The end-to-end performance delivered by a network is impacted by a diverse set of dynamic operational attributes – traffic types and volume, message priorities, and security policies to name a few. The present generation of network simulation tools has demonstrated the ability to provide accurate, at scale network models to predict such end-to-end performance at the application level, such that post-deployment surprises are minimized.
As communication and networks become ubiquitous and mission-critical, the next frontier for network models is embedded network digital twins. There are two primary differences between an embedded network digital twin and offline network models:
- First, the embedded network digital twin is interfaced with an external system such that it can directly reflect the impact of any changes to the network operations on the end-to-end performance of the external system.
- Second, the embedded twin is configured to periodically update its state, as represented by key operational attributes, in real-time. Such updates may be provided manually, come from the external system, or the embedded twin may be interfaced directly with the corresponding physical network.
Embedded network digital twins expand the application of Model Based Systems Engineering (MBSE) by using the embedded network twins to dynamically improve the quality of operational decisions for networked systems and system-of-systems in real-time, including mitigating the impact of developing cyber threats.
Figure 1: Embedded network digital twin
As shown in the above figure, the embedded digital twin may optionally be interfaced directly with its physical counterpart, the live network, such that changes to specific network state (e.g. channel quality or device configuration at an access point) are periodically and automatically uploaded to the digital twin via appropriate APIs. In some cases, the information transfer may be bi-directional, where specific network or device configurations are downloaded to the physical network by its digital twin to improve operational performance in real-time. When significant degradations or operational inefficiencies are observed or anticipated by the digital twin in the performance of the network and/or the external system, alternative mitigations can be investigated using the digital twin, and potentially deployed back to the physical network.
A key attribute of an embedded network digital twin is that it is interfaced with an external system. The external system may be another simulator (e.g. power system or wargaming simulator), physical devices (e.g. sensor network), application (e.g. situation awareness or video conference) or a combination of these systems; the goal of the embedded digital twin is to continuously compute the specific network metrics (e.g. end-to-end latency) based on the current or projected future state of the network and to predict the resulting impact on the performance of the external system. We note that for many use cases, it may be feasible to host the embedded twin remotely, perhaps in the cloud.
As a simple example, an embedded network twin might be used to predict the impact of hot spots within a 5G campus area network on the quality of a video application, where condensed traffic data from the physical network is periodically uploaded to the digital twin; or alternately, the impact of a mobile jammer on video being transmitted by a sensor on a military training mission, where the current trajectory and signal strength of the jammer together with the Blue Force priority traffic is periodically uploaded to the digital twin. The predictions from the digital twin may be used to deploy mitigations that include, for instance, to temporarily reroute or delay lower priority traffic. We note that some operational decisions may be automated, while others may require human intervention.
SCALABLE is investing continuously to both advance the network digital twin technologies and the ability to interface the digital twins with an expanding set of external systems. An important application area for the embedded network digital twin is to accurately anticipate the impact of dynamic cyber threats on the external systems and deploy relevant mitigations in real-time to reduce their severity. In future blogs, we will present a diverse set of use cases for embedded network digital twins across both commercial and military applications, to include:
- Military wargaming
- Network security
- Critical infrastructure applications
- 5G deployment