DENVER -- NFV & Carrier SDN -- As virtualization continues to alter traditional wide area network architectures, the edge of the network is becoming harder to define.
The edge is being redefined as a broader range of facilities, including customer premises equipment (CPE) elements, expanding the number of processing and compute locations in a network by an order of magnitude, in some cases from thousands to tens of thousands or even millions of locations. So said Roz Roseboro, principal analyst, cloud infrastructure and management, at Heavy Reading , here Tuesday ahead of a panel that focused on the opportunities and problems that arise from distributed IT facilities. (See CenturyLink's McBride: NFV 'Introduces a New Kind of Cloud' .)
And with 5G bringing the likelihood of more extensive small cell deployments, those locations become even more valuable than before, she said.
In a survey of service providers, Heavy Reading found that the bulk of services and resources are still housed at centralized data centers, central offices/nodes and network PoPs, while a small fraction of those computing resources are contained in customer premises equipment, mobile basestations and metro aggregation sites.
But a subtle shift is underway, as those compute resources become more distributed on the network and get woven into these smaller (and much more numerous) edge locations. When the survey asked those service providers to size up that situation five years from now, it showed a small, 1.2% gain in compute resources being housed in customer premises, and a 1.3% gain in compute being housed in mobile basestations. At the same time, the survey found that the concentration of compute in central offices/nodes will dip by just 2%, and drop by 1.9% at PoPs.
Those results indicate a minor change in how those resources will be distributed in the near future, but also illustrate that the picture on where the vast majority of network resources and compute will be located won't change overnight -- the number of challenges associated with the deployment, management and monitoring of edge computing resources is legion. "It's important to keep [this change] in context," Roseboro said.
Among the challenges faced by network operators are those related to the physical location of edge resources, and the characteristics of those locations -- these will mostly be environments with space, power and cooling constraints: Doors can be too short to accommodate some equipment; floors might not be sturdy enough to handle the weight; enclosures may need to be weather and vandal-proof. In addition, there will be tricky compliance and regulatory components to factor in.
"Those are real issues," she said.
So where is the edge, or where will it be? Panelists had some different thoughts.
It's starting to move to the customer premises and dissipate from the core and the metro, said Reinhard Florin, GM of North America at Telco Systems (BATM) .
"We are following the CPE … the edge can be there," depending on the use case, added Michael Heffner, VP PLM, Ensemble, at ADVA Optical Networking .
The edge, according to Vikram Saksena, chief solutions architect, office of the CTO at NetScout Systems Inc. (Nasdaq: NTCT), is the outermost extent of where there is "service awareness" -- whether that's at the CPE or behind the radio access network. Regardless of the location, visibility [[into network and application performance]] will still be critical," he said.
And while standards in this area are helpful, they aren't an end-all/be-all problem solver.
Standards offer "a great place for people with common interests to get together and share ideas," Bill Long, VP of interconnection services at Equinix Inc. (Nasdaq: EQIX), said. "But in reality … the predominant solution in the market becomes the de facto standard. The industry is just moving too fast for the standards bodies."
— Jeff Baumgartner, Senior Editor, Light Reading