Dr. Kireeti Kompella, Senior Vice President and Chief Engineer WAN BU, Juniper Networks
A digital twin, a virtual clone of a physical process or object, is not a new concept. Institutions and organisations (including NASA, which coined the term) have used digital twins for decades. Digital twins can generate everything from models that emulate the harsh environments encountered in space through simulations of bulldozers for training purposes, but in a safe setting where operations can be carried out without physical harm.
Although digital twins have been envisioned for years, recent AI-powered advances in data capture, analysis, and automation are finally realizing their full potential—and the possible benefits are virtually limitless.
This is especially true in the field of networking, and within that, for sectors as critically important as research and education.
Understanding network digital twins
From the notion of a digital twin as a virtual clone of a physical object, that of a network digital twin network follows naturally: a virtual replica of an entire network, everything from the physical asset andthe individual devices and users that exist within it – routers, switches and optical gear (or the underlay), their connections, the running software, configurations and associated management systems.
While not yet a full reality, the benefits of network digital twins are numerous, and their use cases are compelling. By creating a network digital twin, network architects and operations personnel can work with sandbox environments to educate, experiment, run simulations and diagnose real-world problems with far less associated risks compared to operating within a production physical network environment, where any glitches immediately impact user experiences.
For a traditionally cautious industry such as networking, there are clear benefits to this approach. Within networking, too much interfering with the production network can lead to significant network downtime and user disruption which is especially precarious across sectors managing critical national infrastructure or expensive, hard-to-reproduce science experiments.
A natural extension is the digital “experience” twin, which simulates user or device behaviour—either in a sandboxed twin environment or directly on the live network. For instance, automating a Wi-Fi client’s connection and authentication sequence can immediately reveal a misconfigured DHCP server. By running these simulations after network changes—ideally during off-hours when employees, students or shoppers are absent—teams can proactively detect and resolve end-user issues before they ever hit production.
How educators and researchers can benefit from network digital twins
While still in its infancy, network digital twins could have far-reaching benefits for the education and research sector. This includes driving efficiencies in the underlay network’s operations thanks to enhanced real-time automation enabling faster, more efficient network maintenance (with far less downtime for users), and helping researchers optimise their modelling and run tasks in safer virtual environments.
For instance, students could learn to run projects on network digital twins and experiment with new ideas like “power steering” – traffic engineering to reduce power utilisation in the network – with no impact on the production network. Faculties in Computer Networking would be able to create clones of different aspects of the production network, scale them, and configure them differently depending on research topics to offer students a more realistic yet simulated network with which to play.
Imagine, say, campuses and research facilities spread across numerous countries. A researcher at one location wants to conduct an experiment at a research facility several hundred kilometres away. Traditionally this would require the researcher to initiate a remote session, connecting across multiple entities, and the physical network would provide the basic transport path to enable this connectivity to the research facility and associated data transfer capabilities. With a network digital twin, the researcher would first simulate this connection to the offsite research hub, validate that it meets their QoS and resiliency requirements and ensure that it functions correctly – knowing full well that any changes they made to the clone would not have any real-world impact. Once satisfied, they can proceed to make the connection in the production network.
Exploring the AI capabilities of network digital twins
There is also the potential that the network digital twin would be able to run simulations much faster than the physical network, by “speeding up time”. This would result in a higher volume of data that machine learning algorithms can use – allowing them to analyse, anticipate, predict trends, find anomalies and fine tune responses for more accurate “closed loop operation”. This could have significant impact on research and learning, accelerating the development of new skills, technologies and understanding of network operations.
By leveraging AI and ML, network digital twins shift operations from device-centric management—routers, switches, firewalls—to holistic, end-to-end user-experience optimization. This AI-driven approach simultaneously maximizes network uptime and elevates user satisfaction.There is still much development needed to realise cost-effective, scalable network digital twins. But we are on an interesting path toward realising its potential. And for the education and research sector in particular this is a very exciting prospect.
For further insights, please attend Kireeti’s presentation at TNC25, taking place in the Exchange Theatre on Tuesday, 10 June.
Dr. Kompella is senior vice president and chief engineer for the WAN at Juniper Networks, He has deep experience in Packet Transport, large-scale MPLS, VPNs, VPLS and Layer 1 to Layer 3 networking, and is very active in the IETF as author of several Internet Drafts and RFCs.
Read the full online magazine here
