuff into production, we need to learn from this, we need to get data, we need to get feedback, and we need to iterate faster and faster by trialing this stuff out.”
He added: “And we need to figure out what deployment, business, operational models look like. It is still very early, this industry does not move lightning-fast, but we are going to try to change that with what we are doing in TIP.”
Of course, there is an elephant in that very room where the MNOs huddle together with the web and data center suppliers. Lowering barriers for vendors also lowers them for operators – commoditized networks combined with shared spectrum will enable many service providers to build wireless platforms, often targeting the most lucrative user bases, such as the industrial sectors where MNOs have often failed to make much impact. And in turn, as the network equipment providers risk losing their position with the large operators, they will be more incentivized to seek out new clients, and support those new service providers to enter the market, as Nokia is already doing via its managed services and cloud core offerings.
Facebook, like Google, has no particular interest in supporting one type of operator – or vendor – over another, as long as they are prepared to invest in bringing data connectivity, and web services, to every person on the planet. The MNOs may believe TIP exists to usher in a brave new world for their beleaguered businesses, but they cannot embrace the new economics of open source without also changing the way they change and respond commercially – or they will find more agile challengers grasping OpenRAN and outwitting them in the market.
Slicing and AI/ML feature in new TIP working groups:
Other new working groups were announced at the TIP Summit. One is focused on AI and ML, and will be co-chaired by Deutsche Telekom and Telefonica. There will be three streams of work:
- Running ML-based network operations, optimization and planning.
- Exploring how latency-critical services such as autonomous cars can be supported by systems that predict user behavior.
- Multivendor data exchange methods to ensure that ML models can work across the industry.
DT’s CTO Bruno Jacobfeuerborn said: “The application of machine-based decision-making and auto-remediation approaches will be key to enabling the accelerated deployment of new services while supporting hyper traffic growth at a lower cost structure.”
Another is exploring end-to-end network slicing, developing use cases and identifying barriers. This will also be chaired by the prolific BT whose CEO of technology, services and operators, Howard Watson, is joining the TIP board, along with Telefónica’s Del Val Latorre, Vodafone’s Tenorio, and Aaron Berstein, director of community ecosystem programs at Facebook.
The full list of TIP working groups is:
- Edge Compute
- vRAN Fronthaul
- Solutions Integration
- Open Optical Packet Transport
- mmWave Networks
- E2E Network Slicing
- Artificial Intelligence and Applied Machine Learning
- People and Process
- System Integration and Site Optimization
An important aspect of TIP’s work is to support the start-ups which may be able to disrupt the wireless equipment ecosystem. These are incubated in TIP Ecosystem Acceleration Centers (TEACs), which are hosted by operators (and in some cases, the start-ups also gain funding from the operators’ venture arms). DT, BT, SK Telecom and Orange already have TEACs established in their respective home countries.