The most fundamental issues faced by the Internet architecture are the scalability and quality (stretch, convergence, and stability properties) of its inter-domain routing system. Solving them requires to address multiple dimensions altogether i) the routing table size growth resulting from a larger number of routing entries, and ii) the routing system dynamics characterized by the routing information exchanges resulting from topological or policy changes. Both dimensions increase memory requirements but also processing capacity of routing engines.

The overall objective of this project is to design, validate, and evaluate by experimentation a cognitive routing system and component that combines advanced machine learning and routing techniques in order to efficiently address the Internet routing problems described above.