What is a Digital Twin?

What is a Network Digital Twin?

Formally, a digital twin is a "virtual representation of a physical network" (https://www.itu.int/rec/T-REC-Y.3090-202202-I/en). However, in telecoms it is becoming more broadly known as the set of capabilities to automate network operations. Since we're unable to safely and quickly query data directly from the netework in a normal operations environment, a digital twin provides the next best thing - an up-to-date duplicate of the network we can view and interact with.

I find it helpful to think about digital twins as a maturity journey:

1. Digital model: This is a model of the network in a database, often manually updated. We can query this data to get information on the network.
2. Digital shadow: The network data flows one way from the network into our copy.
3. Digital twin: While it is used as a generic term today, a digital twin should give a two way interaction with your network, allowing you to view the information (pull) and make changes to the network (push).
4. Cognitive digital twin: This digital twin runs intelligent use cases on the network such as root cause analysis or network planning.
5. Autonomous digital twin: Fully automated use cases are run on the network such as self healing, auto-remediation and other AI-enabled use cases.

Every org

Read more about digital twins

I have a deep dive (personally written, not AI-generated!) on my website.

In other words, a digital twin is an up-to-date representation of your network. "Representation" can mean multiple things - from an interactive point & click dashboard to a database. The important thing is that a digital twin has a recent set of your network data you need to perform whatever task you need.

We should note, digital twin today encompasses a category of functionality, rather than a single definition.

Why can't we get the network data directly from the network?

Ideally, every application would be able to query live data directly from the network whenever it wants. However, there are a couple of challenges:

• Every query on the network puts additional strain on the network. For critical infrastructure, this is too risky.
• For the previous point, network management systems don't allow direct querying of the network.
• Every integration between the NMS and an application is another security risk. CSPs want to limit access to their network as much as possible.
• CSP's do not want unrestricted access to sensitive network information. Their users phone numbers, internet behaviour, names, addresses and so forth are all highly sensitive information.

Where do we get the network data from then?

So, if we can't access the data directly, then we need a way of accessing the most up-to-date version of the network. Orgs today will export their network data on a periodic basis and store it in a database. For example, many companies will export their network data once per day over 15 minute intervals. This data can be stores in a file system or a database for future use. If we wanted to build a digital twin, we would put the data into a database so we could read and analyze the network from that point in time.

Why is there debate about what a digital twin should do?

I think the confusion abounds from each organisation having a different set of requirements from a digital twin. One company wants to view the alarms on their network, so they need alarm data and network data together. Another company wants to plan an extension to their network, so they need network topology with performance and configuration data to simulate what an extension would look like.

We can map these requirements onto a digital twin evolution chart.

Short version

What is a digital twin?

What are the capabilities of a digital twin?

• Digital model:
• Digital shadow:
• Digital twin:
• Cognitive digital twin:
• Federated digital twin:
• Autonomous digital twin: