Big Data Is on CenturyLink's 'Automation' Wish List

Big data has found its way into CenturyLink's NFV efforts, as Enterprise Architect Kevin McBride tells Light Reading.

And maybe that was inevitable. One hope for any new-age networking project -- whether it's SDN, NFV or security -- is to use analytics to inform the network what to do next. "You can start asking questions of that data and have it place a workload," McBride says. "The feedback loop is what makes this powerful."

Ultimately, the feedback loop could become predictive, so that the network places virtualized network functions (VNFs) as the need starts to arise.

"There's still the need for human decision-making to do placement" of VNFs, McBride says. NFV orchestrators aren't automated, "not to the extent that we need them to be."

The idea of the feedback loop is nothing new. But one thing that's changed for CenturyLink has to do with the data collected by analytics: It's coming from many different network domains.

That's led CenturyLink and other carriers to consider a multidomain service orchestration layer -- kind of a global orchestrator overseeing smaller orchestrators. And that multidomain orchestrator needs to find needles in that haystack of information. That means creating a data lake and building up big data expertise, McBride says.

For more on automation, visit the dedicated automation content channel
here on Light Reading.

Big data thus becomes one more piece of the sprawling, evolving NFV puzzle. CenturyLink itself is in a third generation of SDN/NFV deployment and has learned along the way. The multidomain concept, for example, grew out of the need to cope with multiple NFV orchestrators and with vendors' not necessarily compatible VNFs.

Among its latest strategies is a shift to agile software-development processes and the focus on creating an abstraction layer around old OSS/BSS systems. The latter step should help simplify the implementation of NFV. (See CenturyLink: Kill Complexity to Speed NFV.)

Among the pieces still missing, in McBride's view, is a common language for describing services. That would help with interoperability among different carrier and vendor NFV efforts. It would also be a necessary step for creating an intent-based framework, one where you tell the network the desired outcome, and the network elements configure themselves to make it happen. (See Cisco's 'Network Intuitive': A Risky Transition.)

"We need to be focused on adding more intent-based logic to description files" and "being more collaborative to one another to describe things," he says. "We work with other carriers. When they bring workloads into the CenturyLink NFV, they should be able to describe their workloads."

That would be a key step, but like the feedback loop, it's not a new idea, says Heavy Reading analyst James Crawshaw.

"In order for that magic to work, you need to have an agreed language and agreed models," Crawshaw says. "It comes back to this topic of information models and data models, which is reminiscent of the work the TMForum has done in the past and was never really implemented."

— Craig Matsumoto, Editor-in-Chief, Light Reading

kq4ym 8/18/2017 | 11:16:28 AM
Re: Top down vs Bottom Up It will be interesting to watch the developments in this area as operator like CenturyLink try to get to the automation platforms that will bring not only a better return on investment but some sense of simplifications. As noted, "In order for that magic to work, you need to have an agreed language and agreed models," and there's lots of work towards that end in progress.
ChiefMar76368 8/10/2017 | 4:34:31 AM
Top down vs Bottom Up I think most operators have recognised the need for the concept of an "orchestrator-of-orchestrators" of some form. What tends to get overlooked however is that the higher up the layers you go, the nature of that "orchestration" is different, and involves different information (more commercial and less network) and different evaluations. To the best of my knowledge, no "orchestrator" can tell you whether it will be cheaper to spin up a VNF in one datacentre vs another. Or whether the fact you're trying to migrate away from vendor X means you have an overriding business rationale for choosing vendor Y's VNF. The process of analysing data (even Big Data) is completely different than the process of deciding which, of all the possible options to choose from, is the optimum course of action. Big Data technologies may help with the first, the second is much more in the realms of AI: automated decision-making is really where the breakthrough is. 
James_B_Crawshaw 8/9/2017 | 4:30:29 PM
Before I get lynched by the TMF community ... ... what I meant to say was that intent-based modelling is reminiscent of the work TMF has been doing for many years on the Shared Information and Data model. The SID was arguably too complex with more of a focus on lower level data modelling than higher level information modelling. The overly ambitious scope of the SID may have made it impractical but they were on the right tracks architecturally speaking. 
Sign In