User-driven Path Control through Intent-Based Networking

The UPIN (User-driven Path verification and control in Inter-domain Networks) project aims to implement a mechanism for a user to control the way data are traversing the network. Here we investigates the possibilities and limitations of Intent-Based Networking (IBN) for user-driven path control. Exploring several intent translation techniques allows us to define four main factors that influence the design of an Intent-Based Networking approach. The level of control, level of required knowledge, type of language, and network all influence the design. Based on the UPIN project demo, we design two approaches: technical-centric, which focuses on enabling Intent-Based Networking, and human-centric, which focuses on relieving restrictions from the expression methods for the user. Rasa is used to create a chatbot interface for the human-centric approach. We experiment with several configurations to discover the optimal pipeline for our training and testing data. Results indicate an 85 percent accuracy of intent recognition and 93 percent accuracy of entity extraction. Although the accuracy is not high enough to allow the Intent-Based Networking implementation to make decisions without supervision, the implementation proves to be a viable method of expressing intents for user-driven path control.