Performance Analysis of Adaptive Cloud through Workload Prediction Supporting Resource Provisioning

Abstract

Virtualized resource allocation to cloud users in accordance with their requirement is a pivotal step for newlineapplications deployment. To handle continuously changing workload on cloud infrastructure is the newlinecomplex task. During provisioning, the major concern is that if demand is low, excess of resources are newlineavailable which leads to over provisioning and if demand is high, the resources may not be enough, newlinewhich leads to under provisioning and poor QoS (Quality of Services). In case of cloud environment, newlineresources are provided on demand. Efficient resource provisioning needs a proactive approach which newlinecan predict the future load before provision of resources to reduce the under or over provisioning newlineproblem. The above-mentioned challenge has to be handled through proactive resource provisioning newlineapproach, which can predict the future demands of resources. This approach helps in deploying and newlineprovisioning of resources efficiently based on the demands without loss of QoS. A prediction model newlinereleases the unused resources from the pool of resources by maintaining Quality of Services (QoS) for newlineresource provisioning in advance. To reduce the latency and improve the performance of cloud, newlineaccurate workload prediction strategy for provisioning of resources efficiently is the dominant aspect in newlinethe cloud-based services. In our research work, used model predict the future workload to consider the newlinearriving requests from cloud servers for resource provisioning efficiently. A prediction model which can newlinepredict the resource demands in advance for dynamic resource provisioning from the observed or newlinehistoric database in a virtualized environment is proposed. The prediction model ARIMA-PERP newline(Autoregressive Integrated Moving Average-workload Prediction for Efficient Resource Provisioning) newlineevaluated the implementation so as to satisfy the on-demand need of end users for efficient resource newlineutilization. The accuracy of prediction model is assessed for proposed QoS parameters and the crossvalidation newlinemethod. newlineThe proposed approaches