Designing Cost-Latency Optimization of Task Scheduling Algorithm Over Fog-Cloud Computing Environments

Abstract

The Internet of Things has recently generated large amounts of data, which has an impact on cloud computing infrastructures. Fog computing is an extension and complement to cloud computing in which services like computation and storage are provided at the network's edge. However, the variation in fog resource capacity has introduced new challenges, such as latency reduction, deadline meeting, and cost reduction, in order to efficiently utilize resources and provide services. Task scheduling is essential for assigning tasks uploaded from IoT devices to efficient fog and cloud nodes while accounting for latency, resource utilization, deadlines, and costs. As a result, if proper task allocation to available resources is not strictly followed, it results in wasted resources and unsatisfactory service quality. Over the years, different approaches such as task scheduling and resource allocation have been proposed to optimize the quality of service in a fog-cloud environment. However, it is still challenging to get an accurate estimate of the optimum quality of services such as latency, resource utilization, deadline, and cost. This study proposed an improved hybrid grey wolf optimization algorithm for allocating IoT device-generated tasks to appropriate cloud and fog nodes. The suggested approach modifies the goal function to handle multiple objectives into a single objective using the scalarization method in order to solve the task scheduling problem that includes latency and cost while taking the deadline of tasks into account. On the cloudsim simulator, the effectiveness of the proposed approach has been evaluated. Based on experimental results, the proposed algorithm has reduced latency by 22.45%, cost reduction by 17.67%, outperforming FCFS, and has reduced latency by 21.9%, cost reduction by 18%, outperforming SJF algorithms. It has also reduced latency by 4 %, cost reduction by 7.1% outperforming PSO, reduced latency by 4.4%, cost reduction by 6.7%GWO, and reduced latency by 5%, cost reduction by 6.65 % outperforming GA. Additionally, compared to existing algorithms, the suggested approach more successfully meets the requirement of a task deadline