Published On: Wed, Jan 25th, 2017

Visibility & Analytics Key to Realizing the Full Value of SDN & NFV

Service providers are banking on SDN and NFV to reduce capital outlays and transform how they build and operate networks to deliver services that are resilient, adaptive and customer-driven, lower cost networks that are far more intelligent and agile than what they have today. Who can argue with that?

Network equipment vendors and the open source community have been busily developing a wide array of standards and products for software-controlled, virtualized infrastructure in both data centers and wide area networks. It’s early days in terms of full-scale operational deployments, but there is no shortage of options for service providers to consider. However, one key piece of the puzzle has not yet fallen into place.

Truly intelligent services require feedback loops based on sensor input that drive automated orchestration for service provisioning and activation, as well as automated network and workload performance management.
Although it’s understood that visibility data provides the sensor input and analytics engines generate the critical operational intelligence needed for closed loop automation, many questions remain unanswered.

What are the right telemetry data sources? How should the virtual infrastructure be instrumented at each layer to provide telemetry data? What is the role of virtual probes? What mechanisms should be used to collect data? How is data collection controlled and managed?

What takes place at the analytics layer? How is data analyzed in real-time and actionable intelligence provided to drive those feedback loops? What types of specialized analytics tools are required?

No doubt, many smart people are working hard to answer these questions and more, but until we reach a consensus on the answers, service providers will be unable to realize the full value of SDN and NFV.

In future blog posts, I’ll be exploring this topic in more detail.

For more information, contact me at scollins@acgcc.com.