During her 20-year career at Verizon, Radhika Venkatraman, senior vice president and CIO of network and technology at Verizon, has helped the company navigate many different network technologies. In her current role she leads a large global team that is in the midst of automating major functions like inventory management, engineering and construction, service assurance, surveillance, and more. Her team is also responsible for implementing software-defined networking (SDN) and network functions virtualization (NFV). She spoke with Sue Marek, editor in chief of SDxCentral, about how Verizon is trying to model Facebook and Google in its network transformation instead of other telcos.

You have been involved in Verizon’s implementation of SDN and NFV. About a year ago Verizon published a reference architecture on this. Can you give me an update?

We have been thinking about network virtualization in a few different ways. We are thinking of it both as it supports revenue generation, which is our enterprise-based services. And second, as it supports our infrastructure.

We plan to leverage a similar construct for both revenue-generating, enterprise-grade services as well as our infrastructure network services.

We have a target architecture, which includes not just building a cloud, but also how we would have resource orchestration on top of it, and how we communicate between the cloud and the resource orchestration. We have this notion of service orchestration that allows us to do service management, and then we have this concept of a portal. That will allow us to communicate via an API [application program interface] to our internal communications and also to our enterprise customers and suppliers that may want us to host their VNF [virtual network function] in our ecosystem.

I am thinking of it like a virtualization platform that includes a marketplace concept, a service configuration concept, and service assurance. The service configuration and assurance will be model-driven so we can actually enable service and network design, cloud resource management, and visualize the service.

What is the status of your network virtualization for your enterprise services?

You may have seen our communication about our SD-WAN product. We did some of it on white box CPEs [customer premises equipment]. We are hosting it in the cloud. We are launching SD-WAN [software-defined wide area networking] routing security in a big way in the marketplace soon, but we are already doing trials with about a dozen customers.

We have deployed an orchestrator. We have numerous VNFs — about 13 of them — that support security, SD-WAN routing, and WAN optimization from a host of vendors on our platform, and they are coming at varying time frames.

We are thinking about a self-service portal for our end customers so they can do their own provisioning and policy management. They can do their own bandwidth management and swap out VNF services. They can change and service the policy management and do their own policy diagnostics.

The whole platform will be an ongoing journey. We are building new capabilities incrementally every month. We will be hosting more VNFs every month and adding more customers each month with the notion of having some sort of marketplace.

What about Verizon’s network infrastructure?

From an infrastructure perspective we have figured out all the things that we want to virtualize.

We have said we want to virtualize our multimedia subsystem, or IMS core, and we want to virtualize our mobility services, our converged voice and data services, our wireline voice services, and wireless voice services, and then there’s about 5 percent miscellaneous that we want to virtualize. We think we will have about 150 functions to be virtualized. My belief is that in 2017 we will go tactile with 60 or 70 of them. I don’t want to say whether they are the biggest or the smallest at this time.

All of the functionality that I described for the enterprise we will have to build in support of our operations. We’ll have a portal for vendors to support their VNF software, an automated way to track and manage services.

We have to partner with the suppliers. Some will be new companies and some incumbents. We are trying to define a common information model in terms of what we will accept from our suppliers. And we will use that to do our service assurance.

I feel very good about the progress that we are making. Now it is about day-to-day execution.

You mentioned an orchestration platform. Can you provide more details?

We think our framework can support multiple resource orchestrators. We have picked Ericsson for our hosted enterprise services, and we have picked HPE’s NFV Director for infrastructure services. I also have a startup company that I’m working with. We are using that for automation and for resource orchestrating some of the VNFs.

If we can move to a model that is based upon standards like NETCONF/YANG , and we can be very descriptive about our southbound communications between the resource orchestration and the cloud, then I don’t see why I can’t use multiple orchestrators to handle the VNF.

We only plan to have one service orchestrator. We are not planning to have multiple service orchestrators. We are looking at both the startup and HPE –  and this is something we will decide soon.

Do you prefer to work with startups or bigger vendors?  In the past, operators were concerned about working with startups because they thought they might not be able to scale quickly enough. Is this still a concern?

When you have a small vendor, you have to work much more closely with them. But if you have the skills to work with a small vendor, it goes much faster.

I’m seeing that with SDN — everyone wants to do something on their own. When AT&T built ECOMP, they used 400 people. I have 23 people on the orchestration team. We may add another dozen to that team. I’m convinced small teams will be like speed boats in this ecosystem.

We are trying to learn from the Facebooks and Googles about how they did this instead of trying to learn from Telefonica or AT&T.