LONDON -- OSS in the Era of SDN and NFV -- There is a sea-change in how telecom network operators are talking about network management in the virtualized world, shifting away from the centralized orchestration and toward the idea of management fabrics combining smart, self-managing network elements, Heavy Reading Principal Analyst Caroline Chappell said here today.
Kicking off Light Reading's "OSS in the Era of SDN and NFV" event, Chappell said the network operators who pioneered deployment of virtualization are now starting "to grapple with scaling and dynamism" and finding that the traditional approach to automating management, using centralized and hard-coded automated workflows, becomes unworkably complex.
"What I am hearing is talk of fabrics and choreography, rather than orchestration, which implies centralized control," Chappell said. "There is going to come a time when [network operators] will have thousands of compute locations and tens of thousands of VNFs -- how is all of that going to be managed?"
The newer thinking is that network elements need to be programmed to manage themselves and that these smart elements must be able to respond to changes in their environments, she noted. Increasingly, they are organized in closed-loop management systems that are much more distributed and less centralized.
Because these kinds of solutions aren't generally available yet, many network operators "are building these for themselves, using a new generation of IT tools" such as Puppet, Chef and more. Instead of handing things off to their internal IT departments or OSS vendors, the virtualization pioneers are developing their own approaches, she said, that are service-driven.
"The OSS department's approach is unlikely to be intent-based," Chappell commented. "There has been very little abstraction in traditional OSS -- services are described in terms of individual boxes that need to be managed," not logical collections of network functions. "That's an agility killer and it leads to a huge waste in all that network engineering knowledge and talent."
The problem is that many operators don't have engineers that want to write Puppet manifests or Yang models, she said. But savvier operators need to empower the staff that are willing to work in this fashion and not curb their enthusiasm. "These are the tools that the digital natives you are trying to recruit are going to use in the future," she said. "Getting to grips with this tooling is going to be important."
Another major change is the blurring of the line between management and control, as more management functionality "disappears into the network," Chappell said. In traditional telecom networks, OSSs operate entirely separately from the network, but that is changing. The close control loops of the future will be critical to the operation of the network, and offer more real-time controls, she noted.
Chappell pointed to AT&T and its announced ECOMP (which stands for enhanced control, orchestration, management and policy), the operations system for its virtualized network, which is not an external OSS but is a virtual ecosystem embedded in the virtual network infrastructure. In AT&T's case, different network domains -- transport, access and cloud infrastructure -- will have their own service and resource models and closed control loops, which AT&T is building in an agile, iterative manner. (See AT&T's Chiosi: Unite on Open Source or Suffer.)
"So the end-to-end network management will be a world of many models, and the architecture is metadata-driven, rather than requiring programmers to write code," she said.
So one question going forward is how the MANO -- the management and orchestration layer defined by the ETSI NFV ISG -- fits into this more distributed model. "Operators are falling out of love with MANO," Chappell said. "It's too complicated."
One reason is the confusion around service orchestration, which is increasingly where the operators are focused, and the MANO's approach to NFV management, which is based on an architecture plan that is now three years old, she said. Those criticisms could well be addressed by future developments within the ETSI NFV ISG or other open source efforts such as OPNFV.
— Carol Wilson, Editor-at-Large, Light Reading