Prior to the COVID-19 pandemic, digital twins were a somewhat esoteric emerging technology that was being used to, for example, enable engineers to maintain aircraft engines more efficiently. With the arrival of the pandemic, however, the level of interest in applying digital twin technologies to the management of various aspects of IT increased. IT teams, like everyone else, were required to work from home. The ability to remotely manage an IT environment using a digital twin went from being an intriguing idea to something that would enable IT teams to work anywhere more easily.
What is a Digital Twin?
One of the areas where digital twins are starting to be applied is network management. A digital twin is a virtual representation of any object or system. It is typically updated in real-time using data collected from sensors that feed a simulation based on an artificial intelligence (AI) model created using machine learning algorithms. A digital twin on any complex system can be created, including an extended networking environment or even a single router or switch.
Benefits of applying digital twin technologies to network management include being able to adjust performance by running what-if simulation before implementing a change that might, hopefully, reduce costs or improve performance to onboarding network managers that can now visualize the interdependencies that exist between all the components that make up an extended network.
Additionally, cybersecurity teams in the wake of a breach can employ that same platform to better identify how malware might be laterally moving across that network environment.
Networking with Digital Twins
Building a digital twin of a network today is not all that easy. A network digital twin needs to reflect configuration, topology and traffic load at a sufficient level of fidelity to accurately reproduce the behavior of the physical twin. The digital twin of a network also needs to integrate with a wide range of software that has typically been deployed as an overlay on top of routers and switches as well as firmware that is embedded in those same devices.
Finally, data generated by those devices needs to flow across a network in real time to the location where the digital twin has been deployed. Given the amount of data that needs to be processed by a digital twin they are typically deployed on a public cloud.
Fortunately, vendors are starting to make pre-configured models of network components available. Those models can then be customized to create a digital twin made up of models provided by different vendors. Most providers of networking platforms are building digital twins of their own platforms that will be employed within the context of a network management platform they provide. However, over time IT organizations will be able to combine models from multiple vendors within a heterogeneous network management framework.
Broadening Digital Twin Technologies
The Digital Twin Consortium, which counts Microsoft, Dell Technologies, Autodesk, and GE Digital among its founding members, earlier this year announced an alliance with LF Edge, an umbrella organization within the Linux Foundation that is focused on an open, interoperable framework for edge computing platforms.
The Digital Twin Consortium plans to leverage EdgeX Foundry, an open source, loosely coupled microservices framework being advanced by LF Edge, to collect data. The goal is to collaborate on open-source projects that facilitate the implementation and consumption of a reference architecture defined by the Digital Twin Consortium platform. Microsoft is also trying to drive adoption of a Digital Twin Definition Language (DTDL) for building models.
Last year the Digital Twin Consortium formed a similar alliance with the Industrial Internet Consortium (IIC) to accelerate the development, adoption, and monetization of digital twin technologies.
AI and Digital Twins
It’s still early days as far as digital twin technologies being applied to manage networks is concerned. Very few network managers today have access to a digital twin of their networking environment. However, as IT organizations invest more in AI to manage IT operations (AIOps) the more common digital twins will become. As AIOps continues to evolve, machine learning algorithms will crawl all over extended IT environments. The digital twin models of an IT environment are likely to incorporate data collected from applications, servers, and storage systems as well.
“I think it will be broad into the fabric of AIOps,” says Mitchell Ashley, CEO and principal analyst for the research firm Accelerated Strategies Group. “It will be a specialty arm.”
One way or another digital twins and other AI technologies will be applied to network management. In some cases, it will be driven by networking specialists. In other cases, adoption of these technologies will be driven more broadly from the top down by IT leaders as the number of applications and platforms that need to be managed continue to expand beyond the capabilities of any IT team to manage on their own. Regardless of approach, the way networks are managed is about to change.
Read next: AIOps Trends & Benefits for 2021