Semi-Oblivious Traffic Engineering: The Road Not Taken

USENIX Symposium on Networked Systems Design and Implementation (NSDI)


Networks are expected to provide reliable performance under a wide range of operating conditions, but existing traffic engineering (TE) solutions optimize for performance or robustness, but not both. A key factor that impacts the quality of a TE system is the set of paths used to carry traffic. Some systems rely on shortest paths, which leads to excessive congestion in topologies with bottleneck links, while others use paths that minimize congestion, which are brittle and prone to failure. This paper presents a system that uses a set of paths computed using Räcke’s oblivious routing algorithm, as well as a centralized controller to dynamically adapt sending rates. Although oblivious routing and centralized TE have been studied previously in isolation, their combination is novel and powerful. We built a software framework to model TE solutions and conducted extensive experiments across a large number of topologies and scenarios, including the production backbone of a large content provider and an ISP. Our results show that semi-oblivious routing provides near-optimal performance and is far more robust than state-of-the-art systems.

Related Publications

All Publications

TSE - June 29, 2021

Learning From Mistakes: Machine Learning Enhanced Human Expert Effort Estimates

Federica Sarro, Rebecca Moussa, Alessio Petrozziello, Mark Harman

IEEE ICIP - September 19, 2021

Rate Estimation Techniques for Encoder Parallelization

Gaurang Chaudhari, Hsiao-Chiang Chuang, Igor Koba, Hariharan Lalgudi

RecSys - September 27, 2021

Jointly Optimize Capacity, Latency and Engagement in Large-scale Recommendation Systems

Hitesh Khandelwal, Viet Ha-Thuc, Avishek Dutta, Yining Lu, Nan Du, Zhihao Li, Qi Huang

MLSys - June 9, 2021

Value Learning for Throughput Optimization of Deep Neural Networks

Benoit Steiner, Chris Cummins, Horace He, Hugh Leather

To help personalize content, tailor and measure ads, and provide a safer experience, we use cookies. By clicking or navigating the site, you agree to allow our collection of information on and off Facebook through cookies. Learn more, including about available controls: Cookies Policy