Composable digital twins predict network latency across changing topologies
Toward Composable Network Digital Twins: A Subgraph-Based Latency Prediction Study
Networking and Internet ArchitectureMachine Learning
Summary
Networks often change their connections and traffic, making it hard to predict how fast data will travel. The authors created a new way to break down a whole network into smaller pieces and build flexible models for each piece. These small models can be combined to predict delays in the entire network, even when things change. Their method works well on both real and synthetic data, and it is more reusable than older models that treat the whole network as one block.
What this means in practice
- •For network operators: Quickly estimate latency impacts of configuration or topology changes without rebuilding full models.
- •For cloud infrastructure teams: Plan network upgrades or traffic shifts by combining reusable subgraph models to predict performance effects.
Authors
Shenjia Ding, David Flynn, Paul Harvey
Abstract
Modern networks must support changing topologies, configurations, and performance objectives, motivating fast and reliable performance estimation. Network digital twins (NDTs) enable what-if analysis for performance estimation in such network scenarios, however, existing machine learning-based NDT approaches often rely on entire topology representations, which are inherently monolithic and lack reusability under topological or traffic changes in the network. This paper introduces a composable NDT approach that decomposes networks into subgraphs represented by reusable unit twins that capture subgraph structure, configuration and traffic behaviours. A lightweight composer aggregates unit twin combinations to create NDTs that predict per-route end-to-end latency through an overall topology. Evaluation across controlled synthetic topologies and diverse traffic scenarios, real-world Topology Zoo topologies, and a public NDT challenge dataset demonstrates that the composable NDTs achieve high in-distribution accuracy while remaining stable under out-of-distribution scenarios. Comparison with monolithic full topology NDTs demonstrates that our composable approach achieves reusability, while achieving comparable or superior accuracy.