Open World Learning Graph Convolution for Latency Estimation in Routing Networks

TL;DR

This paper addresses the challenge of estimating SDN routing latency in open-world settings where network sizes, configurations, and traffic patterns vary beyond training distributions. It introduces a domain-knowledge–driven approach that reformulates the routing problem on a directed line graph and employs a combination of Directed Spectral Graph Convolution, NALU-based embedding/readout for robust extrapolation, and Routing Role Recognition attention to capture long-range dependencies. The method demonstrates improved accuracy, faster inference, and stronger generalization to unseen graph sizes and routing configurations compared to state-of-the-art RouteNet variants and graph models. The work advances practical open-world learning for network latency estimation, with implications for scalable SDN monitoring and control, while identifying future work on dynamic, bursty traffic scenarios.

Abstract

Accurate routing network status estimation is a key component in Software Defined Networking. However, existing deep-learning-based methods for modeling network routing are not able to extrapolate towards unseen feature distributions. Nor are they able to handle scaled and drifted network attributes in test sets that include open-world inputs. To deal with these challenges, we propose a novel approach for modeling network routing, using Graph Neural Networks. Our method can also be used for network-latency estimation. Supported by a domain-knowledge-assisted graph formulation, our model shares a stable performance across different network sizes and configurations of routing networks, while at the same time being able to extrapolate towards unseen sizes, configurations, and user behavior. We show that our model outperforms most conventional deep-learning-based models, in terms of prediction accuracy, computational resources, inference speed, as well as ability to generalize towards open-world input.

Figures (6)

• Figure 1: Schematic Representation of SDN Network State Estimation Problem. Directed line denotes network traffic flows with different throughput in the given topology, where one can see that there exist different distributions of line type between training and validation/test time, as mentioned in $(i)$. Undirected line denote the link communicating pair of network elements, with the line's strength denoting the link's capacity level.
• Figure 2: Comparative Example of the Problem Formulation, with Ground Truth being Red. Left: Previous works' formulation. Right: The proposed formulation
• Figure 3: (a) Pre-processing: Incorporate OD Information in Node's Local Structural Feature; (b) Learning and Inference: Extract Neighboring Impact and Long Range Dependencies
• Figure 4: Model Generalization Ability Towards Open-World Input
• Figure 5: Ablation Study of the Proposed Model w.r.t NALU and MLP Embedding & Readout Function