The combination of virtualisation with small cells is viewed as a critical aspect of the continued development and improvement of mobile networks by operators seeking a radical overhaul of architecture and cost-efficiency. Alan Law of the Small Cell Forum discusses the work the organisation is doing to drive the adoption of this technology throughout the industry.
Virtualisation has been the surprise game-changer of the past two years, stealing up behind 5G to offer a more immediate prospect of radical change in mobile network architecture and economics. As mobile operators start to weigh the advantages of this approach in their access networks, they are often looking to introduce virtualisation in tandem with small cells. That has made it important for the Small Cell Forum (SCF) to take a leadership role in driving the technology forward.
Increasingly based on (European Telecommunications Standards Institute) ETSI's network functions virtualisation (NFV) specifications, the technology runs key network elements in software as virtual machines on commodity hardware, rather than on dedicated hardware appliances. The technique is well established in the data centre but immature in the carrier network, and most early trials and deployment plans are focused on non-RAN areas such as the evolved packet core.
Why small cells?
Virtualising RAN, as envisaged by a few visionary operators like China Mobile, has challenging issues, not least that most carriers are currently investing heavily in conventional LTE macro networks. In addition, there are questions over whether standard servers can ever support the hugely demanding network tasks currently carried out on dedicated digital signal processors in the base stations. Of course, RAN also cannot be entirely virtualised - at a minimum, a stripped-down radio/antenna unit must be left in physical form at the cell site. This must, in turn, be linked to the baseband processor - which may be in the cloud - by a very fast, ultra-low-latency connection, called fronthaul, that demands ready availability of affordable dark fibre in most cases.
These challenges mean that most operators will not embark on macro RAN virtualisation until the next decade. However, as they come under pressure to densify their networks, the deployment of a new layer of small cells will often be a trigger to introduce a first measure of access network NFV.
This is for several reasons. Firstly, as operators embark on densification programmes, the small-cell layer can be deployed from scratch and so a new architecture is possible without having to risk disruption to the established LTE macro network - which will need to perform several more years of useful service before it would be justifiable to replace it again.
Secondly, the deployment challenges are somewhat fewer. Small cells do not have the same intensive performance demands as large base stations and will often be fronthauled to a virtualised small-cell controller that is relatively nearby, and so can be linked by low-latency wireless technology if fibre is impractical.
Thirdly, virtualisation may improve the economics of a new small-cell roll-out far more quickly than those of the macro RAN, because very low-cost cell-site equipment can be combined with a virtualised small-cell controller.
Finally, the main justification for large-scale small-cell projects is to enable the 'HetNet' heterogeneous network, and a virtualised approach, combined with self-optimising network (SON) technology, can make it far more viable to integrate different cellular and Wi-Fi platforms into a seamless pool of capacity.
An operator survey conducted by Rethink Technology Research in June 2014 found a clear overlap between small-cell densification and vRAN in carriers' plans for next-generation network architectures. With about 50% of MNOs planning some elements of vRAN or cloud-RAN in their five-year plans for LTE, 90% of those also had urban small cells on their road map. More than a third of those respondents said vRAN activities would be specifically motivated by the need for densification.
In a subsequent survey a year later, more than 55 operators were asked to name their key reasons to kick off vRAN projects in the small-cell layer (see figure, left). The top five reasons related to resource and cost-efficiency, and the need for a dense network that could scale up and down easily according to demand.
All of these could potentially be achieved using a virtualised approach but, for the confidence that leads to mass adoption, most operators would need to see unified platforms and interfaces that would reduce their cost and risk.
The SCF's role
This is where the SCF comes in. It has put virtualisation at the heart of its next phase of work, announcing a work programme in mid-2014 and publishing the latest installment in its Release Programme, Release 5.1, a year later, focused on virtualised small cells (see 'The Small Cell Forum and virtualisation', opposite. This provides a collection of technical guidelines, best practice advice and other resources to ease the operator's path to vRAN.
In September 2015, it announced a new work programme focused on the next generation of small-cell deployment and based on six topics of particular interest to operators. One of these is virtualisation.
The SCF has taken the lead for two key reasons. One, in an immature sector, it is critical to avoid fragmentation, which is a barrier to deployment and vendor investment. Two, while there are many technical debates about the optimal virtualised architectures, the only successful choices will be those that relate directly to operators' business cases and commercial requirements.
The SCF provides a valuable hub of activity in both regards. It has experience of developing interoperable platforms and open interfaces that provide a unified base on which operators and vendors can differentiate without fragmentation. These developments have always been tied tightly to the real-world requirements, as fed into the SCF's work by its operator members.
The SCF has found clear benefits and drivers towards centralisation and virtualisation of the small-cell network. These facilitate the scalability of small cells, and enable functions to be moved around depending on loading conditions or availability of compute and transport resources.
The role of the SCF in driving common platforms will be more important than ever in virtualisation because of the wide range of approaches that could be taken and the risk of wasting the potential of this transformation by failing to harmonise the ecosystem.
At the heart of the SCF's virtualisation effort is the network functional application platform interface (nFAPI), which will define the optimal interface between the radio head, or physical network function, where most of the latency-specific RF activity remains, and the virtualised small-cell controller, in which most of the intelligence is concentrated and that may be in the cloud or more localised.
The issue of how the functions should be split between physical and virtual is fundamental to the economics and workability of NFV-based small cells, but there is no simple answer, and the nFAPI team is evaluating several possible splits. The SCF workgroup is examining virtualisation models that deliver key benefits over a baseline distributed-RAN approach, taking account of important capabilities like dynamic operations - allowing the functions to be scaled up or down as required.
Real-world choices will depend on the individual operator's objectives and resources, particularly access to dark fibre. This is, for most operators, the most daunting challenge in considering vRAN, and the SCF's nFAPI work specifically enables approaches that allow 'non-ideal' backhaul and fronthaul - those based on IP and Ethernet technologies - to be used effectively. With the envisaged nFAPI splits, the physical unit can be linked to the virtualised small-cell controller using the same IP connections as small-cell backhaul, though with some decrease in latency tolerance.
As well as nFAPI, the SCF will be engaging in other activities around virtualisation, including 'plugfests' for vendors to test their products for interoperability; proof-of-concept demonstrations of various architecture splits, including cooperation with ETSI; and work to identify management and orchestration (MANO) requirements.
It is a key tenet of the SCF that all its work must be grounded in business models, not just technology. Most operators are looking at virtualisation initially from the standpoint of saving on cost, and on other resources like energy, by shifting from specialised hardware to software on general-purpose servers; however, in the operator survey, only 24% said those efficiencies would be a sufficient justification on their own for moving to NFV in the access network.
In addition to those benefits, they were looking for an architecture that would be so flexible and scalable that it could support entirely new revenue streams or use cases. This may mean combining NFV with software-defined networking, and these two related concepts are becoming entwined with a third - mobile-edge computing (MEC). This is also closely related to the SCF's work on small-cell-service-enabling architectures and, like NFV, is being driven by ETSI. MEC posits a cloud-based IT services environment at the edge of the mobile network, which provides benefits such as ultra-low latency, precise positional awareness and agile applications.
This combination of technology allows resources to be allocated where required; new instantiations of the VMs to be called up, or removed, according to need; and the software to be customised for different customers and partners, because all the elements are programmable. This provides several business opportunities to the operator so that it can, for instance, dedicate certain resources to a quality-reliant service such as mission-critical enterprise connectivity, launch and remove new consumer applications rapidly in response to uptake, or allow enterprise customers to access their own secure 'slice' of the network capacity.
The SCF is particularly studying the potential for virtualised networks that support multiple operators, as well as the emerging 'network-as-a-service' (NWaaS) model. In NWaaS, a mobile operator, virtual operator, OTT service provider or enterprise can access - and pay for - capacity as it needs it, selecting and combining different elements, such as security and signalling, from a service catalogue, and receiving its own unique instance of the network that is logically separated from other traffic.
These are the kind of business cases that are impossible to scale up on conventional networks and that may well be an important element of 5G - along with virtualisation itself. By taking a leadership role in this important development at the earliest stage, the SCF is confident that it will also be in a position to influence those 5G evolutions, giving its members a strong voice in how the next generation of mobile networks will look.
SDN and NFV
Virtualisation implements some or all network functions as virtual machines running on off-the-shelf hardware, saving on the cost of ownership of a string of dedicated one-function boxes. It also delivers an integrated orchestration environment so those virtual machines (VMs) can be reconfigured on the fly without affecting the whole system, adding to flexibility and allowing new instances of VMs to be deployed or taken down rapidly.
Software-defined networking (SDN) can be implemented entirely separately but has specific benefits for virtualisation, such as the ability to scale the control-plane aspects of the network independently from the data-plane aspects. SDN separates the control and data planes, enabling centralising all the control of the huge number of distributed VMs supporting data-plane operations.
Among mobile operators, virtualisation is widely seen as a nearer-term strategy, at least for certain parts of the network such
as the packet core, than full-blown SDN. However, the general consensus is that the full benefits of virtualisation will only be achieved as part of an SDN transition, even if that takes ten years to achieve.
The ETSI set of specifications for NFV has been an important factor behind operators' unexpectedly intense interest in near-term deployments - giving them the confidence of a widely supported set of specs that have been carrier-driven and tailored for telcos' specific needs.