How to automate network operations centers through AI

Commercial context

Across the globe, CSPs are rolling out 5G connectivity which employs network function virtualization (NFV) and software defined networking (SDN) to enhance scalability. But they are being held back by complex legacy OSS/BSS tools, which have grown both organically and inorganically through acquisitions, and the need for domain-specific solutions. The result can be severe operational challenges.

In particular, a CSP’s service management layer can be fragmented with disparate tools, processes and interfaces across operational groups. This fragmentation leads to inefficiencies, delays and increased operational costs - the complex architecture means it takes a long time to launch a product, while the order process becomes design-intensive because of the overlapping functionalities in each component. Furthermore, the use of one-off integration definitions requires substantial end-to-end design efforts. As CSPs’ legacy networks evolve to be more IP-based and virtual, their IT stacks need to evolve. To avoid major incidents, CSPs need more proactive service assurance solutions that can predict and resolve potential issues in advance.

The solution

This Catalyst is therefore developing a cloud-native “DarkNOC” framework, using AI, which will enable CSPs to automate their network operations centers (NOCs). The goal is to create a reference architecture for the telecoms industry, which will help CSPs to scale up their network assurance and service management systems easily. As the DarkNOC framework uses standardized definitions of business entities across all platforms, it can provide a consolidated view of inventory, alarms and service management. As a result, operational users will now get all the information they need in one place, while AI can be employed to automate remediation.

The framework uses TM Forum’s Autonomous Network concept and is aligned with the TM Forum ODA (Open Digital Architecture) framework. To provide an accurate snapshot of network inventory, which is critical to building an autonomous network solution, the project is using TMF639 to support a network discovery solution that keeps the service management configuration management database up to date with accurate resource inventory data. By making use of TM Forum Open APIs, the Catalyst aims to simplify the integration process and enable the reuse of each capability, offering greater value from each delivery.

The Catalyst is also looking to use AI to enhance CSPs’ existing domain-specific solutions across fault and performance management, thereby enabling these applications to work in autonomous mode, without any human in loop. More broadly, the project team is using generative AI to help convert business requirements and text into software code, thereby reducing the dependency on IT.

Applications and wider value

The DarkNOC framework is designed to reduce the number of incidents by proactively identifying potential problems. It should also lower MTTR (mean time to resolution) with better impact assessment, assignment and resolution, while cutting the average call handling time. It will also eliminate the ‘swivel chair scenario’ where a user needs to access multiple systems to manage operational tasks.

“Overall this change automation will deliver immense benefit to a telco as an organization and the industry as a whole by freeing up our NOC agents to allow them to focus more on policy decision than driving mundane operational tasks, supporting our end customers in a more proactive manner,” explains Girish Mahajan, Senior Technical Manager at BT. “This should translate to MTTR reductions on reactive incidents and improve change success by removing the need for these agents to be manually making changes directly on the network.”