A Novel Resource Management Framework for Fog Computing by Using Machine Learning Algorithm

Introduction

In this era, the most upcoming technique is the edge devices IoT and mobile device users are increasing in drastic manner. By this same speed, the high complexity real time applications and data intensive applications gadgets such as Virtual Reality, Augmented Reality, Drones are developed. The resources limitations, energy consumption, mobile network traffic, high computation and storage are the main difficulties facing by the user. It is very difficult to solve these limitations with cloud computing model to support the cloud, a new concept to deploy task and storage, edge computing, fog computing, mobile edge computing are introduced.

To overcome the challenges and to support the cloud fog computing is proposed by Cisco, virtual platform provides storage, computation, network services and plays its role between the end devices and cloud (Bonomi et al., 2012). A decentralized networks which performs actions as cloud server but it relays on the cloud computing for high computation (Mahmud & Buyya, 2017). It performs wonders for edge devices, IoT devices and mobile devices due its low latency for real time applications and low mobile network traffic. In fog computing, main challenge is resource allocation and heterogeneous devices.

Offloading (Hyytia, Spyropoulos & Ott, 2015) is the way of transferring the task from the edge device to the fog nodes or to the cloud. Due to large complex applications and lack of resource its task or the part of the task offloaded is to the cloud or to the fog node to overcome the above mentioned challenges. Offloading decision is made based on the prediction technology and KNN algorithms to provide the best resource management for fog nodes. The important of fog computing and its characteristics are represented the energy minimization IoT devices resource management is discussed (Dastjerdi & Buyya, 2017).

Machine learning is way to increase the intelligence of the system and it is important to analyze. It is the way of learning the system for high computational real time applications (Angra & Ahuja, 2017). Detailed learning process about the data to understand and to classify accordingly will lead to improve the knowledge about the data (Mitchell, 1997). Machine learning algorithms applied to large data set to learn the data. Machine learning algorithm is used to improve problem solving with the prior knowledge about the data. It leads the dataset to train according to the learned information and it will find the way to reduce the problem and improves its performance. There some challenges in the machine learning while using in the data like how much data need to learn?, what algorithm is fit for this data? etc.,

The proposed method of this system used the machine learning algorithm KNN for resource allocation in fog nodes. It will show the nearest fog node to the user by using the machine learning KNN algorithm for efficient offloading and resource management. K- Nearest Neighbors (KNN) (Parsian, 1015) is the machine learning classification algorithm, here K is a numerical value always greater than 0.

KNN algorithm is used to find out the nearest data points in the dataset using the standard Euclidean distance. The (KNN) is a non-parametric method used for classification, to calculate the accuracy and for the validation of the KNN classification confusion matrix and the statistical method likelihood-ration is also used (Huang et al., 2013).