Windstream embarks on orchestration journey with SDNow service

Windstream is taking some of the lessons it has learned from SDNow into other areas of its network. (Pixabay)

Windstream is taking on orchestration and virtualization one bite at a time with its SDNow service.

Windstream announced a major expansion of its SDN Orchestrated Waves (SDNow) service in October. SDNow is an optical wave service that features 10G point-to-point circuits for 1,500 long-haul route combinations. SDNow, which is available in more than 50 markets nationwide, utilizes multivendor service orchestration and automated provisioning to deliver services to Windstream's transport customers.

In this Q&A, which was lightly edited for length and context, Arthur Nichols, vice president of network architecture and technology at Windstream Enterprise, talks about how SDNow is the starting point for orchestrated, intent-based provisioning of other services.

Fierce Telecom: Does SDNow include machine learning and artificial intelligence that will feed into more automation at some point?

Arthur Nichols (Windstream)

Arthur Nichols: In the realm of automation and orchestration, we are not at the stage of leveraging what I would categorize as machine learning. The notion with SDNow is leveraging SDN controllers to abstract the network. We kind of started at the core with that wavelength service, our long-haul optical services, but it's all about exposing intent-based provisioning across multiple vendors and multiple domains.

That allows us to push away from largely human driven design and more towards systems that have control of work-based path computation and controller-based service creation in connectivity, which allows for automation.

RELATED: Windstream weighs potential sale of certain fiber assets, consumer CLEC assets

FierceTelecom: What have you learned so far with SDNow, and what's next?

Nichols: I think that what we found is that for wavelength services there hasn't been a great demand for it to be in real time with an API sort of integration with our wholesale customers.

What we're working towards next is exposing that same intent-based provisioning model and network abstraction through centralized SDN controllers and orchestration beyond our long-haul optical core into the metro access edge for Layer 2 and Layer 3 services. Taking that SDNow service automation and orchestration and propagating it into more metro services.

We started with optical services at the core of our network and now we're developing these same techniques of centralized control and network abstraction and intent based provisioning across multiple domains into Carrier Ethernet arenas, and into IP and MPLS services in our metro edge services.

FierceTelecom: Where is Windstream right now in terms of machine learning, artificial intelligence and automation?

Nichols: I would characterize our automation, our machine learning and our artificial intelligence as largely at the beginning. We are focused on how we apply machine learning not only for network efficiency, but also for customer experience. The start of this journey for us is all about getting the data. So transforming how we ingest and retain data for the purposes of network efficiency and predictive analytics, and hopefully customer experience from the network itself.

I think streaming telemetry and stream processing informs the machine learning and then potentially artificial intelligence sorts of models. So it is early for us as we are trying to develop our expertise and build frameworks around that.

I will say that it seems like every vendor that I've talked to, for the last two or three years probably, has had some analytics package to sell us that centers around concepts of machine learning. The challenge for us is, how do we take the expertise that those vendors bring in with their very individual solutions and extract the information to form of models and algorithms to work with our own machine learning framework work and our own big data infrastructure?

That's really where we're focused on. How we collect the data most efficiently. How do we take that intelligence in machine learning—the correlations and algorithms from others as well as from our own development—into a common sort of framework and common infrastructure? That's where we really see value being unlocked, both for our network efficiency and how we are able to optimize the customer experience.