Network virtualisation has moved from the periphery to the centre of operators’ strategic plans as they grapple with a range of operational and competitive challenges – from keeping costs under control to becoming more agile. If established players fail to adapt there are plenty of challengers emerging who will be only too happy to show the way, so the industry has braced itself for change.
Clear signs of this include a willingness to work together, and to establish some open standards that will smooth the transition – initiatives which are already starting to converge.
“We’ve had a raft of announcements recently,” says Ken Dilbeck, TM Forum’s virtualisation expert. These include the announced merger of two big open source network functions virtualisation (NFV) orchestration and management initiatives: the AT&T-led open source ECOMP and China Mobile-led Open Orchestrator (OPEN-O) projects. Some expect them to be joined by Open Source MANO, another open source effort, led by Telefónica and sponsored by ETSI.
All of these initiatives, and their open source emphasis, are designed to accelerate virtualisation by addressing the need for end-to-end OSS/BSS (management and monetisation). “There’s an understanding among service providers that they need to join up the touch points, which needs an industry solution, not an AT&T- or Orange-specific solution,” Dilbeck says.
AT&T is so serious about virtualisation that it plans to replace 75% of its hardware with computer-operated software systems by 2020. To date it claims to have converted 34% of its network to a software-defined model; its goal is to get this to 55 per cent by the end of 2017.
But it can’t control everything. “Jointly creating software code that is free for anyone to use represents a fundamental shift for an industry built on proprietary solutions and vendors’ IP,” adds Alla Goldner, director of industrial affairs and standardisation at Amdocs. “It will be easier for SPs to combine best-of-breed technology from different sources in their networks.”
Operators have already come to rely quite heavily on external cloud providers for IT capacity, which has given them a greater appreciation for the potential of virtualisation.
Colt’s strategic shift to virtualisation started internally. Today, the reinvented ICT provider sees itself competing with the likes of BT Global Services, AT&T, Verizon and Deutsche Telekom, providing secure, high-quality, low-latency connectivity services to the international enterprise market.
In Colt’s case, the focus is very specifically high-bandwidth connectivity enabled by software-defined networking (SDN). Its SDN-enabled platform allows customers to set up and modify network connectivity in real time and pay for it by the hour - a network service consumption-based model which it claims puts the customer in control. In 18 months, it has launched four services based on this model: DCNet On Demand; Ethernet On Demand; Dedicated Cloud Access On Demand; and SD-WAN.
Colt’s virtualisation journey started in 2012, but originally this was to improve internal efficiency. “In the last couple of years, our focus has shifted to customer propositions,” says Mirko Voltolini, Colt’s VP of network technology. The agile, on-demand services are a significant departure from the two-to-three year contract/fixed-price models the industry is used to.
Although it’s still early days, Voltolini says at least half of Colt’s development effort is concentrated on new technology and services. Customer take-up of virtualised services is currently under 25% but Colt’s offerings are new and growing interest in services such as SD-WAN is expected to drive rapid growth.
Colt is playing an active role in the major open standards initiatives. Late last year, working with AT&T, it completed the first successful SDN interoperability trial, demonstrating that SDN architectures from different network service providers can interoperate with each other across continents.
What’s needed now is continued consolidation of standardisation efforts, and greater coordination and governance, according to Manish Gulyani, VP of marketing for IP and optical networks at Nokia. He believes vetting by a wider community, where operators have significant power, will lead to better solutions and better outcomes for everyone. As an example he cites the re-thinking of the packet core as a cloud-native proposition – leading to cloud-native product architectures that are best suited for NFV and also for evolution towards 5G.
Nokia sees particular potential in virtualising specific IP functions such as route reflectors; network function virtualisation in data centres; virtualised provider edge functionality (for improved enterprise connectivity and agile expansion into new markets); virtualisation in mobile (driving the need for other virtualised IP functions such as security gateways, WLAN gateways, etc); and virtualisation in fixed networks.
Beyond “slicing and stitching” network resources to enable new services, Gulyani says Nokia is seeing strong demand for network-sourced knowledge (ie. associated big data and intelligence) which could drive operational efficiencies using automation.
SDN can be taught to respond to changes in network behaviour – useful for honing network performance and responding quickly to cyber attacks. “Automation enables carriers to mitigate security threats in real time without manual intervention,” explains Louis Scialabba, director of carrier solutions marketing at Radware. “By analysing a rich set of data, carriers can be more predictive about potential attacks too, and stop the most complex network assaults as (or even before) they happen.”
Virtualisation of the network function also enables SPs to be more proactive in averting network issues and failures, via ‘active probing’. “This is an emerging trend whereby a virtualised operator can deploy virtual testing tools in the data centre that actively test for potentially faulty network nodes,” explains Carsten Brinkschulte, CEO of mobile core network startup CND.
The virtualisation opportunity is potentially so broad that operators need to be clear about what they aiming for and why, according to David Noguer Bau, of Juniper. Where existing services run in physical infrastructure, automation is likely to be the way to go – using APIs to ensure updates are still available across customer digital interfaces. In some cases, existing services will migrate into virtualised network platforms for efficiency and scale.
“Digital native services are easier to automate end to end, and more malleable for the creation of new infrastructure,” he notes. “For example, virtual managed security doesn’t require a specific infrastructure. As long as x86 compute nodes are available, new virtual firewalls can be spun up, configured, and managed.” Operators also need to be realistic about their ambitions and timescales, as this is a significant change for many and the culture/business model implications are considerable.
“Overall, the migration to programmable network infrastructure has been slower than predicted,” says Gerry Donohoe, technical director at Openet.“To date two uses cases are leading the way: SDN applied to WAN connections to connect enterprise networks, and virtual customer premises equipment to replace physical appliances at customer premises. These are the low hanging fruit with a straightforward business case,” he notes.
“Telcos are striving to deliver digital services with the same agility, innovation, speed and personalisation as web-scale companies and, while the adoption of cloud technology for network infrastructure can level the playing field to compete, deploying SDN/NFV based infrastructure is complex and there will need to be a period of transition to replace existing infrastructure.”
But transformation does need to get underway. “SPs must strike a balance between minimising expenses and maximising market power,” says Tsutomu Kushima, senior manager of the telecom carrier business planning division at NEC, whose virtualisation customers include Globe Telecom and NTT DoCoMo. “Network and service virtualisation are now business imperatives.”
As the IoT matures, the need for dynamism and reliable automation will only increase, notes Bejoy Pankajakshan, senior VP of technology and strategy at Mavenir, a provider of voice over LTE, voice over WiFi, advanced messaging and cloud RAN solutions. “5G is on the horizon, promising fast, lower-latency networks. An optimised core network solution, provided through the implementation of NFV, is exactly what will help SPs reduce total cost of ownership and be ready for the next evolution of wireless to drive new incremental service revenue.”
Visibility across networks is critical too, becoming more complex as elements become virtualised and the handing off points between providers blur. Open interoperability will help, but having the right tools will be important. “As networks and services virtualise, the ability to offer SLAs has sometimes been set aside in favour of the flexibility that more dynamic services offer,” warns Scott Sumner, director of business analytics at network performance and service assurance specialist Accedian. “For example, bandwidth on demand is rarely offered with a throughput assurance-based SLA, since it’s nearly impossible to validate the bandwidth is available as requested – at least, not using a traditional, standards-based service activation test.”