Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FIGRET: Fine-Grained Robustness-Enhanced Traffic Engineering

Ximeng Liu, Shizhen Zhao, Yong Cui, Xinbing Wang

TL;DR

Traffic Engineering faces a fundamental tension between normal-case performance and robustness to traffic bursts. FIGRET addresses this by applying fine-grained robustness that is customized per source-destination pair through a burst-aware loss and path-sensitivity constraints, learned in an end-to-end fashion. The method demonstrates substantial improvements in $MLU$ and solution speed over Jupiter and DOTE across WAN and DC topologies, and provides resilience to failures and traffic drift with limited retraining. This work offers a practical, scalable approach to robust TE that can adapt to diverse traffic characteristics while maintaining high utilization and responsiveness.

Abstract

Traffic Engineering (TE) is critical for improving network performance and reliability. A key challenge in TE is the management of sudden traffic bursts. Existing TE schemes either do not handle traffic bursts or uniformly guard against traffic bursts, thereby facing difficulties in achieving a balance between normal-case performance and burst-case performance. To address this issue, we introduce FIGRET, a Fine-Grained Robustness-Enhanced TE scheme. FIGRET offers a novel approach to TE by providing varying levels of robustness enhancements, customized according to the distinct traffic characteristics of various source-destination pairs. By leveraging a burst-aware loss function and deep learning techniques, FIGRET is capable of generating high-quality TE solutions efficiently. Our evaluations of real-world production networks, including Wide Area Networks and data centers, demonstrate that FIGRET significantly outperforms existing TE schemes. Compared to the TE scheme currently deployed in Jupiter data center networks of Google, FIGRET achieves a 9\%-34\% reduction in average Maximum Link Utilization and improves solution speed by $35\times$-$1800 \times$. Against DOTE, a state-of-the-art deep learning-based TE method, FIGRET substantially lowers the occurrence of significant congestion events triggered by traffic bursts by 41\%-53.9\% in topologies with high traffic dynamics.

FIGRET: Fine-Grained Robustness-Enhanced Traffic Engineering

TL;DR

Traffic Engineering faces a fundamental tension between normal-case performance and robustness to traffic bursts. FIGRET addresses this by applying fine-grained robustness that is customized per source-destination pair through a burst-aware loss and path-sensitivity constraints, learned in an end-to-end fashion. The method demonstrates substantial improvements in and solution speed over Jupiter and DOTE across WAN and DC topologies, and provides resilience to failures and traffic drift with limited retraining. This work offers a practical, scalable approach to robust TE that can adapt to diverse traffic characteristics while maintaining high utilization and responsiveness.

Abstract

Traffic Engineering (TE) is critical for improving network performance and reliability. A key challenge in TE is the management of sudden traffic bursts. Existing TE schemes either do not handle traffic bursts or uniformly guard against traffic bursts, thereby facing difficulties in achieving a balance between normal-case performance and burst-case performance. To address this issue, we introduce FIGRET, a Fine-Grained Robustness-Enhanced TE scheme. FIGRET offers a novel approach to TE by providing varying levels of robustness enhancements, customized according to the distinct traffic characteristics of various source-destination pairs. By leveraging a burst-aware loss function and deep learning techniques, FIGRET is capable of generating high-quality TE solutions efficiently. Our evaluations of real-world production networks, including Wide Area Networks and data centers, demonstrate that FIGRET significantly outperforms existing TE schemes. Compared to the TE scheme currently deployed in Jupiter data center networks of Google, FIGRET achieves a 9\%-34\% reduction in average Maximum Link Utilization and improves solution speed by -. Against DOTE, a state-of-the-art deep learning-based TE method, FIGRET substantially lowers the occurrence of significant congestion events triggered by traffic bursts by 41\%-53.9\% in topologies with high traffic dynamics.
Paper Structure (41 sections, 10 equations, 20 figures, 8 tables, 1 algorithm)

This paper contains 41 sections, 10 equations, 20 figures, 8 tables, 1 algorithm.

Table of Contents

  1. introduction
  2. motivation and key insights
  3. Necessity of managing bursts
  4. Diversity in traffic characteristics
  5. Dig deep into the trade-off
  6. TE model
  7. FIGRET design
  8. Design for managing traffic bursts
  9. FIGRET framework
  10. How to compute a TE solution
  11. Leveraging deep learning to implement the end-to-end method
  12. Design loss function of FIGRET
  13. Loss for MLU
  14. Loss for fine-grained robustness
  15. Design DNN structure of FIGRET
  16. ...and 26 more sections

Figures (20)

  • Figure 1: Comparing the impact of using anti-burst Hedging mechanism on maximum link utilization (MLU), with MLU values normalized to the maximum MLU.
  • Figure 2: The variance of traffic demand by source and destination (The value of variance has been normalized). The larger the variance, the more likely it is that the traffic demand for that source-destination pair will experience a burst. Regardless of which network in \ref{['fig:GEANT']}, \ref{['fig:Facebook_pod_a']}, or \ref{['fig:Facebook_tor_a']}, the traffic demands of different source-destination pairs exhibit different traffic characteristics.
  • Figure 3: A simple example comparing three TE schemes. TE scheme 1 has the optimal MLU in the normal situation. When traffic burst situations 1/2/3 are all possible, TE scheme 2 exhibits the best resilience against bursts. If only traffic burst situation 3 is likely to occur, TE scheme 3 outperforms TE scheme 2.
  • Figure 4: Cosine similarity analysis using a window of 12 historical TMs vs. the currently-seen TM.
  • Figure 5: Performance of FIGRET and baselines under the objective of minimizing max link utilization (MLU). The y-axis value represents the MLU normalized by that of the Omniscient TE.
  • ...and 15 more figures