Q-learning approach for load-balancing in software defined networks

Abstract

In this paper, we propose a Q-Learning approach for load balancing in Software Defined Networks to reduce the number of Unsatisfied Users in a 5G network. This solution integrates Q-Learning techniques with a fairness function to improve the user experience at peak traffic conditions. With typical high rates offered by 5G and future networks single user behavior shall have a significant impact on the Quality of Service (QoS) on the rest of the users. Therefore, we are in need of responsive networks based on their utilization and on the number of users occupied. In this paper we classify users into different groups and normalize the resources to provide the best QoS. The simulation results verify the improvement in terms of the number of Unsatisfied Users and of the connections dropped. Additionally, it enhances per-flow resource allocation while avoiding over-utilization of certain network resources. In a nutshell, this proposal will serve any future network with high traffic conditions to deliver the best QoS to their end users.