Many NFV use cases call for deployment of service elements at the provider network edge. 5G nodes, PGW, CPE, caches, all need to be placed near the end user to ensure efficient and cost-effective performance of the corresponding services. With NFV, elements previously implemented as proprietary hardware appliances, can be virtualized, deployed on COTS servers and centrally managed. In this presentation we'd like to share an approach to deploying distributed OpenStack-based NFV in a carrier network with an ultra-light edge composed of optimized Nova Compute Nodes and a centralized control plane placed in a provider DC with integrated SDN solution. We'll discuss different deployment options, their advantages and disadvantages as well as present the learnings derived from a PoC deployment with a demo of the resulting solution. In addition to addressing architectural challenges inherent to the distributed deployment model, we'll explore performance and distance related implications.
We'd like to share with the audience our learnings derived from an implementation of a distributed approach to OpenStack-based NFV deployment and the effect of physical distance and remote management on the performance of edge-intensive provider services.