Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Identification of Path Congestion Status for Network Performance Tomography using Deep Spatial-Temporal Learning

Chengze Du, Zhiwei Yu, Xiangyu Wang

TL;DR

This work tackles the challenge of inferring internal link congestion from end-to-end measurements by introducing Addictive Congestion Status (ACS), a three-class qualitative and a quantitative congestion metric for paths. It couples an Adversarial Autoencoder with Long Short-Term Memory networks (AAE-LSTM) to learn spatio-temporal representations of probing data and align them with the ACS distribution, enabling accurate classification of path congestion and estimation of the number of congested links. The approach is validated on real network topologies, showing improved recall, precision, and F1 scores for congested-link localization and lower NRMSE for link-performance inference, especially when ACS information is correctly used to constrain the solution space. The results demonstrate that leveraging ACS with deep learning yields more accurate network tomography while reducing probing load, offering a practical tool for robust network performance assessment in large and complex networks.

Abstract

Network tomography plays a crucial role in assessing the operational status of internal links within networks through end-to-end path-level measurements, independently of cooperation from the network infrastructure. However, the accuracy of performance inference in internal network links heavily relies on comprehensive end-to-end path performance data. Most network tomography algorithms employ conventional threshold-based methods to identify congestion along paths, while these methods encounter limitations stemming from network complexities, resulting in inaccuracies such as misidentifying abnormal links and overlooking congestion attacks, thereby impeding algorithm performance. This paper introduces the concept of Additive Congestion Status to address these challenges effectively. Using a framework that combines Adversarial Autoencoders (AAE) with Long Short-Term Memory (LSTM) networks, this approach robustly categorizes (as uncongested, single-congested, or multiple-congested) and quantifies (regarding the number of congested links) the Additive Congestion Status. Leveraging prior path information and capturing spatio-temporal characteristics of probing flows, this method significantly enhances the localization of congested links and the inference of link performance compared to conventional network tomography algorithms, as demonstrated through experimental evaluations.

Identification of Path Congestion Status for Network Performance Tomography using Deep Spatial-Temporal Learning

TL;DR

This work tackles the challenge of inferring internal link congestion from end-to-end measurements by introducing Addictive Congestion Status (ACS), a three-class qualitative and a quantitative congestion metric for paths. It couples an Adversarial Autoencoder with Long Short-Term Memory networks (AAE-LSTM) to learn spatio-temporal representations of probing data and align them with the ACS distribution, enabling accurate classification of path congestion and estimation of the number of congested links. The approach is validated on real network topologies, showing improved recall, precision, and F1 scores for congested-link localization and lower NRMSE for link-performance inference, especially when ACS information is correctly used to constrain the solution space. The results demonstrate that leveraging ACS with deep learning yields more accurate network tomography while reducing probing load, offering a practical tool for robust network performance assessment in large and complex networks.

Abstract

Network tomography plays a crucial role in assessing the operational status of internal links within networks through end-to-end path-level measurements, independently of cooperation from the network infrastructure. However, the accuracy of performance inference in internal network links heavily relies on comprehensive end-to-end path performance data. Most network tomography algorithms employ conventional threshold-based methods to identify congestion along paths, while these methods encounter limitations stemming from network complexities, resulting in inaccuracies such as misidentifying abnormal links and overlooking congestion attacks, thereby impeding algorithm performance. This paper introduces the concept of Additive Congestion Status to address these challenges effectively. Using a framework that combines Adversarial Autoencoders (AAE) with Long Short-Term Memory (LSTM) networks, this approach robustly categorizes (as uncongested, single-congested, or multiple-congested) and quantifies (regarding the number of congested links) the Additive Congestion Status. Leveraging prior path information and capturing spatio-temporal characteristics of probing flows, this method significantly enhances the localization of congested links and the inference of link performance compared to conventional network tomography algorithms, as demonstrated through experimental evaluations.

Paper Structure

This paper contains 20 sections, 1 theorem, 7 equations, 12 figures, 2 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. PRELIMINARIES
  4. Network Model
  5. Probing Flow
  6. METHODOLOGY
  7. Addictive Congestion Status
  8. Long Short-Term Memory Network
  9. Adversarial Autoencoder
  10. Reconstruction Process
  11. Adversarial Process
  12. Numerical Evaluations
  13. Network Setups
  14. Evaluations of ACS
  15. Optimizing Probing Flow
  16. ...and 5 more sections

Key Result

Proposition 1

$\forall \mathbf{a}, \exists \mathbf{b}^{\prime}$, such that $d\left(\mathbf{a}, \mathbf{b}^{\prime}\right)<d(\mathbf{a}, \mathbf{b}) \Longrightarrow \mathcal{M}\left(\mathbf{a}, \mathbf{b}^{\prime}\right) \geq \mathcal{M}(\mathbf{a}, \mathbf{b})$

Figures (12)

  • Figure 1:
  • Figure 2:
  • Figure 3:
  • Figure 5: Illustration of the probing flow.
  • Figure 6:
  • ...and 7 more figures

Theorems & Definitions (1)

  • Proposition 1