Verizon tunes up AI internally for SDN and its business customers

While artificial intelligence has yet to hit its full stride, Verizon is using it both inside of its network and for business services and applications.

Internally, Verizon's Vickie Lonker, vice president, network and security product management, said her company was using AI to further enable automation. 

"We're using AI on the service and delivery side of the house to make sure we understand where we can automate tasks that people are doing," Lonker said in an interview with FierceTelecom. "Where we have conditions on the network that repeat themselves, or that have certain characteristics, then we can use AI to actually learn from those characteristics and automate the actions that the network or the controllers take to correct whatever that network condition is.

"Also, some of the capabilities within our managed SD-WAN and wireless LAN solutions include some AI."

Closed-loop automation is a key element of the service chains that Verizon builds. The service chains are prepackaged offerings that combine Verizon's infrastructure with software and hardware from vendors. Combined with orchestration, closed-loop automation helps Verizon build service chains composed of multiple functions. The service chains connect a set of users to a set of functions without requiring business customers to buy bundled functions that they don't want.

Related: Verizon: Service chains are essential and automation is a key focus

"So we're using it (AI) not just in what we're doing with customers in the SDN space, but also on our own networks as we continue to build out this infrastructure throughout Verizon to support our growing platforms," Lonker said.

Verizon's Shawn Hakl, senior vice president, business products, said Verizon has traditionally used machine learning and AI extensively in its security offerings, but now the company is using both for network correlation, network management and "deep skill security."

"The automation we put in place for the orchestration engine and the API exposure is what lets you tie into the machine intelligence layer," Hakl said. "We've got what I consider first-generation machine learning solutions around resiliency and protection as part of our SDN solutions.

"Let's say, for example, I know that a function is operating on node A, but node A hasn't come up so many times in a row now. I can replicate that same service chain in another node, reconfigure the network parameters and do a safe failover. So we've got the root of it for stuff like that, that use a combination of statistics and heuristics, but I think we're at the very early stages."