D-NDT: Distributed Network Digital Twin Architecture - Use Case on 5G Network Slice Management

Network Digital Twins (NDTs) are set to play a crucial role in next-generation networks. However, their deployment in the real world is constrained by high synchronisation overhead and scalability limitations. To address this, while distributed computing is a promising paradigm, it cannot be implemented directly in the NDTs due to their tightly coupled architectures. Therefore, we present Distributed NDT (D-NDT) and decouple synchronisation from model execution. In this, dedicated components handle bidirectional telemetry flows, preprocessing, and aggregation near the physical network, while separate execution components perform model inference using the abstracted, synchronised representation. This separation enables independent scaling, eliminates redundant data collection across domains, and supports distributed deployment. Consequently, we present a use case for 5G network slice management and compare the performance of D-NDT with centralised and partially distributed architectures. The results show that D-NDT reduces synchronisation latency and maintains scalability as NDT complexity increases.