Shooting Large-scale Traffic Engineering by Combining Deep Learning and Optimization Approach

The rapid growth of global modern wide area networks has posed significant challenges to traffic engineering (TE). Existing TE methods often struggle to balance optimality with tractability, while recent machine learning based approaches fail to develop reliable strategies across diverse network scenarios. To address these issues, we introduce LO-TE, a novel TE solution integrating deep Learning and Optimization techniques. LO-TE operates in two phases: obtaining an initial solution and refining it to achieve a near-optimal TE solution. Our approach utilizes a scalable graph attention network for finding the necessary flows for refinement, paired with a refining algorithm based on linear programming. We demonstrate the application of LO-TE on three typical TE problems. We evaluate LO-TE on both real-world and self-generated large-scale topologies, demonstrating its strong generalizability across various TE problems and traffic models. The evaluation results indicate that LO-TE is 12x-188x faster than traditional TE optimization methods on large-scale network topologies, with an average performance gap of less than 6% compared to the optimal solution. Moreover, LO-TE outperforms state-of-the-art deep learning-based TE methods using limited training data, achieving only 1.8%-69% maximum link utilization under dynamic traffic conditions.