Operators Must Cloudify at the Edge
Matt Haines has done time in the cloud. He was in the business when centralized cloud computing took off in the early part of the decade, and he's watching the cycle of product and standards development repeat itself today as the cloud drifts inevitably outward toward the network edge.
The difference this time around is the players involved, and the fact that they can learn from the lessons of their forerunners in the centralized cloud movement.
"The reality is everybody today, all the operators today, actually do edge work; they just tend to do it in special-purpose appliances and applications that sit out at the edge," says Haines, referring to the specialized equipment and workloads that network operators manage out in the field. "And so the conversation I keep having is: Why don't we start to consider those as general-purpose cloud resources?"
Now a fellow at Accenture -owned Interactive Broadband Consulting Group LLC (IBB) where he leads the Cloud and Software Transformation practice, Haines spent several years first as an engineering vice president for HP Cloud Services, and then as a cloud services vice president for Time Warner Cable.
With IBB, he's preaching the idea that operators who are leading development at the network edge can and should apply the principles pioneered by established cloud providers. Instead of building out single-purpose equipment systems, operators should be considering deployment of commodity hardware with software architectures that are flexible enough to be leveraged for all kinds of applications.
"When you think about what a cloud is," Haines says, "it's really this ability to provision it on demand and schedule something out on to it that I haven't had to work out with you [the application provider] ahead of time... Why don't we start to transition that [cloud model] into your own use of what you do today at the edge?"
Conveniently, network operators are already investing in increasing their edge capacity. They're adding fiber, distributing network features to increase local availability and virtualizing functions to optimize the use of network resources. As operators make these network investments, comments Haines, they should also be taking a long view on how those enhancements can be put to use in the future.
"Pretty soon," notes Haines, "you're at the point where you can sell capacity in your edge [sites], in your 500 or 5,000 edges based on how close it is to a customer, and deploy it exactly the way I would deploy it to a public cloud."
For cable operators, one component of this process is today's move toward distributed access architectures. Cable companies are spending money to push cable modem termination system (CMTS) functions out to the edge of the network at the same time they extend fiber deeper and exponentially increase the number of optical nodes they deploy across their subscriber territories.
"Right now as all the companies continue to push ... more and more capacity into the fiber transition point to handle the build-outs for faster, faster speeds and so forth, this is the right time to think about how we start to look at those resources as turning them into cloud," say Haines. "And turning those [resources] into a cloud is really a software exercise of putting down the layer of software that is the cloud software, the OpenStack stack that sits there that all of a sudden turns commodity hardware into something you can provision a virtual machine or more likely a container into."
"Eventually," he adds, "you get to where you're provisioning capacity and applications out to those edges, [some of which] aren't your applications because you're selling to somebody else just like the public cloud companies do today."
Is this idea of a cloudified edge something that others in the industry are also thinking about? Absolutely.
Consider how content delivery network company Qwilt Inc. is working with operators to architect software nodes into their last-mile networks. Qwilt has built these nodes to run on off-the-shelf hardware, and the nodes can be programmed to accommodate content from a wide range of providers across a wide range of access network types. Each node becomes a resource at the edge for content delivery, which operators can use for their own purposes or, at least theoretically, as an asset in negotiations with partners and customers for edge capacity. (See Qwilt Starts Caching on Verizon Edge.)
Or consider Senet . The service provider has created a system for managing and billing out capacity on a global virtual Internet of Things (IoT) network. The concept is that any entity can own a part of that virtual network by installing standardized low-power wide-area network (LoRaWAN) gateways. The capacity enabled by those gateways can then be employed by the owner to deliver IoT applications, or sold off to customers with applications of their own. (See Senet Virtualizes Worldwide IoT Network.)
There are a growing number of applications that can take advantage of storage, computing and connectivity resources at the network edge -- from video delivery, to applications that analyze local behavior (think connected traffic cameras) to latency-sensitive IoT services like asset tracking. And the more these applications develop, the more edge network resources they'll need.
"What we're about to witness is the same cycle that... we saw in what I would call the centralized cloud area," says Haines.
Demand is growing and solutions are proliferating. And if network operators want to take advantage of the cycle this time around, they'll need to cloudify at the edge sooner rather than later.
— Mari Silbey, Senior Editor, Cable/Video, Light Reading