Neural-Hill: A Novel Algorithm for Efficient Scheduling IoT-Cloud Resource to Maintain Scalability

Abstract

The age of the Fourth Industrial Revolution (4IR) is the era of smart technologies and services. The Internet of Things (IoT) is at the heart of these smart services. The IoTs are resource-constrain devices. They act as middleware in intelligent systems and maintain communications between cloud servers and smart services. Processing related to intelligent decision-making, including data processing, cleaning, feature extraction, and analysis, is performed on the cloud servers. The IoT devices respond according to the decisions the applications run on the cloud servers make. The massive number of internet-connected devices is increasing by 8% per year. The cloud infrastructure backing these enormous numbers of IoT devices must be scheduled efficiently to maintain Quality of Service (QoS). An optimized scheduling scheme is essential to minimize the cost and enhance scalability. This paper proposes an innovative and novel algorithm, Neural-Hill, which combines the Deep Neural Network (DNN) and Random Restart variant of the Hill Climbing algorithm to schedule IoT-Cloud resources efficiently and ensure scalability. It is a preemptive scheduling algorithm designed to operate in dynamic task scheduling. The performance of the Neural-Hill algorithm has been evaluated in terms of optimal solution-finding time, execution time, routing overhead, and throughput. The experimental results demonstrate the significant quality of service improvement with the assurance of better scalability.

Keywords

• Internet of Things
• cloud computing
• efficient scheduling
• cloud resources
• deep neural networks
• hill climbing