It is clear that virtualization will be a significant element of the 5G network picture but there is a serious risk of disconnect between standards work in these two key areas. Carrier network virtualization is increasingly focused on ETSI’s NFV (Network Function Virtualization) specifications, though there are many other relevant activities too, many in the open source arena. This technology is starting to impact on mobile operators’ plans at the 4G stage but is likely to be refined further for 5G platforms – but of course, work on the core air interfaces for 5G is only just beginning, despite the welter of ‘pre-5G’ demonstrations and launches.
The latest initiative funded by the huge European Commission 5G PPP (Public-Private Partnership) scheme is designed to ensure these two areas coalesce at the right points, and that NFV is integrated into 5G at an early stage to avoid reinventing the wheel. Called the ‘Superfluidity Project’, it will be run by a consortium of partners from the telecoms and IT businesses, which are increasingly being forced to work together as a result of trends to virtualization, IT/network convergence and 5G. This, of courses, raises issues all along the chain, from operational levels (skills and company organization) to standards bodies. Standards projects in IT, internet and telecoms will increasingly need to take account of one another, creating a new layer of complexity and inter-communication.
There are 18 players attempting this act of convergence in Superfluidity – the big name operators are BT and Telefonica, while Alcatel-Lucent and NEC lead the telecoms contingent and Citrix, Intel and Red Hat join from the IT world, along with smaller companies and academic institutions. The aim is to develop a converged, cloud-based platform for virtual network functions and services.
The core goal of “superfluidity” in the network will be achieved, the group says, by defining and abstracting all current network components (and future ones as they emerge) and “decomposing” them, as Saverio Niccolini, deputy general manager of NEC Laboratories Europe, put it in an interview with TelecomTV. That means breaking them down in to basic, reusable units of software (virtual machines) which can be inserted anywhere in cloud-based architecture to support 5G services as these, in turn, are defined.
Those services can be overlaid on the fluid and infinitely reconfigurable software network created by all these virtual machines, creating advantages for the operator which include, the consortium argues:
- Location independence of network services in heterogeneous networks
- Time independence through instant and agile deployment and migration of services
- Scale independence – almost unlimited and transparent service scalability
- Hardware independence
Operators will be able to launch services instantly and run them anywhere on a network, from the core to the edge, (and remove them again), addressing many issues of how to provision services in an agile way, and also avoid over-provisioning which is inevitable when current rigid architectures operate amid varying demand levels.
If the group achieves its aims, it will be an important step in the process which the NFV initiative so convincingly started – translating virtualization, an approach created for the IT world, into telco-speak and making it suitable for the intensive and specialized demands of a carrier network. ETSI’s work certainly brought virtualization, which was hardly discussed by operators in a commercial context beforehand, into the carrier’s world, as well as providing a much-needed point of unity, but it remains quite “data center-oriented”, in Niccolini’s view.
One of Superfluidity’s aims will be to take more account of the mobile edge – a critical element of 5G thinking is to move a large number of functions to the very edge of the network, and to push that edge further out, even as far as the device or sensor gateway. So getting the decomposed elements out to the edge wherever possible, to reduce latency, will be critical.
Niccolini said: “You need smaller and smaller virtual machines so we can push functions out to the edge of the network” and effectively extend the reach of the data center. But in that environment, it is increasingly vital to have strong umbrella layers to manage and orchestrate these huge numbers of dynamic and distributed VMs.
Andy Reid, chief network services architect at BT, told a recent LightReading OSS conference in London that one of the important aspects of Superfluidity will be to take the basic ETSI specifications a step further. The problem with drawing up specs in such an immature market is that requirements and technologies change rapidly, and Reid already feels that the ETSI approach to MANO (management and orchestration) is behind the times, failing to take account of issues that have arisen since it did its first work.
He said “I wish I’d pushed harder” for ETSI to take a more holistic view of the NFV processes when it drew up its initial specs, a process in which BT was a leading light. “Shouldn’t we take an end-to-end process perspective and understand how they’ll work logically and then map that back to the architecture?” Instead, he told the event, the result was an API specification that is valuable but inadequate to deliver the “interoperability and the end-to-end systems we want. These processes are not going to join up just with what we’ve got.”
Yet orchestration will be vital to a network that needs to be highly automated as well as highly fluid. There is a “lack of precision” about the responsibilities of the various functional blocks in NFV MANO, he says. For instance, “if you start including FGPAs (field programmable gate arrays) there is a whole set of capabilities that become more programmable and that creates a new optimization problem.”
Nik Willetts, chief digital officer at the TMForum, also argues that the NFV discussion has been too narrow, and too focused just on transitioning from physical to virtual, not on the “colossal challenges of managing a virtualized network alongside physical systems”, as he told this year’s annual TMForum conference.
“While NFV is a key piece of the agility puzzle for service providers, you can’t look at it in isolation,” he said. “NFV creates the opportunity to revolutionize how we create, deliver and manage services, but, in order to succeed, they will need to overhaul their operations and business support systems, as well as how they think about operations.” The Forum’s own ZOOM project aims to “create best practices and frameworks to guide this transformation, complementing the work of bodies such as ETSI”.
BT has also been critical of other standards which could underpin virtual carrier networks and NFV, including the OpenStack open source platform, highlighting the immaturity of this space and the huge amount of work that standards groups need to accomplish in a short time to meet carriers’ shrinking deadlines.
Recently, the telco’s chief data networks researcher, Peter Willis, told a conference in Germany that the firm was considering proprietary alternatives to OpenStack – which it has previously supported – for its virtual enterprise CPE roll-out, unless a set of issues was addressed. These relate to the connection of VNFs, service chain modification, scalability, ‘start-up storms’, security and backwards compatibility. “If these six issues are not addressed we will not use OpenStack for virtual enterprise,” Willis said.
The Superfluidity project, which is receiving almost €8m ($8.7m) in EU funding under the Horizon 2020 innovation program, will run until the end of 2017 and so should dovetail with the 3GPP’s work on 5G. This recently started in the form of a work item, and other relevant activities in other bodies. Of course, the next challenge will be to communicate with these other initiatives to ensure that respective projects really do interface with one another while they are in the works, rather than jostling for the upper hand later.
In fact, ETSI is rising to this challenge too, setting up an inter-standards development organization focused on information models to describe infrastructure and virtual network functions requirements. This will look to pull together efforts by various standards bodies including the Metro Ethernet Forum and the Open Networking Foundation, something ETSI is well positioned to do as it is not a standards body itself, and so can sit above their politics to some extent.