Sometimes, memorandums of understanding between multiple large companies are followed by years of silence as the parties struggle to coordinate any concrete deliverables. Not so with the MoU announced on January 20 by Deutsche Telekom, Orange, Telefónica and Vodafone, joined a couple of weeks later by Telecom Italia (TIM). Less than four months later, the five partners have already issued their first joint paper, which outlines their key technical requirements for Open RAN, and sets an ambitious target of commercial readiness by 2022.
This sets out their demands on any vendor that wishes to be part of the European operators’ supply chains, and it provides far more detail than previous statements, or the Open RAN RFIs issued by Vodafone and Telefónica from 2018. The operators say that “vendors should provide a comprehensive set of features for Open RAN solutions to be on par with the best-in-class traditional RAN networks” which should “minimize any potential gaps in the features roadmap of interest for Europe. RAN features should support high performance in terms of user throughput, capacity and latency”.
The breakneck speed, by large mobile partnership standards, of producing the document is admirable, though should not be allowed to derail the process through ‘more haste less speed’ (see comment at the start of this report). And there are very big challenges for an emerging ecosystem, consisting partly of players with limited at-scale deployment experience, to meet the rigorous demands set out by the operators.
They are right to be demanding, if they want to introduce open, cloud-based and disaggregated architectures not just to enterprise or rural networks, but to the heart of their 5G macro RANs – networks that carry most of the responsibility for their business models, and that will face unprecedented demands in terms of capacity, quality of service and critical availability in the 5G era.
Whether they can really be both demanding, and so urgent in terms of timescales, remains to be seen. It appears very likely that, in real terms, they will stick to timeframes they have already, individually, discussed – to introduce Open RAN in small networks from 2022, and then to work with labs, trials and other efforts with their suppliers, to get it ready for the big time in the middle of the decade.
But in their introduction to their ‘Technical Priorities Document’, the five operators specifically state:” In terms of timeframe, the operators wish to ensure the readiness of Open RAN solutions for large scale network roll-out from 2022 onwards. Macro deployment is identified as the primary target for the operators.”
On the one hand, making a commitment to large-scale commercial roll-out next year does both boost confidence that the vendors – especially the new ones with limited ongoing revenues – will be able to move on from trials and tests to commercial deals from next year. And it certainly injects a sense of urgency, or perhaps panic, into the current efforts to turn specifications, from O-RAN Alliance and groups, into robust networks, that will also provide operators with a low-risk migration path to cloud-native RAN.
However, operators will be very aware of the risks of deploying anything half-baked in their main networks – even small teething troubles can alienate customers and spook other operators. It seems unrealistic that all the requirements in this document could be met to the level needed in a fully loaded 5G network within a year, but the operators do seem keen to change the current narrative, that Open RAN will be mainly a technology for small cells and secondary networks in its first few years. In so doing, they may also be making a case for European national and EU governments to offer more financial support to accelerate progress in Open RAN; and seeking to place a European industry, over which they would have significant influence, in the driving seat when it comes to macro Open RAN.
Without significant government and/or operator funding, it is unlikely that smaller suppliers could tick all the boxes in the document, and that could leave operators forced to work with established major vendors, and their chosen partners, rather than achieving the longed-for open, multivendor ecosystem.
Of course, as we have noted before, it depends on definitions. An operator may choose to deploy a few macro sites as a pilot network, or to enable just open fronthaul interfaces initially, rather than a full new containerized architecture. Those steps could be achieved in 2022 and count as the start of a macro network roll-out, even if full-scale deployment came more gradually and towards mid-decade.
However the roll-out pans out in the coming years, the document provides the most detailed and real world description yet of what an Open RAN might look like, if it were to support the MNOs’ core businesses. Since it has been agreed by five huge operators, it will also provide a strong dose of guidance and encouragement – any vendor working on the options set out here will have reasonable confidence that it could target business with five major operators.
That will not be an easy task though. The paper states that: “To become a competitive alternative to traditional RAN, the operators require solutions that will not compromise on network quality, security, high energy efficiency, as support for 4G and 5G …efficient RAN sharing and legacy band support.” The operators also want the flexibility to implement “various network topologies” including centralized and fully virtualized RAN, and to use either an edge cloud or regional cloud for the basebands, depending on the scenario.
And while the demands for a macro solution seem challenging, there is also a need for small cell and indoor solutions. These are often assumed to be the first networks in which Open RAN will be deployed commercially by non-greenfield operators, but the document suggests that these smaller roll-outs may not be a ‘training ground’ for macro deployments, but will instead have their own topologies, architectures and ecosystems.
“Small cell deployment for both outdoor and indoor scenarios is being targeted with different architectural splits and centralization options,” the paper says. “Indoor deployment does also include a multi-operator scenario with shared or dedicated infrastructure.”
This area of interest would naturally bring Small Cell Forum’s nFAPI interfaces, based on functional Split 6, into play, and there are already vendors and chip suppliers (including Qualcomm) developing solutions. But much of the focus of the document is on the O-RAN Alliance’s specifications, and so might be argued to be narrower than the multi-interface architectures being drawn up by some other operators such as the UK’s BT or the USA’s AT&T. But it goes way beyond foundational specs and highlights additional performance and technology requirements as well as some architectural preferences.
It does talk about the basic interfaces of course, and particularly highlights those that the five operators consider most essential, and the preferred priority for development work by suppliers. The primary interfaces are O-RAN Alliance’s Open F1, which links the centralized unit (CU) and distributed unit (DU) in a disaggregated RAN; and Open X2/Xn for connecting different base stations. The Open F1 fronthaul is, the document states, “the prime interface to be supported in a fully interoperable manner, without compromising network performance, especially for Massive MIMO”.
Though this is a clear endorsement for F1, which is based on the 3GPP Split 7.2x, at least in the macro network, there is still work to be done to make it fully functional. In particular, the operators want the industry to “further investigate uplink enhancements for the 7.2x split in order to improve performance and robustness particularly in mobility scenarios”.
In spectrum terms, the priorities reflect those of European operators, with far less focus on millimeter wave bands than US equivalents would assume. For these MNOs, the key bands on which to focus Open RAN 4G and 5G development are the midband
3.4-3.8 GHz spectrum as well as legacy FDD bands such as 2.1 GHz or 1.8 GHz.
One of the big obstacles to Open RAN in many European markets is that operators intend to keep 2G networks (and, less commonly, 3G) running for many years and so require interoperability with legacy systems. This is something that was highlighted in operator testimony to the UK government’s committee on supply chain diversity, in which BT stressed that Samsung would not be a suitable macro network supplier at present because it had no backwards compatibility with 2G/3G.
Some Open RAN suppliers, such as Parallel Wireless, do have 2G/3G products, but these would not necessarily interwork easily with installed legacy networks from other suppliers. The operators said they were “interested in interoperability between 2G/3G baseband units and radio units, based on proprietary interfaces, since no open interface has been specified successfully. This applies mainly to hybrid radio units supporting 2G/3G/4G/5G, but also for legacy 2G/3G-only RUs already deployed.”
That raises the interesting prospect of proprietary interfaces remaining in place between fronthaul gateways and legacy radios, that would “allow reuse of already deployed radio units as an interim solution”. This could be far more than a solution to a niche requirement – it would allow operators to hold onto their installed 2G/3G, and even 4G and 5G NSA radios for as long as performance and return on investment dictated – which might indicate a longer and more messy migration to Open RAN than the visionaries depict (but also, perhaps, a pragmatic approach that would offer an earlier entry point to the big operators for new vendors).
Another important aspect of the requirements is support for RAN sharing. In past generations, most MNOs were reluctant to share macro RANs, because of regulatory restrictions or because they regarded control of their networks as a key to differentiation. However, sharing the RAN and then differentiating with services (and in future, by accessing slices of the network) has become an increasingly popular strategy.
This will clearly play a role in the general push to drive down cost of ownership for 5G RANs as they get denser and as their performance requirements escalate. An open competitive supply chain should reduce costs by improving price competition, argue supporters – though most operators accept that, until more ‘plug-and-play’ Open RAN solutions emerge, the cost of deploying and integrating a multivendor network is likely to be higher than for an integrated solution.
Most of the medium term savings will be in opex, and will come from automation and the flexibility of cloud-based networks, but these benefits would accrue from any vRAN, not necessarily an open one. And many studies show that the impact on TCO of easier sharing of passive and active networks will dwarf the savings to be had, at least in the next 5-7 years, from automation and cloudification. Indeed, the biggest impact of moving the RAN to the cloud may be the increased simplicity of sharing, since each operator can deploy its own virtualized or cloud-native functions (xNFs) on common cloud infrastructure.
The document states: “While MORAN [multi-operator RAN] with shared O-RU (open radio unit) only, and MOCN [multi-operator core network] support is unanimously requested, both shared infra and dedicated infra per operator is relevant, depending on whether the infra is deployed on the same site or deployed autonomously by each operator in their target location (e.g. in their own cloud). Efficient RAN sharing management is required to allow sufficient independence between operators to manage their own xNFs functions on a shared infra, while avoiding any potential conflicts.”
Cloud-native architecture is also central to the document, even though these platforms are at the very early stage even for 5G core networks, let alone RAN. Rakuten Mobile has worked on a cloud-native, containerized RAN with partners such as Robin.io, but most established operators indicate, in Rethink’s recent surveys of MNOs, that they see a cloud-native macro RAN as a development for the late years of the 2020s. However, that is not stopping the five operators from being ambitious – for instance, they want a Kubernetes-based open cloud platform, capable of supporting the RAN software, to be developed and proven.
Another challenging requirement concerns energy efficiency. This is an important goal for many operators embarking on 5G, to meet cost reduction targets and green commitments, but the density of 5G offsets the energy efficiency of the new radio on a per-unit basis, while running intensive RAN functions on general purpose processors is, with current technology, far more power-hungry than using dedicated ASIC platforms. Even Nvidia, which is developing a GPU-based architecture to support the vRAN distributed unit, has acknowledged that its first efforts are extremely power-hungry when supporting advanced functions such as Massive MIMO beamforming or dynamic spectrum sharing.
The operators are steering the industry to prioritize development resources for this challenge, with the onus likely to fall mainly on the processor companies such as Intel, Nvidia and ARM. “Energy efficiency is a global requirement encompassing various domains of Open RAN,” the document states. “Hardware should be the most energy efficient as possible, starting with radio transmitters, but also considering RAN infrastructure, where energy efficiency will be a criterion in the choice of accelerators.” It also points to the requirement for network flexibility to save energy by turning idle units down or off. It adds: “Vendors should also implement features to reduce the energy consumption in both networks and devices in low load situations.”
Of course, security is high on the list. There have been heated debates over whether Open RAN is more secure than traditional networks (because so many eyes are on these platforms, and so many experts contributing to the development of security mechanisms); or less so, because of the uncontrolled nature of some open development. All agree that 5G potentially opens up more points of entry and vulnerabilities than previous networks because it will enable new Internet services including Internet of Things, which will involve millions of sometimes unsecured end points, that cannot necessarily be tracked and controlled as effectively as a smartphone.
“Security is a global priority for the operators requiring strict policies and features across all entities of the Open RAN network, which raises new challenges due to the multiplication of vendor solutions and open interfaces,” says the document.
All these requirements are a very useful handbook, for all stakeholders, of what operators need and what should be prioritized for development. But the pragmatic acknowledgement, implicit in the whole document, that Open RAN is very much a work in progress, seems at odds with the bullish timescale for macro deployments. The operators state that they “consider further standardization is required” but they do not plan to define any standards, and so are beholden to the timetables of the industry bodies, especially the O-RAN Alliance. “Standardization will be subject to the established standard-setting bodies such as the O-RAN Alliance,” they said clearly.
The paper will undoubtedly focus the minds of those bodies on trying to tick as many operator boxes as possible when prioritizing their work items. But it is to be hoped that, by making 2022 such as big milestone, the European group is not setting its own fledgling ecosystem up for failure.
Dedicated 5G coding can boost O-RAN performance and reduce cost, says analyst:
One of the challenges for emerging vendors in the Open RAN ecosystem is that they may lack the resources to develop optimized solutions to all the challenges set out by operators such as the group of five Europeans. Even if they have strong partnerships with larger vendors, and only need to focus on the particular network element or software they provide, there will still be demanding requirements to address in a short period of time, as the European technical white paper has made very clear.
But some aspects of the designs are not worth compromising on, argues network analyst firm Mobile Experts in a new report commissioned by AccelerComm, which specializes in solutions for spectral efficiency, advanced decoding and ultra-low latency.
The report claims that supporting the most advanced channel coding could save a “typical” US operator $193m when deploying O-RAN compliant networks. The report, entitled ‘The dollars are in the details’, argues that dedicated solutions for coding and decoding, though they come at more cost than generic offerings, can deliver better performance and lower cost for customers.
But the upfront cost can be off-putting to a new supplier trying to get a product to market quickly. Joe Madden of Mobile Experts said: “The major OEM vendors in the 5G market have invested a lot of resources into optimising their LDPC [low density parity coding] decoder implementations, but emerging O-RAN vendors just don’t have the resources to optimize each step in the processing chain. Many are using general-purpose LDPC coders and decoders, with generic architectures and algorithms derived from WiFi, Bluetooth or 3GPP networks, and therefore are not optimized for 5G environments.”
In a hypothetical US national network modelled by Mobile Experts, a dedicated solution tailored for 5G Open RAN was estimated to reduce the number of gNodeBs (5G base stations) required by 2,550, saving $193m (€159.77). The exercise compared an O-RAN deployment using standard LDPC decoder implementations with one using AccelerComm’s Advanced LDPC IP.
LDPC is a linear error correction scheme that helps to reduce noise and interference in 5G data transmissions.
Rob Barnes, VP of sales and marketing of AccelerComm, said: “While there are significant benefits to O-RAN technology, there are still areas where it lags behind the major vendors, in particular in the area of 5G NR uplink sensitivity.” This uplink capability was called out in the five European operators’ technical requirements paper as an area that needed special focus from the ecosystem, if O-RAN was to meet the needs of their 5G macro networks in future.