Carta Blanco: NFV at Telefónica

The October 2012 white paper that brought the topic of Network Functions Virtualization (NFV) to the wider industry consciousness—and introduced the ETSI NFV Industry Specification Group (ISG) tasked with making NFV happen—was noteworthy for a number of reasons. First there were the bold claims made about the ecomonic and operational benefits that NFV might bring. Then there were the fundamental architectural changes that the paper outlined; the meshing of the telecom and IT functions and the move away from bespoke hardware. Third was the breadth and depth of impact the paper suggested NFV could have on the telecom equipment vendor landscape, and the power shift that this implied.

The last of these was, in fact, the first point to which the paper drew attention. Its opening line read: “This is a non-proprietary white paper authored by network operators.” A loaded statement if ever there was one.

Those authors, representing 13 fixed, mobile and integrated operators from across the world, did not downplay the challenges inherent in bringing about such seismic technological, structural and cultural change. Because of the openness that NFV is conceived in part to encourage, the development of a unified interface between software and hardware was identified as an imperative. The likelihood of a negative impact on performance caused by the use of standard hardware, the migration process from physical to virtualized operation (and the period of coexistence), security, network stability and the vital importance of automation were just some of the obstacles listed.
The paper offered a vision of the future—but suggested limited scope for a long, comfortable period of adjustment. Virtualization is already happening, its authors wrote, and within a few years transformational change will have taken place.

One of the most vigorous operators within the NFV ISG is Spanish incumbent Telefónica and, true to the message of accelerated change that the paper sought to communicate, it has not been sitting idly by. The firm marked the opening of MWC 2014 by setting out aggressive plans for the virtualization of its network functions, including detailed timelines relating to different network elements. The project has been dubbed ‘Unica’ and has as its headline aim the virtualization of 30 per cent of all new infrastructure by 2016.

Ahead of the announcement, at the firm’s sprawling headquarters complex in Madrid, Telefónica’s global CTO Enrique Blanco briefed a small number of industry press about the firm’s plans.

Familiar problems

The factors converging to create Telefónica’s need for change will be familiar to operators in many markets. The traffic across its network footprint is growing annually at between 25 and 30 per cent, Blanco says. Customers are calling for higher performance across fixed and mobile networks, and demand for the interconnectedness of the experience is increasing all the time. While 94 per cent of the total traffic that Telefónica carries is on its fixed network, more than 80 per cent of this traffic is driven by smartphones and tablets using wifi connections, he says.

Blanco’s team are squeezing 15 per cent efficiency improvements out of its networks each year, which is good but not good enough. Meanwhile Telefónica’s marketing team are demanding a network they can sell as the best in every market in which the firm plays and the firm’s vendors are being driven hard as part of what Blanco describes as “frenzied” network activity.

Improvements in performance, cost and differentiation on an altogether larger scale are needed, Blanco says—and those requirements call for “radical” transformation in the network. “We can’t wait,” he says, “because if you wait then nothing happens.”

Telefónica is “reaching the limits” of the performance it can drive from proprietary vendor platforms, Blanco says and the “lever” with which the operator can exert the greatest force for change is virtualization.

telefonica_fig.twoTo embrace such a radical architectural overhaul would be extremely challenging for a small, unaffiliated mobile operator; for an international carrier like Telefónica, which has more than 323 million customers in 24 countries, it is a monumental undertaking. It is hardly unusual for the sheer size of a project to deter operators from embarking upon it, so where is the impetus coming from in this instance?

The firm’s prominent role in the NFV ISG—Blanco identifies US market leaders AT&T and Verizon as well as German incumbent Deutsche Telekom as the three other leading influencers in the group—is clearly important to the momentum of Telefónica’s own project. The Spanish incumbent has 14 patents of its own relating to NFV, offering more evidence that this is a technological movement being led as much by the operator community as its suppliers (which is not to downplay the work of the vendors).

Meanwhile the virtualization work already carried out by Telefónica’s IT division, under the leadership of CIO Phil Jordan, has broken crucial ground for the operator. By the end of this year Jordan’s team will have virtualized some 40 per cent of the group’s IT servers and Blanco is emphatic in his depiction of Unica as an evolution that will eradicate historical boundaries within the operator. “Today we don’t see a clear differentiation between network and IT,” he says.


Wide-reaching benefits

But perhaps the most powerful force driving Telefónica into such dramatic change is the conviction with which Blanco identifies the benefits that wide-reaching virtualization will bring.

“Think about Telefónica in terms of how many countries we are operating in and how many different infrastructures we are operating,” he says, “and then think about what virtualization means. For us it means that we can change how we operate, change our governance model, exploit our capabilities in the datacentre, reduce the total cost of operation by up to 30 per cent and at the same time upgrade our platforms.”

Instead of having one IMS or one HLR per country virtualization will enable Telefónica to run single, central instances. Or, where regulation around data privacy and geographical location might create a problem, virtualization will allow for systems to be replicated on general purpose hardware in near real time.

Expanding on the example of IMS, Blanco says deployment times can be reduced from months to days. “Today we have IMS in all countries and in each case it is single vendor technology. Today we mainly use it for offering VoIP for our FTTH customers but in time all the traffic that is going through the large PABXs of our enterprise customers will be managed by the IMS.

telefonica_fig.three“But what if we need to bring 1,000 new enterprise customers online, or increase our FTTH customers by 1.5 million? I have to talk to my vendor, I have to increase specific hardware, create a new project, test it, deploy it and plan it—and this takes four months. If we do this using virtualized IMS using resources running in our datacentre we can do the same thing in four days,” he says.

Virtualization offers greater future-proofing in terms of new application development as well as expansion of existing services, he says. “We don’t know which of the services running on top of the IMS will be the most successful, we don’t know exactly when we will need to be fast and agile. But with Unica we have the infrastructure to deploy different services no matter if it is corporate VoIP or consumer video. We can create the IMS core and the application server on top of it in a matter of hours.”

The fact that the same underlying infrastructure will be used to provision network functions like the IMS or HLR and SaaS offerings for enterprise customers is central to the cost savings that Telefónica expects Unica to deliver.



The project, or the series of projects that combine to form Unica, as Blanco has it, has been underway since 2011. He likens the early phase to a large scale swap-out programme in the sense that the energy required to establish momentum behind the project was substantial.

In publishing a clear timeline for the project Blanco intends to make sure that this momentum will not diminish. But that timeline is fundamental to more than just the project, he says:. “Without a roadmap an operator doesn’t exist.”

That roadmap is the first of its kind that the industry has seen and the first milestone relates to the CPE. Telefónica has been running a trial in Brazil for more than a year with “one of the big Japanese vendors” that has taken the intelligence traditionally built into CPE such as domestic routers and moved it into the datacentre. This is “very easy to say and very difficult to do,” Blanco explains, but he believes it offers significant benefits.

“All of the software that traditionally manages the network and all the services that the customers can access are taken out of the customer equipment. So we are simplifying the hardware in the customer premises, reducing the cost of delivering that hardware to the premises and reducing the number of times we need to visit the premises to change things or solve problems.

“But it is not just good for us. All the services are now defined in the network premises and the customer that is moving [house or corporate location] can get all of their services switched across in bulk,” he says.
“This is a very ambitious project that has been engineered by our own R&D team and it will be launched and deployed in a massive way in the last quarter of the year.”

Work on the virtualization of the Evolved Packet Core is also underway as the vEPC, along with vIMS, vDNS and vDHCP—among others—are all within the first phase of Unica, which concludes at the end of 2014 (see slide). The vEPC should be online in June of this year, Blanco says.

As the firm progresses into the second and third phases the project becomes more complex, he says, picking out the Broadband Remote Access Server as a particularly challenging network element to move into the datacentre. Slated for achievement by the end of 2016, Blanco describes virtualization of the BRAS as “one of our dreams”.

He continues: “This is not a network function that is hosted on a standard platform but rather in specific hardware. It is very close to the customer premises and it needs a huge capability in terms of fiber and transport. This is a project that will take us much deeper into virtualization.”


Virtualized RAN

Meanwhile Telefónica is currently working on proof of concept trials for a virtualized RAN in its home market of Spain. Blanco nods to the pioneering work of South Korean operators in this area, which have “true hyperconnected networks,” he says. Success with a vRAN is some distance away; it is noticeably absent from the roadmap that the operator published in February.

“We are focusing first on control and signalling capabilities,” Blanco says, “because they are similar to IT capabilities and easier to virtualize. Then we will go to payload and then access. Maybe one and a half years ago this was just talk but today we are defining it and, from June, we will start to run network as a service in the datacentre.”

telefonica_fig.oneThe stated aim of Unica is to have 30 per cent of new infrastructure virtualized by the end of 2016, so another challenge is clearly going to be maintaining a huge portfolio of legacy infrastructure.

Telefónica might be able to virtualize many of its network functions but it can’t take quite the same approach to its human resources. Change on this scale brings great technical challenges but these are usually matched by cultural issues. And when a project is focused on centralisation, as this one is, these issues could be quite serious. So how is the human side of this process being managed?

There is widespread conviction among the network team, Blanco says, but he concedes that “it is more difficult to change the culture than the technology.” He makes reference to “internal resistence to what we’re proposing” but returns to the importance of a clearly visible and well communicated roadmap in persuading the wider employee base to embrace change.

There is a vitally important human element in the customer relationship as well, he says. Telefónica cannot be seen to treat data protection issues as anything other than paramount, and SLAs for enterprise customers who might have doubts about the performance and reliability of newly virtualized iterations of services on which they have come to depend cannot be breached.


Standardized hardware

The credibility of NFV is another significant challenge to Unica, Blanco says, and this comes into play when discussion turns to the hardware that will run all of these virtualized functions. For all the talk of using commercial, off-the-shelf (COTS) hardware for NFV, the issue is far from settled. Blanco says that different decisions will be made for different functions.

“In many cases you can use completely standard IT hardware, but not for all,” he says. “But there have to be improvements and the NFV ISG is pushing for that. When you virtualize you lose some efficiencies in the hardware while you gain in flexibility. But we need efficiency as well and operational issues, like trouble shooting, are not as good [in a virtualized environment] as we are used to in networks,” he says.
Modifications to standard IT products will be necessary, he says, in the hypervisor and orchestration layers for control and signalling and in hardware for payload. “The hope is that these specific modifications will be common in the IT industry within two years,” he says.

Threat to vendors

It is telling that Blanco talks in terms of improvements in the IT industry. The wider shift of which Unica is a part has been positioned, not unreasonably, as a threat to the traditional telecoms vendor community. And when Blanco says “we are doing this to break vendor lock-in” it’s not hard to see why.

There are two elements at play here. The first is the coalescence of traditional telecoms and IT vendors, bringing new competitors for everyone, and the second is the likely intensification of competition among those traditional telecoms vendors.

“One of the key things is that Unica should be a multi-vendor approach,” Blanco says. “The traditional vendors don’t like to break this lock-in. But if we define a multi-vendor reference architecture and say that we will not have any network functions that do not respect this, sooner or later they will come to us.” If a new IMS can be provisioned and live within a few days, after all, an existing IMS can be taken offline just as easily.

In keeping with this approach, the list of vendor partners involved in Unica is a long one. Telefónica is working with Huawei in its Spanish datacentres, Alcatel Lucent in its Mexican datacentre and NEC in its Miami installation. Blanco says he is working “very closely” with Broadsoft and namechecks NSN and HP, Cisco, Ericsson, Juniper and Intel as well. Different vendors within this group have different stories when it comes to NFV and it will be interesting to see how each one progresses within Unica.
But Blanco is not about to change his relationship with legacy vendors for a similar one with new suppliers. His team are investigating different hypervisor solutions and he insists that cost efficiency is essential: “If I’m saving money by breaking vendor lock-in, I don’t want to pay that money to the hypervisor vendor.

“There are some that are more expensive and have more benefits and some that are less expensive and we’re testing all of them,” he says. “But this is nothing new for us; this is our life.”

Blanco talks about NFV with a great deal of energy and projects a real sense that this is seen as an operator movement with which the vendor community must fall in line. Change is inevitable and only by embracing it can Telefónica hope to have an influence on its direction, he says.

“The border between the people of the network and the people of IT is disappearing,” he says. “This is an absolutely different way to think. In the next three years, if you walk around our own data centres, you will not be able to distinguish whether the server you are looking at is providing a service to a large enterprise customer, a cloud service for our own guys or if it is running the HLR, the IMS or the EPC.”
Blanco employs an overused phrase when he talks of “changing the paradigm of the industry.” For once, though, it doesn’t sound so hackneyed.

One comment

  1. Avatar fvunen 06/05/2014 @ 10:39 am

    Nice article. Is there a possibility to include a readable version of the roadmap graph? The resolution of the roadmap is so low that you can’t readily see when what is targeted.

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.