Using Machine Learning for Task Distribution in Fog-Cloud Scenarios: A Deep Performance Analysis

Abstract

For efficient utilization of Internet bandwidth and reducing the response time for delay-sensitive applications, we propose Machine Learning Based Task Distribution (MLTD) technique, which uses the Artificial Neural Networks for smart task distribution between the fog and cloud servers. In this paper, we evaluate the efficiency of MLTD in different conditions to detect the parameters that can impact its performance. Also, we compare the performance of MLTD with other similar methods in terms of Internet bandwidth utilization, response time, and resource utilization. The achieved results show that the performance of MLTD can be better or worse than the other methods, and the training procedure of the neural networks plays an important role in increasing the efficiency of MLTD.