Perspective Indicators for Service Orchestration in Zero-touch network and Service Management Context

The pursuit of more efficient computer network services has propelled the development of Zero Touch Network and Service Management (ZSM). ZSM aims to automate and simplify the operation and management of telecommunications networks and services. The objective of this approach is to eliminate the need for human intervention or direct interaction in the setup, provisioning, supervision, and resolution of associated issues. In this context, this paper presents a Machine Learning-based service orchestration proposal that uses five perspective indicators to represent different viewpoints of a service operation: Availability, Demand, Cost, Utility, and Perturbation. To achieve these objectives, the system analyzes user service request logs and integrates them with other collected monitoring data to generate five indicators. Furthermore, a map of IP address blocks in different regions is employed to ascertain the geographic location of demand and the available computational resources for the specified service instance. These indicators are then used to configure a Machine Learning engine based on Deep Q-Learning. The engine dynamically generates and oversees applications across the infrastructure, optimizing the cost-benefit ratio of service operations. We conducted experiments in simulated environments to validate the efficacy of the proposed solution. The results showed that our developed mechanism could dynamically adjust the allocation of instances, leading to improved performance and efficient use of available resources.