SDN/NFV for optical networks

Workshop I: SDN/NFV for optical networks

Although Software Defined Networking (SDN) and Network Functions Virtualization (NFV) concepts, main architectures and potential benefits are well-known and understood, their applicability in optical transport networks is still to be fully explored and exploited. In addition, due to the analogue nature of the optical transmission domain, optical technology is still highly proprietary, closed, and difficult to model and standardize.

SDN/NFV are behind several opportunities and initiatives supporting the control and management of optical networks, including: i) the control of disaggregated optical networks, which involves utilizing whitebox components without compromising overall network performance, but reducing cost (e.g. in reduced geographical ranges that relax the optical performance parameters); ii) the implementation of autonomic networking using Artificial Intelligence and Machine Learning techniques to exploit data made available through network telemetry for optical monitoring and performance guarantees and iii) network virtualization and slicing, including the virtualization of optical hardware.

This workshop addresses these topics, with emphasis on the following questions:

  • What standards, interfaces, technologies seem to be more adequate for the widespread adoption of SDN and NFV within the optical community? Can OpenFlow be extended to fully cover the foreseen requirements, or will Netconf and YANG supersede it?
  • What are the challenges regarding optical disaggregation from a control and management perspective? is it feasible in a short term? Can we expect uniform information and data models? Are current modeling languages enough or will we have an explosion of de facto and de-jure standards, open models and control frameworks?
  • What is the role of the optical technology in network virtualization and 5G network slicing?
  • Next generation programmable hardware for optics should support some form of virtualization. What hardware functions can be virtualized? Can NFV be extended to “generalized” functions (e.g. Transmitter / Receiver functions, …)? What is the trade-off between virtualization and performance penalty?



Comments are closed.