Improvement in auto scaling mechanism of cloud computing resources using Composite ANN

In cloud computing, auto scaling has attracted considerable attention from researchers and organizations for acquiring and releasing resources based on real-time demand from large data centers. Elasticity enables auto scaling resources on-demand; thus, a pay-as-you-go model can be implemented. Auto scaling suffers from various complexities and challenges because allocation and de-allocation of resources is dynamic as per the workload demand. The false prediction of resource allocation may lead to reduced quality of service and violation of service level agreements. For predicting resource requirements, various machine learning and deep learning (DL) technologies have been used. Although DL methods are accurate with predictions, they require Big Data. With increase in the size of input data, the model's complexity increases, which leads to overhead on the system. In this study, an artificial neural network (ANN) with linear regression is used to efficiently predict resource requirements. Aim is to achieve optimal resource allocation for systems with limited computing capacity and minimum overhead on the system. To minimize the input and utilize less hardware resources for prediction while retaining accuracy in the output that was produced, ANN with optimized linear regression is used. The results demonstrate that linear regression optimally minimizes weights on the nodes through each layer. Furthermore, when this minimized output is allocated to an ANN, it produces an ameliorated output with minimum overhead on the system