Real-Time Task Scheduling and Resource Scheduling in Fog Computing using Deep Learning Techniques

Abstract

Cloud capabilities may now be accessed from the network's edge thanks to fog computing. In order to fulfill the needs of organizations that have particularly high standards for the quality of the services they provide, it is essential to choose the most suitable site for the execution of applications. So, a cloud service has to have a dependable task scheduling system in order to establish in which geographic zone a certain piece of information may be put into action. We provide two different schedulers that are able to maximize work scheduling by using the resources of cloud computing and fog computing respectively. Both planners are grounded on nonlinear mathematical programming as their primary methodology. When deciding which processors should be assigned a certain work, the schedulers do not use any rules that have already been established; rather, they take into consideration a range of characteristics, such as availability and performance, when making their choices. The recommended algorithms usually perform much better than the scheduling problem that are already in use, such as the Randomised and Round Robins approaches, without sacrificing the quality of the service provided, as shown by the numerical solution. This research aims to enhance the total effectiveness of execution of tasks in IoT systems by consciously selecting certain real-time activities to carry out at the fog layer. This is done with the intention of increasing overall efficiency. We present a way for scheduling activities in a cloud-fog computing environment that takes use of fuzzy logic to split up work between the cloud and fog layers. This approach can be found in our paper "A Technique for Creating Schedules in a Cloud-Fog Computing Environment. The approach selects appropriate processing units to carry out the designated tasks in the fog nodes that include homogeneous devices, taking into consideration the requirements of the job (such as compute, storage, and bandwidth), as well as the limitations imposed by those requirements (e.g., deadline, data size). According to the findings of the simulation tests, the proposed method is superior to the current best practices in terms of the percentage of jobs that were successfully completed, the average turnaround time, the amount of time required to calculate the results, and the latency rate.