DOTE - Rethinking Predictive WAN Traffic Engineering

Explore a groundbreaking approach to traffic engineering on wide-area networks (WANs) in this award-winning conference talk from NSDI '23. Learn about a novel method that directly optimizes traffic flow using only historical data, eliminating the need for explicit future demand estimation or prediction. Discover how this stochastic optimization technique, enhanced by deep learning, achieves near-optimal traffic engineering quality while significantly reducing runtimes. Gain insights into the extensive empirical evaluation conducted on real-world traffic and network topologies, demonstrating the superiority of this approach over previously proposed methods. Understand the potential implications of this research for improving WAN performance and efficiency in large-scale network environments.