AI and automation in private networks

Private network operators need to be able to deploy, manage and optimise their operations more efficiently if they want to meet customer expectations and scale quickly. Operators will rely on AI and service automation to monitor the network closely and get a better understanding of how different components react during their lifecycle and therefore make better decisions and avoid future faults.

Private network operators might face several challenges in maintaining network quality of service and experience at all times. In previous work by STL Partners, we identified three areas where AI and automation can help in mobile networks. These are fault detection, prediction and resolution; network optimisation; and network planning and upgrades. These three use cases also apply to private networks, where specific deployment and management challenges arise from the isolated and the complex nature of the networks.

Management and deployment challenges of private networks

For enterprise customers managing mission critical applications, the expectations for the network availability exceed those for the public networks by a large margin. Depending on the application, the enterprise might be looking to guarantee availability of at least 99.9999% (less than 31.5 seconds of downtime per year). With critical communications, even a slight degradation in network quality can result in serious shut down of several applications. One of the main reasons for deploying and investing in private networks is to provide a consistent level of service in terms of optimised coverage, capacity, security or latency that cannot be guaranteed by public mobile networks. Moreover, the threshold for mission critical and business critical applications might evolve, as the network capabilities evolve. Many data and latency demanding use cases such as real time video analytics will become a default mission critical application in different industries.

As the number of networks increase the complexity of managing them will increase. A single telco or a private networks provider could be operating different private networks for different types of customers with various customised KPIs, which makes management fundamentally difficult. They need a way to coordinate between these networks and limit the management challenges. Adding to the complexity, private networks have different deployment models ranging from a complete isolation in standalone networks to fully integrated with public networks as in network slicing. In addition, they could be supported by different interconnected network structures and technologies, including distributed multi-cloud and multi-edge computing. Guaranteeing availability and reliability in addition to security and interoperability across these complex scenarios will be challenging. Both traditional and new telcos need to be able to automate deployment and management operation in order to scale with ease and manage the number and volume of networks as they grow without overwhelming and draining their resource and team capacity.

How AI and automation can help

Technology providers anticipate that AI and automation can help accelerate deployments and simplify lifecycle management across all stages through:

Lowering deployment cost. Enterprises and private network providers are not currently utilising analytics to deploy network infrastructure cost effectively. They might be overbuilding for resilience just like what telcos did with macro networks in the past. Using AI in the planning stage can help tackle this and lower the deployment cost. It would be very helpful for providers adopting the Opex model (i.e. having customers paying for the network service rather than pay for the infrastructure)

Automation of networking testing orchestration and KPI measurements. The complexity of the network will require frequent testing for various parameters including varying levels of reliability, coverage, security and latency for different users and use cases. Automation engines can help simplify that in addition to triggering certain measurements in response to system alerts.

Early detection of signs of service degradation across the network layers is key to uphold an acceptable level of service all the time. The use of advanced analytics and machine learning engines is necessary for accurate real time predictions of the network performance based on real-time information.

Added layers of granularity to tracked and monitored telemetry data within the network or the slice. The network operator should leverage tools that can identify the importance of these different elements and the impact they will have on the wider network users at any given time. Tools should be able to prioritise critical situations and adapt responses based on the changing requirements.

Examples of technology vendors solutions

Many security and service assurance vendors are starting to tackle private networks requirements in their solutions.

• Providers such as EXFO and Radcom are integrating AI/ML capabilities to assist with private 5G monitoring and service assurance.
• In April 2021, Empirix (acquired by Infovista) also launched an AI/ML based platform dedicated to 5G SA monitoring and assurance.
• Rohde & Schwarz is setting up a private network in Germany to develop and validate its own network testing and monitoring solution for 5G.
• Zeetta Network which spun out of the University of Bristol’s High Performance Networks Group is building on the group’s research in network virtualisation and network slicing to offer software solutions and automation tools for network optimisation and management for both enterprises and service providers for private network management.
• ADVA has launched AI-NET-PROTECT project in September 2021 which is a consortium of industry and research partners working together to develop an AI-based management and network security solution leveraging real-time telemetry. The initiative aims to drive automated resilience and security for private networks and critical infrastructure.

Service assurance and maintaining SLAs will be a challenging role for telcos

For traditional telcos that are also managing public networks, the move towards more distributed and virtualised operations such as in private networks and edge computing will introduce new challenges in network management and service assurance. The emergence of network slicing as a means of customising network services, in addition to providing dedicated private networks, will further complicate the management and security needs for telcos competing in the private networking market. Whether network slices will enable telcos to offer similar levels of SLAs to that of dedicated private networks is far from clear. However, telcos will still need to guarantee higher quality of service, reliability and security for every slice than what public networks can offer.

As seen in Figure 1, the network slice lifecycle resembles that of the any network. Ultimately network slices need to be dynamic in how they are created, launched and operated as they will be shared by multiple users. This means that telcos should have real-time monitoring for every slice and the public network capacity.

Figure 1: Automation across all parts of a network slice lifecycle

Telcos need to understand and demonstrate clearly what aspects of the network performance they can guarantee and work to create value and differentiate their offerings around them while continuing to develop other KPI measurement and management tools for differentiated granularity.

Integration of automation and AI into private networking solutions will be particularly important for operators aiming to deliver in-life management and orchestration for their customers, whether through a slice or in dedicated networks.  Ultimately, this is far from most operators’ core expertise, so we expect to see increased partnerships between operators and service assurance companies such as those above to deliver cost effective and efficient network management and ensure SLAs are met.