Enabling machine learning-assisted resource monitoring for network slice creation and management in OSM: design, implementation, and validation

As a novel concept introduced in 5th generation (5G) mobile networks, network slicing allows configuring a shared physical infrastructure into multiple logical networks that are virtually isolated from each other. Through these logical networks, resource allocation for specific use cases is tailored, facilitating rapid and flexible development of new services and applications. To create and manage network slices, open source management and orchestration (OSM) has emerged as a powerful tool for software developers and network operators. Within the scope of automatic slice generation and management using OSM, however, whether and how machine learning can play a role remain as a hardly addressed research question. In this paper, we explore the feasibility of applying machine learning to OSM for the purpose of improving slice creation and management efficiency. To do so, we introduce an enhancement in the monitoring module of the OSM architecture. More specifically, a machine learning based alarm monitoring sub-module is developed, such that a new field value is automatically generated every time when an alarm is identified. In addition, we create a prediction model for resource utilization prediction so that the most suitable resources can be allocated when a slice is created. Furthermore, we have implemented our solution in a network slicing platform we developed based on OSM and performed proof-of-concept validation through two network slicing scenarios. Through simulation-based validation and testing, we reveal that the proposed method achieves reliable performance and demonstrate the effectiveness of our solution towards automation network slice creation and management in OSM.