Intelligent VMs Prediction in Cloud Computing Environment

To fulfill the requirement for dynamic execution of customer's applications in cloud, efficient VM (virtual machines) forecasting techniques are required. Current researches are unable to accurately predict VMs usage for user's applications. Hence, we need a mechanism to overcome this problem so that VMs in cloud environment do not suffer from being unutilized. We propose a Bayesian model to determine VMs requirement for applications run in the cloud environment on the basis of workload patterns across several data centres in the cloud for different time interval during days of the week. The model is evaluated by considering CPU and memory benchmarks. The model is evaluated by using SamIam Bayesian network simulator and Benchmark traces collected from CloudHarmony benchmarking services. The simulation results indicate that the proposed model involving random demand scenarios provide insights into the feasibility and its applicability to predict the VM and its utility for customer applications, which helps in proper capacity planning. Further, it is able to predict VMs in Cloud environment with accuracies in 70% to 90% range, as compared to existing prediction models.