Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deeper and Wider Networks for Performance Metrics Prediction in Communication Networks

Aijia Liu, Shiqing Liu, Xiaobing Pei

TL;DR

The paper addresses end-to-end latency and jitter prediction in Digital Twin Networks (DTNs) by introducing DWNet, a deeper and wider heterogeneous GNN. DWNet models networks with two path-state components $h_p$ and $\tilde{h}_p$, along with a link state $h_l$, and uses a GRU-based message passing scheme to capture both direct path effects and interactions among paths sharing edges. It constructs a two-type heterogeneous graph with path and link nodes and employs a dedicated feature fusion readout $F_p([h_p^T,\tilde{h}_p^T])$ to predict path KPIs, while $h_l^T$ informs link-level outputs. Experiments on NSFnet, Geant, and Synth topologies from an OMNET++ dataset show that DWNet achieves lower MAE/MAPE and higher PCC than RouteNet, especially in unseen topologies, demonstrating stronger generalization. This work advances DTN modeling by enabling more accurate performance predictions, which can improve network planning and autonomous optimization in real-world networks.

Abstract

In today's era, users have increasingly high expectations regarding the performance and efficiency of communication networks. Network operators aspire to achieve efficient network planning, operation, and optimization through Digital Twin Networks (DTN). The effectiveness of DTN heavily relies on the network model, with graph neural networks (GNN) playing a crucial role in network modeling. However, existing network modeling methods still lack a comprehensive understanding of communication networks. In this paper, we propose DWNet (Deeper and Wider Networks), a heterogeneous graph neural network modeling method based on data-driven approaches that aims to address end-to-end latency and jitter prediction in network models. This method stands out due to two distinctive features: firstly, it introduces deeper levels of state participation in the message passing process; secondly, it extensively integrates relevant features during the feature fusion process. Through experimental validation and evaluation, our model achieves higher prediction accuracy compared to previous research achievements, particularly when dealing with unseen network topologies during model training. Our model not only provides more accurate predictions but also demonstrates stronger generalization capabilities across diverse topological structures.

Deeper and Wider Networks for Performance Metrics Prediction in Communication Networks

TL;DR

The paper addresses end-to-end latency and jitter prediction in Digital Twin Networks (DTNs) by introducing DWNet, a deeper and wider heterogeneous GNN. DWNet models networks with two path-state components and , along with a link state , and uses a GRU-based message passing scheme to capture both direct path effects and interactions among paths sharing edges. It constructs a two-type heterogeneous graph with path and link nodes and employs a dedicated feature fusion readout to predict path KPIs, while informs link-level outputs. Experiments on NSFnet, Geant, and Synth topologies from an OMNET++ dataset show that DWNet achieves lower MAE/MAPE and higher PCC than RouteNet, especially in unseen topologies, demonstrating stronger generalization. This work advances DTN modeling by enabling more accurate performance predictions, which can improve network planning and autonomous optimization in real-world networks.

Abstract

In today's era, users have increasingly high expectations regarding the performance and efficiency of communication networks. Network operators aspire to achieve efficient network planning, operation, and optimization through Digital Twin Networks (DTN). The effectiveness of DTN heavily relies on the network model, with graph neural networks (GNN) playing a crucial role in network modeling. However, existing network modeling methods still lack a comprehensive understanding of communication networks. In this paper, we propose DWNet (Deeper and Wider Networks), a heterogeneous graph neural network modeling method based on data-driven approaches that aims to address end-to-end latency and jitter prediction in network models. This method stands out due to two distinctive features: firstly, it introduces deeper levels of state participation in the message passing process; secondly, it extensively integrates relevant features during the feature fusion process. Through experimental validation and evaluation, our model achieves higher prediction accuracy compared to previous research achievements, particularly when dealing with unseen network topologies during model training. Our model not only provides more accurate predictions but also demonstrates stronger generalization capabilities across diverse topological structures.
Paper Structure (15 sections, 3 equations, 7 figures, 1 algorithm)

This paper contains 15 sections, 3 equations, 7 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. Problem Formulation
  4. Graph Neural Networks (GNN)
  5. RouteNet Network Modeling Method
  6. DWNet Modeling Method
  7. Heterogeneous Network Modeling
  8. Message Passing Procedure
  9. Feature Fusion Process
  10. Algorithm Description and Analysis
  11. Experiment and Analysis
  12. DataSets and Parameter Setting
  13. Evaluation Indicators
  14. Prediction Outcome Evaluation
  15. Conclusion

Figures (7)

  • Figure 1: Communication network.
  • Figure 2: Heterogeneous diagram in GNN.
  • Figure 3: Message passing and status updates for the model.
  • Figure 4: Comparison of forecast delay statistics in first scenario.
  • Figure 5: Comparison of forecast jitter statistics in first scenario.
  • ...and 2 more figures