Machine learning informed segment routing path selection

Here we present a solution for practically demonstrating path-aware routing and visualizing a self-driving network. From the AI-enabled traffic engineering (TE) side, we explore AI with supervised learning to help understand traffic flows and deep reinforcement learning to learn optimized network utilization results. From the path-aware network side, we explore the segment routing architecture, which introduces a novel approach by utilizing the polynomial residue number system (RNS) in contrast to traditional SR solutions that rely on port switching. This method enhances performance and offers advanced routing capabilities, including flexible path migration and robust failure recovery. In our proposal, various network aspects such as performance, latency and utilization are improved with APIs, which communicate traffic snapshots and prediction supervised learning algorithms to predict future available bandwidth on the network, returning options for choose the best path.

Contact:

To learn more about this technology, email partnerships@ornl.gov or call 865-574-1051.