Two service provider executives from companies deeply engaged in virtualization offered some words of caution Tuesday: When it comes to using SDN and NFV to automate network operations and improve services, there are no short cuts or off-the-shelf solutions.
Bill O'Brien, director of Adaptive Platforms at CenturyLink Inc. (NYSE: CTL) and Andrew Kusminsky, COO & chief strategy officer for Perseus , a global managed services company, spoke in a virtual roundtable sponsored by marketing firm JSA and focused on real-world deployment issues for NFV, SDN and big data. Although their companies and services are quite different, both men emphasized the need for companies to develop their own expertise, and to take the time necessary to do that.
"Individual companies will need to build up their capabilities in data science in order to allow for this transformation to happen," O'Brien said. That means putting the right platforms in place for automated SDN and big data, he said, but also understanding what those platforms look like and what they can do. "It's not trivial. You have to put in the time. When you look at large web firms, they have invested almost a decade in this and they are doing some interesting things now."
Looking to a single vendor for integrated solutions may have its appeal but it won't work in the long run because no one vendor can address all issues, O'Brien said.
"It's not something you can buy completely off the shelf, you are going to have to make that investment in your organization as well as the knowledge within your company so you can begin to evolve and develop the algorithms to make the smart systems and smart networks," he commented.
Zeev Draer, VP of strategic marketing for MRV Communications Inc. (Nasdaq: MRVC), and the lone vendor on the panel, agreed with O'Brien. "The biggest myth of SDN is that it is one-size-fits-all," he commented. "This is a long process -- there are no short cuts."
Kusminsky's firm serves the demanding financial services industry and has deployed SDN as a means of delivering on-demand services that customers can turn up themselves through a portal. Only by insuring its own internal expertise is Perseus able to bring high-value customers onto a new technology platform with confidence, he said.
"Fear of going onto a new platform is always a little scary," Kusminsky said. "There are no short cuts to building a network like this, and as long as you don't take any shortcuts and you hire the right people with the right skills, and go through the right processes and the right inventories, and do things the right way, then you can overcome the objections any clients will have. We have a lot of clients already using this platform very happily and very successfully today."
Specifically, Kusminsky called out a recent circumstance where two separate network events took down its customer's connectivity to financial exchanges in Hong Kong and Tokyo.
"Our client needed to be able to turn up new capacity between those two points," he said. "By logging into the Perseus SDN fabric, they were able to enable their services to go live within 24 hours. Enabling a customer to turn services up that quickly was a very powerful tool."
Putting services into their customers' hands through a portal is a key first step to "uberization" of connectivity, he said, but right-sizing is also key. Being able to turn up -- or turn down -- capacity based on intelligence that tracks efficiency of usage means customers are happier.
In CenturyLink's case, company officials realized early on that creating the kind of automated systems they want as part of the move to adopt SDN would require collecting massive amounts of data for analysis. The decision was made not to create a separate data collection and analytic system and process but to build that into the network platform itself, and to specifically enable feedback loops.
"With the robust set of APIs that are typically inherent with any SDN-NFV platform and the automation that goes into play, we began to ask ourselves how can we automate the decisions typically made by humans, and then use those and build intelligent feedback loops back into the platform," so the network can adapt itself to operational behavior, as well as customer demands across a range of services, O'Brien said.
"There was a need for building out a big data platform leveraging a lot of the typical type tools we see between Hadoop, Spark and using various classes of machine learning, and bringing data scientists on board to accomplish that," he added.
The data collected can be used to drive service level agreements or to predict the outcomes of network behavior based on previous experience and proactively mitigate issues for customers, he said.
Predictive analytics are a key element for Perseus, Kusminsky said, because they enable the networks to react to applications -- and it is the applications that drive customer behavior and consumption of network services.
As often happens in discussion of NFV and SDN, the talk turned to culture changes and the need to get telecom into more of an IT or software-driven mindset, something Draer strongly supported.
As O'Brien noted, however, the changes in culture, organization and procedures will also take time as they are taking place across these large organizations "and impacting a lot of disparate systems."
"Here we are making this mind-shift change where we are taking about a federation of internal platforms in that shift to DevOps, and this highly integrated data model, where we are having systems and machines make decisions that were the domain of humans and that is a huge shift, so there is a trust layer that needs to happen," he concluded.
That shift also takes time, and probably doesn't come with a short cut.
— Carol Wilson, Editor-at-Large, Light Reading