Predictive analysis for cloud resource management

Abstract

Over the past few years, there has been a considerable increase in the strain placed on the cloud server due to the rapid growth in demand for cloud resources. Resource provisioning is one of the complex issues that arise in a cloud system. Resources should be distributed dynamically based on the application's changing demand. Over-provisioning results in increased costs and energy waste. Contrarily, under-provisioning leads to SLA violations and a drop in service quality (QoS). As a result, resource allocation should be as close to the applications' current demand as possible. So, the Workload prediction is essential in determining the precise resources needed for an application to run successfully on the cloud. If a cloud computing model uses its resources in the most effective way possible, it is considered efficient. By using the Statistical models along with the Machine Learning model, this work shows a comparative analysis for predicting the workload of the virtual machines. Result shows that the machine learning model give better results than the statistical models.