Leveraging Fog Computing for Harnessing the Service Latency in Cloud-Fog Computing: A Deep Learning Approach

Massive volumes of data are produced daily from every aspect of life due to the rapid emergence of internet-enabled devices. The capacity, processing power, and storage required to handle and store this enormous amount of precise and volumetric data are lacking from these internet-enabled devices. To handle these requests, Cloud computing has been suggested as a strong alternative. The massive amount of incoming traffic to the Cloud, however, results in latency overhead because of the distance between end devices and the Cloud datacenter. It is also an NP-hard computational problem to process these dynamic and heterogeneous requests with different requirements. In this sense, Fog computing seems like an alluring addition to the Cloud that can help overcome the aforementioned difficulties. As a result, this study created a collaborative computation framework by integrating a Fog layer between end devices and Cloud datacenters to minimize the incurred latency. Furthermore, resource monitoring is an additional challenge which requires a precise prediction of loads among computing nodes for processing latency-intensive tasks. Therefore, a multilayer LSTM is proposed to predict the workloads of the nodes. A binary JAYA is proposed to fine-tune the hyper-parameters of LSTM and the scheduling of tasks among computing nodes. The effectiveness of the proposed strategy has been validated for disparate scheduling metrics such as service rate, latency, and resource monitoring rate. The simulations showcase the effectiveness of the proposed method over other baselines.