An Energy- and SLA-Aware Double DQN Scheduling Framework for IoT–Fog–Cloud Systems

Abstract

The related heterogeneous tasks in Modern IoT applications are in charge of meeting strict time and resource constraints and must be efficiently processed in the cloud. Requires only communication with the fog-cloud devices infrastructure and usually managing to avoid large delivery delays and consequent network congestion which generates latency on the services. For this reason, Fog computing supports the concept of task execution using intermediate computing nodes near the user devices providing faster response times in addition to system efficiency gains than processing tasks in the cloud itself, However in a fog-cloud architecture the selection of nodes where the tasks should be placed is a non-trivial problem because of varying workloads and limited fog resources. In this paper, we present an adaptive task scheduling framework leveraging Double Deep Q-Learning (DDQN) for IoT-Fog-Cloud environment. The performance of the implemented framework is validated and tested using the iFogSim simulator which is a widely used simulation platform specifically designed for successful IoT-Fog-Cloud methodologies. Heterogeneous fog nodes,network delays,and energy consumption are accurately modeled by iFogSim,which makes it ideal for intelligent scheduling strategy evaluation. When it comes to the performance, we will consider the following:Average latency,total energy consumption,and SLA violation rate. The experiment results reveal that our DDQN-based scheduler is able to effectively lower the latency and energy consumption and low SLA violation rate. The conclusion validates largely the usage of Double Deep Reinforcement Learning for dynamic and energy-efficient task scheduling in IoT-Fog-Cloud systems.