Towards Cognitive Autonomous Networks

10Cognitive Autonomy in Cross‐Domain Network Analytics

Szabolcs Nováczki, Péter Szilágyi², and Csaba Vulkán

Nokia Bell Labs, Budapest, Hungary

Telecommunication networks exhibit complexity on many aspects: large topology, multiple interworking technologies, service variety, diverse, and mobile end‐user demand. Given the scale and depth of complexity, to reach a high level of autonomy in network automation requires a paradigm shift in both system and interface design. Within the system, autonomous operation requires real‐time awareness of end‐to‐end performance via correlated insight to user experience, application state, and network state across multiple network domains. Interfaces within the system should enable collection and sharing of insight to facilitate analytics and decisions. Interfaces should support the principles of intent, where the network receives high‐level objectives and service definitions from the operator and returns system state in a transparent and humanly understandable way. This also redefines the role of the operator, as it means both delegation of responsibility and delegation of trust towards the system.

An enabler for automation is the large volume of real‐time operational data generated in the network. Integrating machine learning (ML) methods into workflows enables leverage of this data for network automation and to unlock the potential of insight‐driven decisions and action within the system. However, a core challenge for ML is the interpretation ...