Announcements |   placeholder

Intent for Configuring High-Performance Science Networks

In this talk, we will describe our approach of using natural language processing for network engineering to dramatically simplify network configuration with intent-based north-bound interfaces. As demonstrated at the 2016 Supercomputing conference, using our tool, INDIRA (Intelligent Network Deployment Intent Renderer), allows end-users to ‘talk’ in English to the network, converting their needs into network commands, automatically instigating OpenNSA (for multi-domain network path provisioning), Globus (for big data file transfers) or catering to other performance needs. Compared to current approaches, INDIRA uses ontology engineering such as semantics (OWL, RDF) and data dictionaries, to allow both, a ‘human-like’ conversation with the network for intent capture and a one-command intent rendering over the network. INDIRA’s focus areas include user-friendly approach, automating network provisions, technology-agnostic integration and allowing network feedback to users. Current results have demonstrated enhancement of user-network experience via a chat-like method, automated path provisioning, configuration of file transfers with QoS and optimizing network performance for science applications.


Part of session

SDN and automation

Related documents