For a long time, the virtualized RAN (vRAN) has been something more of dream than substance. It has been discussed ever since China Mobile’s seminal white paper, a decade ago, launched the concept of a RAN whose digital functions were run as software in the cloud. But commercially, it has remained the preserve of a few very expensive, hand-crafted deployments, mainly in South Korea, Japan and China; and some roll-outs of small cell or secondary networks, where it has often proved less risky and disruptive to test the new technology without touching the primary macro network.
Now, several factors are conspiring to turn on the accelerator. The start of the gradual shift to full 5G, with a 5G sliceable core; the heightened interest in open RAN architectures, which are starting life in disaggregated, virtualized form; the catalyst factor of new entrant, digital-first operators like Rakuten Mobile; the glimpse of mature, deployable vRAN platforms based on standard interfaces to support interoperability and future-proofing. All these are making the vRAN seem like a realistic option for the coming few years, for many operators, rather than an exotic vision confined to a few pioneers.
5G has not, in itself, been a trigger to virtualize the RAN, since it has been deployed in Non-Standalone (NSA) mode with the 4G core. However, operators are starting to plan their routes to Standalone 5G. Since the 5G core will be virtualized – and increasingly operators are considering moving straight to the emerging modern and agile containerized architectures – there will be logic to starting on RAN virtualization at the same time. Some of the most disruptive and difficult aspects of migration to the cloud, including new supply chains, internal organizations and network orchestration processes, will apply to core and the RAN.
Most operators will attend to the core first, and the vRAN certainly brings the highest cost, risk and changes to old ways of any network domain. But as they start to expand their NSA networks, it will make sense to leverage the progress made towards virtualized or cloud-native platforms in all areas of the network.
So the big vendors – conscious that, if they do not move quickly to address operators’ new urgency about virtualization – are rolling out their next generation 5G offerings. Last week, Ericsson unveiled its 5G Standalone offering, which provides the option of a fully cloud-native core and virtualized RAN. The previous week, Nokia had launched its new vRAN portfolio, throwing its weight heavily behind the O-RAN open interfaces, as if to compensate for its troubles in the first generation of 5G deployments by setting the pace in the second.
Hard on its heels came Samsung, also an O-RAN aficionado, showing off its own new vRAN platform. The Korean vendor, like Japan’s NEC, is hoping that taking a lead in open RAN architectures and vRAN will give it a significant play in the global RAN market, rather than just in their respective home countries. With very little legacy installed base to worry about, these vendors nevertheless have deeper pockets and come with less risk for operators than the O-RAN specialists and start-ups.
There is still a long way to go however. There will be some high profile migrations to 5G SA, the cloud-native core and the vRAN in 2020 to 2022 – Verizon, China Mobile, Telstra and a handful of others are leading the charge this summer. There will also be more deployments of disruptive open platforms by new mobile players like Dish Network in the USA. But behind these headlines, the more common pattern will be to trial open, disaggregated vRANs away from the primary macro network – in enterprise, city, campus or IoT networks that can generate new revenues but do not put the quality of experience of the main user base at risk if there are teething troubles.
These secondary deployments can extend the edges of the mobile network into in-building and enterprise environments of many kinds. As Small Cell Forum’s newly published annual Market Status report indicates, many of them will be implemented and managed by non-MNOs, such as neutral hosts or private operators, with their own localized vRANs, vCores and edge computing nodes. This is where many operators, traditional and new, will learn how to run a virtualized RAN and core in anger.
Success in such deployments will increase confidence, and drive availability of commercially robust, easily deployable solutions, rather than the kind of engineering-heavy, highly customized platform that Rakuten has implemented. The pace of mainstream network cloudification will depend on how quickly certain key elements are standardized, or at least reach the point when there are just two or three architecture options, which can support equipment and software at scale.
This is partly related to the open RAN movements. There needs to be consensus rather than fragmentation, and the different open RAN options are too often discussed in terms of industry and global politics rather than technology. Rethink’s recent operator surveys indicate that most operators want flexibility about key interfaces – notably the fronthaul and midhaul connections between radio units, distributed units and centralized units in a disaggregated RAN.
They do not want one rigid specification – two or three options provide the ability to deploy the optimal solution for a particular environment, which may be different in a city from a factory, for instance. Many operators, old and new, say they will support more than one split option in their disaggregated networks, with Option 6 (supported by Small Cell Forum) being prominent in enterprise and private networks, and Option 7.x, backed by O-RAN, in macro and city deployments.
Ideally, operators like BT want a software-defined xHaul interface. And of course, what they do not want is an array of different options, all lacking a wide base of equipment and vendors to choose from; nor for any vendors to implement the most popular splits in a semi-proprietary way, creating lock-ins, as happened with the CPRI fronthaul spec.
Achieving consensus and broad support for a manageable set of interfaces is one important enabler of mass adoption of vRAN. The other is the appearance of trusted, commercially robust containerized solutions, for core and RAN, which operators really believe can perform as well as, if not better than, vertically integrated, physical solutions. In the RAN, there is still a significant question mark over this issue – whether the intense demands of this network domain, greatly increased by 5G, can possibly be met in an optimal way through software deployed on COTS hardware?
This will remain an issue for very low latency applications or those reliant on extreme levels of signal processing, such as dynamic spectrum sharing. But deploying the core and RAN in cloud-native fashion, with applications deployed in containers and microservices, will make the systems more lightweight and agile, and will help achieve some of the key goals of 5G, such as dynamic slicing, without the need for highly specialized hardware. Despite many vendors’ claims to a fully cloud-native core, it is only now that most operators have thought solutions might be deployable any time soon, as seen in a crop of major trials or early deployments, such as Verizon’s and Telstra’s, both working with Ericsson.