Service quality-based network slicing optimization

Abstract

Network slicing and its optimization is a critical concept in today's telecommunication networks. It involves creating and jointly optimizing separate virtual segments within a single physical network to cater to specific users, devices, or applications. This paper presents an innovative framework that optimizes network resources using estimated QoE values for specific slices and inputs provided by operators, like traffic priority. Operators can also set strict Quality of Service (QoS) parameter requirements, such as minimum bandwidth requirements based on their Service Level Agreements (SLAs), which the model incorporates. Our proposed optimization framework capitalizes on packet-level estimations, avoiding the need for end-user side metrics like WebRTC logs. This innovative approach not only allocates resources based on initial estimates but actively and constantly refines resource allocation leveraging estimated QoE and traffic priority (given by the operator), ensuring optimal network resource allocation, thereby facilitating operators in meeting Service Level Agreement (SLA) compliance and maximizing revenue. In addition to this very dynamic scenario and solution, we propose more static algorithms as well, and investigate the impact of the learning and reconfiguration window i.e., the timeframe which is used as the optimizations input and the timefame in which the network is allowed to be reconfigured. In summary we propose a new network optimization framework based on the estimated QoE and inputs given by the operator (i.e., priority of the traffic and SLA constraints) and we investigate the impact of different network degradations and reconfiguration timeframes on the overall profit.