NFV: Virtually ready?
21 May 2015 |
With NFV now moving out of its early experimental phase, what has the industry learned so far? Capacity asks what conclusions can be drawn from early NFV proofs of concepts.
In its embracing of network functions virtualisation (NFV), the telecoms industry has demonstrated an unprecedented sense of purpose and energy. Just a little over two years ago there was no such thing as NFV, just a few operators privately toying with the uncoupling of the hardware and software elements of their network.
The starting gun was fired in November 2012 when seven major carrier names selected standards body ETSI to be the driving force behind the industry specification Group for NFV. Since then, across the telecoms sector and on every continent, carriers and service providers have been busy testing every aspect of the physical network, experimenting to see what can be run as a virtual function. Vendors for their part are turning their back on decades of hardware-led thinking and converting their existing product portfolio to meet the demands of this new environment.
The impetus and relative unity of vision that the industry has shown so far is nothing short of stunning. But the journey is only just underway. As early NFV and SDN trials and proofs of concept (PoCs) start to give way to real deployments, this is a good time to take stock and examine what has been learned so far.
The early NFV pace setters – names like AT&T, NTT Com, Deutsche Telekom, Telefónica – have long left tentative experimentation behind. NTT Com has gone further than most in bringing NFV-based commercial services to market. It is a year since it boldly launched a cloud-based o er that enterprises can activate themselves and pay for on a per-use basis. Built on a platform developed by Virtela, a company it acquired only in January 2014, the services are independent of any single vendor’s vision and available right across the carrier’s global footprint.
“We launched services in the summer of 2014, in 50 locations worldwide covering all continents,” says Ivano Rondelli, director of network services at NTT Com.
“Some were security-related – rewalls, IDS, VPN services. Since then I’d say we have faced internal issues, some relating to the education of our salesforce. Things are picking up slowly. We can now provide functions to customers that would once have been in a box.”
Initial difficulty has been worth it, says Rondelli, with the complexities of the pre-NFV model already giving way to something altogether easier to manage. Even in these early phases, cost benefits are also being noted: “In some areas we’ve calculated a saving of 60% compared to the previous solution,” he claims.
“We’ve got faster and more agile infrastructure. Customers can increase the resources they need from on as they want, not wait for months.”
He believes NTT Com has done ‘a pretty good job’ so far: “But we’re still working on it. We’ll now expand NFV capabilities so we are delivering all services that way.”
He says the plan from the word go was to be an NFV pioneer, not a follower. Rondelli claims NTT Com was the first operator to launch NFV services on global basis – whereas a number of operators have focussed primarily on their home territory.
Sticking with a vendor neutral platform, he says, has been key to its global focus, as the operator has been able to take advantage of any software it likes.
“It is also noteworthy that AT&T made the decision to change absolutely everything, while our strategy was to gain the bene ts without destroying all we’ve done in the past,” he adds.
Colt is another carrier which embarked on NFV before NFV had a name, starting on the virtualisation of its network back in 2011. The transition from physical to virtual has been tricky at times, admits Nicolas Fischbach, director of strategy, architecture and innovation with Colt: “NFV requires significant transformation, and carriers cannot realistically make it work with all legacy networks, nor should they – a demarcation is necessary,” he says.
“Hybrid solutions enable new NFV services to interoperate with existing environments, allowing for a gradual introduction and transformation.”
Some aspects of NFV are mandatory requirements from day one, believes Fischbach, naming in particular proper integration with the OSS stack and e ective orchestration: "The virtualisationof network functions needs to deliver the required network functionality at the required performance, which has many network and compute implications, otherwise large-scale NFV won’t become reality,” he concludes.
“If by virtualising existing services carriers aren’t able to o er the same level of performance and functionality, the NFV model won’t work.”
Colt has already learned a great deal from its SDN and NFV deployments. In particular it spotted early the challenges that lay ahead with OSS, service assurance, performance and customer adoption.
“That is why early on, in November 2012, we included a virtual CPE option within our services portfolio. This allowed our customers to try out virtual functions, and realise the benefits of having standard network access within their premises rather than multiple hardware elements in their buildings with limited flexibility,” says Fischbach.
The Model T Ford of networks
Vendors naturally are exuberant about the potential of NFV. The technology has the potential to be the Model T Ford of networks, says Rob Morrison, director product management, CSG International: “By decoupling from the physical element of networking, the potential economies of scale are transformed,” he claims. “The part we are interested in is virtual BSS. It’s a way to reduce both capex and opex, once you no longer have to procure purpose-built BSS hardware. You get accelerated time to market, better flexibility, you can scale up and down at will, and you eliminate over-provisioning. Capex can be reduced now, and opex too will come down too once you have a uniformity of hardware. It’s a bit like a budget airline that saves maintenance costs by only buying one type of plane. Some benefits are coming quicker than was expected.”
Despite early successes, and the efforts of both carriers and the vendors they have chosen as partners, nobody is pretending that NFV today is the finished article. The technology is still quite immature and needs more work in order to achieve full operational readiness. Solutions in the marketplace are no more than preliminary releases, with limited functionality. NFV architectures and deployments are complex, and the key building blocks to make them work don’t overlay properly yet.
The total cost of ownership of NFV versus traditional networking isn’t well understood. Barry Hill is global head of NFV at software vendor Oracle, which has been working with Vodafone Germany to virtualise the latter’s IMS network for the support of new VoLTE, VoWi and Message+ services. He calls these painful early stages that the industry is experiencing NFV 1.0: “The next challenge is about management and orchestration,” he believes. “How do you plug it all together, when you used to do the job with wires and cables? is is the way to greater e ciencies and more automation that in the past. There’s not much by way of management and orchestration in production yet though. When we get there we can call that NFV 2.0, in my mind. It’s all about taking out the need for human intervention and processes. That’s when the network gets truly dynamic.”
Hill’s vision of NFV 2.0 involves Big Data and analytics, allowing the network proactively to feed back in real time, driving orchestration through a policy engine: “You’ll be able to see what’s trending in the network in real-time,” he enthuses. “When a network is self-optimising, that’s when you start to dramatically remove the costs out of the operation of a network. Capacity can be matched with demand. I don’t see NFV 1.0 creating those sort of savings today, but as a carrier you’ve got to be on board. The train has left the station.”
There is a subsequent phase to look forward to, says Hill, where business orchestration and network orchestration are truly in harmony: "That’s NFV 3.0 and it’s going to result in a very different business model. Will this take time? In some ways perhaps we’re closer than we think. Carriers aren’t there yet – but they can see where they’ve got to get to.”
Alcatel-Lucent, which has been acquired by Nokia in a deal worth €15.6 billion, has traditionally been focussed on hardware but is now adapting fast to embrace a virtualised world. It is already beyond the PoC phase, having helped Middle East carrier Etisalat revolutionise its access network, both in its home market and in Sri Lanka, with its virtualised radio network controller.
Phil Tilley, marketing director of cloud solutions and strategy with Alcatel-Lucent, believes that the industry’s early tests and trials were about simply seeing what was practical and possible. He thinks the next stage is for carriers is take the individual use cases they have experimented with and see how interchangeable they are: “We’ve seen operators like NTT and China Mobile carry out PoCs with various different vendors, us and a number of our competitors, and start to ask what they can mix and match,” he says. “After all, what is the real value of NFV if you can’t mix and match across platforms?”
A decade in the making
NFV is widely viewed as a 10 year transformation. If operators haven’t already implemented the rst phase of NFV, they risk not having initiated the organisational structural change required to make NFV work in the long run.
“It’s as much about organisational change as technical. The most competitive service provider is the one who grasps this quickest," says Tilley.
The development of new skills is paramount, as well as carrying out retraining across the business and ensuring that siloes are broken down: “Can the networking guy pick up IT as quickly as the IT guy can understand network protocols? The two are converging right now and we’re starting to see teams working more closely.”
The experience of Telekom Austria Group (TAG), another carrier in the NFV vanguard, appears to conrm Tilley’s view. Sascha Zabransky, its director of service and IT, says its trials have shown there is room for improvement, especially in regards to architecture and performance: “Firstly employees need to digest the new technologies and understand how things work and secondly, a company’s operational model has to change, moving from a silo environment to horizontal layered structure. is carries a number of challenges like the internetworking of dierent organisational units in case of the need to resolve failures.”
But is NFV primarily about internal streamlining, or providing an enhanced service to customers?
“I see NFV as both about increasing the efficiency of our network and its reliability, and on the customer side it’s about letting our customers benet from those efficiencies without buying the network,” says Mateo Ward, CEO of Neutrona Networks.
The Latin American carrier is presently moving from NFV and SDN trial to actual deployment with Juniper Networks as partner, launching later this year with what will be the region’s most virtualised wholesale over.
“NFV lets us run two side by side networks,” he adds. “Without virtualising our routers we would have had a much riskier migration from one physical router to another."
Many of the early NFV implementations have, like Neutrona’s, been high-profile collaborations between big ticket vendors and their carrier partners - CenturyLink and Cyan working on NFV-enhanced services for enterprise customers, SK Telecom and Samsung harnessing NFV for the support of a national IoT initiative, Orange and Nokia Networks virtualising IMS, and so on.
But NFV is also an opportunity for more specialised vendors to come to the party, often working in complex groupings to which each brings a unique perspective. A prominent example is Intel, Red Hat, Procera Networks, Openet and Amartus which worked collaboratively with ETSI NFV to present a proof of concept of a virtualised real-time OSS/BSS at MWC in Barcelona.
Unlike most of the other headline-grabbing NFV collaborations, this one did not involve companies that were already partners or customers.
Cam Cullen, VP of global marketing at Procera, draws an important distinction between simple virtualisation and ‘real NFV’: “Much deployment at this stage is likely to be simple virtualisation of existing functions, making them software-only,” he says. "The next phase is orchestration, and a good start has been made. We’ve built our own VNF manager, to help customers get to real NFV. We’re working with other orchestration players in the market. As a small vendor, we can’t force an operator to just adopt the one orchestration system. Openness is a critical part of our strategy.”
Only two years into NFV and already it is clear that open source solutions have the power to provide an industry-wide defacto implementation for NFV and SDN.
OpenDaylight, for example, provides a compelling opportunity for mid-range carriers to trial and implement NFV solutions. Tilley concedes that open source is a great way to get innovation moving fast: “But I see it as a bit like a kit car that you have to build yourself,” he warns.
“There’s a lot of nuts and bolts you have to deal with. With CloudBand we bring it all together for you and give you a production-grade car.”
In these early phases, it is hard to call any one model or philosophy a clear leader. Bengt Nordstrom, founder and CEO of independent mobile strategic consultancy Northstream, believes that it is all too early to talk about NFV ‘success stories’ or ‘lessons learned’: “NFV implementation is going to take time,” he warns. “It’ll be a gradual process, since it directly relates to operators’ previous infrastructure investments reaching their end of life. Operators will introduce virtualisation as and when they replace their existing legacy applications and systems.”
If it’s too early to declare NFV an outright success, the telecoms industry can at least allow itself a little congratulation for what has been achieved so far. Top to bottom transformation is never an easy task, but the journey is well underway.
ARE SMALLER CARRIERS GETTING LEFT BEHIND?
It is inescapable that most of the NFV headlines so far are collaborations between Tier 1 carriers and their vendor partners. So is the industry effectively excluding the lower tiers?
“We’re not seeing so much NFV in the lower tiers,” confirms Rob Morrison, director product management, with CSG International. “The same sort of benefits should be available to smaller carriers, like simplification of business continuity and maintenance. I’d expect that now some vendors have quite mature offerings in NFV we’ll see more of it deployed by these smaller carriers.”
There is no doubt that at its present stage of development, Tier 1 vendors are extremely important to NFV. It’s these big names that have the dollars to throw at the sort of R&D and lab trials that are vital to the industry, and it is they who will be laying down the key ground work and overcoming all the early technology challenges. But they don’t have exclusivity in innovation.
“I also think there’s an important role for smaller vendors because they have different business processes, and often work in a more agile manner,” believes Justin Paul, head of OSS marketing at vendor Amdocs. “Tier 1 operators sometimes have a more traditional approach to R&D and testing. Smaller players can be very innovative and look at challenges in a different way. We could see Tier 2s and Tier 3s offering some very innovative NFV services more quickly through a more entrepreneurial approach.”
Since smaller operators have less complex networks, it is in theory easier for them to experiment to discover what works and what does not, says Ravi Palepu, head of telco solutions for IT consulting firm Virtusa: “They need pragmatic approaches as every company’s journey to virtualisation will be different,” he adds.
Certain smaller operators are already weighing with their own solutions, partnering in some cases with smaller vendors. Tele-Post Greenland, for example, is working in partnership with Dublin-based software developer Openet on a move away from reliance on dedicated hardware appliances. By leveraging standard IT virtualisation technology it has been able to consolidate multiple equipment types onto industry standard high volume servers, leading to reduced opex and capex expenditure. Vip Mobile Serbia has achieved something similar through using ‘commercial off the shelf’ equipment.
“NFV creates new opportunities for smaller operators who must differentiate themselves in a highly competitive market,” believes Ben Parker, chief technologist for with operational intelligence platform developer Guavus. “NFV represents a way in which service providers can create a more cost effective network and in turn price services more competitively.”
But other Tier 2 players are pursuing a more cautious approach, and that’s fine, argues Bengt Nordstrom, founder and CEO of mobile consultancy Northstream: “Many mid-range operators have assumed a ‘wait and see’ approach - they’re watching for the results from the Tier 1 operators’ trials and also waiting for more mature NFV products to arrive on the market. It’s a safe strategy, but it’s also important for midrange operators to use this time to make the necessary preparations for their future NFV implementations, such as developing the business case and also putting in place a practical and realistic roadmap for the changes that their networks will need.”