The first discussions about bring virtualization and SDN (software-defined networking) to carrier networks revolved around the cost savings. Implementing key network functions as software applications (often open source) on commodity hardware, rather than investing in expensive, proprietary appliances, would be the way to accommodate exploding levels of data usage without breaking the bank.
This was the strategy of desperation. According to John Donovan, AT&T’s senior EVP for technology and operations, the carrier’s global network experienced 100,000% growth in wireless traffic between 2007 and 2014, and such curves could not be profitably supported by adding new hardware. Instead, AT&T aims to virtualise 75% of its network, stripping down each component to its core and re-architecting it for the cloud, as he put it.
But cost alone does not make the case for this massively disruptive and difficult migration – not only to new platforms, but to new skills bases, suppliers and structures (notably the convergence of IT and telecoms).
Operators need additional justifications to make that long transition, and many of are finding it in the NWaaS (networks as a service) acronym. Broadly, this means enabling customers – which may be service providers, enterprises or consumers – to access bandwidth on-demand, pay for what they use, and automatically define and provision the services they want. If these customers are service providers (over-the-top, MSO etc) they will, in turn, support a large number of end users (people or connected ‘things’), and will want to ensure their security and quality of experience, while receiving detailed data for analytics purposes.
AT&T is one of the most advanced operators in moving towards new business models based on NWaaS. It has already introduced SDN-enabled Network On Demand, which allows customers to dial bandwidth up and down as required in real time. “That means they can use just what they need when they need it,” said Donovan.
At the heart of this dynamic, on-demand, self-provisioning world are two key technologies, based around NFV and SDN. One is lifecycle orchestration (LSO), an area where the Metro Ethernet Forum is seeking to set standards with its Third Network initiative; and the other is the service-oriented mobile core, which is being addressed by, among other efforts, a 3GPP task called Decor (TR 23.707).
LSO, as its name suggests, covers the whole lifecycle of a carrier’s interaction with a customer, from provisioning the order to delivering the service, with all the required configurations, to conforming to SLA requirements, addressing faults and providing analytics. LSO sits at the intersection between SDN/NFV and the established disciplines of OSS/BSS – and indeed, many carriers have said that a move to SDN will only really be justifiable when the OSS/BSS is not just virtualized but fully integrated into a dynamic, on-demand environment. Considering many operators took several years and millions of dollars just to convert the OSS/BSS systems to converge prepaid and postpaid billing and customer care, the omens are not good.
But eager vendors claim LSO will be the answer, and will enable operators to unlock many new revenue streams, with the flexibility to support new types of traffic and services, and therefore customers. According to a research report by Rayno, published in March LSO will be a $2.75bn market by 2019. The MEF, announcing its Third Network initiative in September, summed up the key functions of LSO as “Fulfillment, Control, Performance, Assurance, Usage and Analytics”.
If all these functions are fully integrated, operators can truly achieve the holistic customer view they have been chasing by bringing their IT (BSS) and network (OSS and packet core) functions closer together. That will enable them to support complex requests from enterprise or MVNO clients, which involve multiple cloud and telco providers, with detailed real time analytics and pricing, as well as SLA accountability, security and compliance. To deliver this involves tapping into many previously discrete carrier systems. And most importantly of all, carriers need to be able to launch and deliver new services and connections dynamically and very quickly.
The MEF’s Third Network aims to address these challenges, providing a middle way between the uncontrolled dynamism of the pure internet and the reliable but slow-moving closed network. The initiative builds on the MEF’s Carrier Ethernet 2.0 specifications in areas like service expansion, interconnection and manageability. It also includes service orchestration functions, APIs, and a NWaaS information model.
Naturally, the OSS/BSS suppliers are racing to virtualise their offerings and to embrace LSO. But they will face rivals from adjacent spaces too. For instance, CENX has been best known for running an Ethernet exchange, but last year it announced its Cortx Service Orchestration platform, which is already being used by two operators in mobile backhaul networks and by another for end-to-end service orchestration of SDN and NFV.
And then there are the IT and network giants, all keen to offset the blow to their traditional hardware revenues by taking a lead on SDN platforms. Cisco, NEC, the big four mobile network suppliers, Hewlett-Packard and IBM are all heavily focused on orchestration. IBM has placed its Dynamic Lifecycle Orchestration offering at the center of its NFV platform, which in turn will be at the heart of a new telco business unit. The company will launch its full NFV solution later this quarter, but has already published a white paper, ‘IBM OSS for cloud networking’, with heavy emphasis on orchestration, open source cloud assets, and of course, analytics (IBM’s key differentiator these days).
Brian Naughton, director of global telco sales at IBM Software, said recently: “IBM is investing in making NFV/SDN operationally viable rather than an academic project in ETSI, and IBM is not a [networks] supplier so we are going to help the operators realize their open vision of NFV/SDN.”
Addressing some of the same challenges is the move to a virtualized evolved packet core (EPC). This is one of the first network elements which most mobile operators plan to virtualize, and according to a recent Heavy Reading survey, 70% of MNOs expect to deploy one within three years. However, as with NFV in general, the virtual EPC will deliver optimal benefits only if it is accompanied by new business models, based around personalized and dynamic provisioning.
This has given rise to the concept of the service-oriented core network. This creates a ‘network slice’ for each application or user, configured in software and optimized for that particular type or service. This slice, a full virtual packet core instance, can support all the key LTE features, like prioritization, security and QoS, but also be tailored to the traffic profiles of each service. M2M, public safety, MVNOs and private enterprise networks are the most obvious examples of sectors which can be lured by a secure virtual platform optimized for their very particular needs.
Carriers also see this as a relatively low risk approach to NFV since the software-defined core is deployed in parallel with the main EPC (which need not be touched until a later stage), while the ‘slices’ bring new revenue potential with them from day one. However, the capabilities for separating different types of traffic within 3GPP standards are not ideal (PLMN or public land mobile network identities are too coarse, for instance). This is the focus of the 3GPP’s Architecture Enhancements for Dedicated Core Networks (Decor) activity, which is likely to be included in Release 13 standards, specifying how users connect to dedicated cores without using PLMN-IDs, and without handset changes.
An important use of network slicing is to separate customers with different QoS levels. As Ericsson explains in a recent white paper[1], once selection mechanisms are in place (DNS for the packet gateway, in future Decor for the MME, PLMN-ID if required), user terminals can be directed to the appropriate slice. “By first re-attaching terminals that have a low monthly subscription fee or are prepaid to the new network slice, the risk to subscribers that pay for and require a better service is decreased,” wrote the company.
The move towards self-provisioning, highly tailored services and an application-centric approach is radical for operators, but will only serve to save them from annihilation, not to put them way ahead of the field. This is the view of Leonard Sheahan, senior director at Oracle, who told a recent conference: “Amazon is very, very close to having a globally competitive network that can compete with telcos… very, very close.” And of course, Google, Facebook and even Cisco are in the same position, in fixed line broadband and cloud platforms if not wireless (yet).
[1] Ericsson, ‘NFV and software management’, December 2014