Modification of Reward Function in Reinforcement Q-learning Agent to Improve Quality of Service in Federated Edge Cloud

This study proposed the modification of reward function based on logarithmic approach and reduction variable as well as introduces the additional variables which are MIPS (Million Instructions per Second) and RAM (Random Access Memory).The evaluation of several modification toward reward function used in Reinforcement Q-Learning in FEC (Federated Edge Cloud) has been conducted.The Quality of Services (QoS) evaluation using workload dataset from PlanetLab shows the proposed reward function based on Logarithmic MIPS or RAM with reduction parameter was able to improve the SLA (Service Level Agreement) Violation up to 10.6% followed by SLA-PDM (Performance Degradation due to Migration) up to 10%.Furthermore, energy efficiency was preserved, consistent with the previous study, which had already demonstrated efficient energy usage.Overall, it can be concluded that the proposed modification of reward function has more influence on improving QoS performance (SLA Violation and SLA-PDM) than reducing energy use in FEC environment.