How digital twins are transforming network infrastructure: Future state (part 2)

This is the second of a two-part series. Read part 1 about the current state of networking and how digital twins are being used to help automate the process, and the shortcomings involved.

As noted in part 1, digital twins are starting to play a crucial role in automating the process of bringing digital transformation to networking infrastructure. Today, we explore the future state of digital twins – comparing how they’re being used now with how they can be used once the technology matures.

The market for digital twins is expected to grow at a whopping 35% CAGR (compound annual growth rate) between 2022 and 2027, from a valuation of $10.3 billion to $61.5 billion. Internet of things (IoT) devices are driving a large percentage of that growth, and campus networks represent a critical aspect of infrastructure required to support the widespread rollout of the growing number of IoT devices.

Current limitations of digital twins

One of the issues plaguing the use of digital twins today is that network digital twins typically only help model and automate pockets of a network isolated by function, vendors or types of users. However, enterprise requirements for a more flexible and agile networking infrastructure are driving efforts to integrate these pockets.

Several network vendors, such as Forward Networks, Gluware, Intentionet and Keysight’s recent Scalable Networks acquisition, are starting to support digital twins that work across vendors to improve configuration management, security, compliance and performance. 

Companies like Asperitas and Villa Tech are creating “digital twins-as-a-service” to help enterprise operations.

In addition to the challenge of building a digital twin for multivendor networks, there are other limitations that digital twin technology needs to overcome before it’s fully adopted, including:

• The types of models used in digital twins needs to match the actual use case. 
• Building the model, supporting multiple models and evolving the model over time all require significant investment, according to Balaji Venkatraman, VP of product management, DNA, at Cisco.
• Keeping the data lake current with the state of the network. If the digital twin operates on older data, it will return out-of-date answers. 

Future solutions

Minas Tiwari, client partner for cross-industry comms solutions at Capgemini Engineering, believes that digital twins will help roll out disaggregated networks composed of different equipment, topologies and service providers in the same way enterprises now provision services across multiple cloud services. 

Tiwari said digital twins will make it easier to model different network designs up front and then fine-tune them to ensure they work as intended. This will be critical for widespread rollouts in healthcare, factories, warehouses and new IoT businesses. 

Vendors like Gluware, Forward Networks and others are creating real-time digital twins to simulate network, security and automation environments to forecast where problems may arise before these are rolled out. These tools are also starting to plug into continuous integration and continuous deployment (CI/CD) tools to support incremental updates and rollback using existing devops processes.

Cisco has developed tools for what-if analysis, change impact analysis, network dimensioning and capacity planning. These areas are critical for proactive and predictive analysis to prevent network or service downtime or impact user experience adversely.

Overcoming the struggle with new protocols

Early modeling and simulation tools, such as the GNS3 virtual labs, help network engineers understand what is going on in the network in terms of traffic path, connectivity and isolation of network elements. Still, they often struggle with new protocols, domains or scaling to more extensive networks. They also need to simulate the ideal flow of traffic, along with all the ways it could break or that paths could be isolated from the rest of the network. 

Christopher Grammer, vice president of solution technology at IT solutions provider Calian, told VentureBeat that one of the biggest challenges is that real network traffic is random. The network traffic produced by a coffee shop full of casual internet users is a far cry from the needs of petroleum engineers working with real-time drilling operations. Therefore, simulating network performance is subject to the users’ needs, which can change at any time, making it more difficult to actively predict.

Not only that, but, modeling tools are costly to scale up. 

“The cost difference between simulating a relatively simple residential network model and an AT&T internet backbone is astronomical,” Grammer said. 

Thanks to algorithms and hardware improvements, vendors like Forward Enterprise are starting to scale these computations to support networks of hundreds of thousands of devices.

Testing new configurations

The crowning use case for networking digital twins is evaluating different configuration settings before updating or installing new equipment. Digital twins can help assess the likely impact of changes to ensure equipment works as intended. 

In theory, these could eventually make it easier to assess the performance impact of changes. However, Mike Toussaint, senior director analyst at Gartner, said it may take some time to develop new modeling and simulation tools that account for the performance of newer chips.

One of the more exciting aspects is that these modeling and simulation capabilities are now being integrated with IT automation. Ernest Lefner, chief product officer at Gluware, which supports intelligent network process automation, said this allows engineers to connect inline testing and simulation with tools for building, configuring, developing and deploying networks. 

“You can now learn about failures, bugs, and broken capabilities before pushing the button and causing an outage. Merging these key functions with automation builds confidence that the change you make will be right the first time,” he said.

Wireless analysis

Equipment vendors such as Juniper Networks are using artificial intelligence (AI) to incorporate various kinds of telemetry and analytics to automatically capture information about wireless infrastructure to identify the best layout for wireless networks. Ericsson has started using Nvidia Omniverse to simulate 5G reception in a city. Nearmap recently partnered with Digital Twin Sims to create dynamically updated 5G coverage maps into 5G planning and operating systems. 

Security and compliance

Grammer said digital twins could help improve network heuristics and behavioral analysis aspects of network security management. This could help identify potentially unwanted or malicious traffic, such as botnets or ransomware. Security companies often model known good and bad network traffic to teach machine learning algorithms to identify suspicious network traffic. 

According to Lefner, digital twins could model real-time data flows for complex audit and security compliance tasks. 

“It’s exciting to think about taking complex yearly audit tasks for things like PCI compliance and boiling that down to an automated task that can be reviewed daily,” he said. 

Coupling these digital twins with automation could allow a step change in challenging tasks like identifying up-to-date software and remediating newly identified vulnerabilities. For example, Gluware combines modeling, simulation and robotic process automation (RPA) to allow software robots to take actions based on specific network conditions. 

Peyman Kazemian, cofounder of Forward Networks, said they are starting to use digital twins to model network infrastructure. When a new vulnerability is discovered in a particular type of equipment or software version, the digital twins can find all the hosts that are reachable from less trustworthy entry points to prioritize the remediation efforts. 

Cross-domain collaboration

Network digital twins today tend to focus on one particular use case, owing to the complexities of modeling and transforming data across domains. Teresa Tung, cloud first chief technologist at Accenture, said that new knowledge graph techniques are helping to connect the dots. For example, a digital twin of the network can combine models from different domains such as engineering R&D, planning, supply chain, finance and operations. 

They can also bridge workflows between design and simulations. For example, Accenture has enhanced a traditional network planner tool with new 3D data and an RF simulation model to plan 5G rollouts. 

Connect2Fiber is using digital twins to help model its fiber networks to improve operations, maintenance and sales processes. Nearmap’s drone management software automatically inventories wireless infrastructure to improve network planning and collaboration processes with asset digital twins. 

These efforts could all benefit from the kind of innovation driven by building information models (BIM) in the construction industry. Jacob Koshy, information technology and communications associate, Arup, an IT services firm, predicts that comparable network information models (NIM) could have a similarly transformative role in building complex networks. 

For example, the RF propagation analysis and modeling for coverage and capacity planning could be reused during the installation and commissioning of the system. Additionally, integrating the components into a 3D modeling environment could improve collaboration and workflows across facilities and network management teams.

Emerging digital twin APIs from companies like Mapped, Zyter and PassiveLogic might help bridge the gap between dynamic networks and the built environment. This could make it easier to create comprehensive digital twins that include the networking aspects involved in more autonomous business processes. 

The future is autonomous networks

Grammer believes that improved integration between digital twins and automation could help fine-tune network settings based on changing conditions. For example, business traffic may predominate in the daytime and shift to more entertainment traffic in the evening. 

“With these new modeling tools, networks will automatically be able to adapt to application changes switching from a business video conferencing profile to a streaming or gaming profile with ease,” Grammer said. 

How digital twins will optimize network infrastructure

The most common use case for digital twins in network infrastructure is testing and optimizing network equipment configurations. Down the road, they will play a more prominent role in testing and optimizing performance, vetting security and compliance, provisioning wireless networks and rolling out large-scale IoT networks for factories, hospitals and warehouses. 

Experts also expect to see more direct integration into business systems such as enterprise resource planning (ERP) and customer relationship management (CRM) to automate the rollout and management of networks to support new business services.