Automated Digital Twin Generation for Network Testing: A Multi-Topology Validation (to appear)

Abstract

With the exponential growth of data traffic, projected to nearly triple to 9.7 million petabytes (PB) by 2027 within a global communication market expected to reach USD 2,874.76 billion by 2030, ensuring reliable and efficient network testing has become critical throughout the design, implementation, and management operations.

As testing becomes essential to ensure that changes in configuration or traffic conditions do not degrade user experience, current testing practices rely heavily on manual configuration and simulators. This reliance lead to time-consuming, difficult-to-scale, and expert-dependent processes. To address these limitations, our work explores the role of Automated Machine Learning (AutoML)–based automatically generated Digital Twin (DT) in network testing to enable rapid and scalable testing across diverse network conditions. By integrating this approach with a network service controller for configuration optimization, our results evidence an improvement that DT-enabled testing achieves high accuracy while being approximately 25,000 times faster than simulator-based testing. The implications of these findings, suggest that automated DT generation through AutoML can reduce dependence on manual modeling, allow DTs to adapt to diverse test scenarios, and enhance scalability for complex network.