Minimizing Data Center Energy Consumption and SLA Violation via Prediction based Resource Provisioning

Cloud service provider’s main aim is to provide better services to cloud customer and proper utilization of the cloud resources in data centers, such that energy consumption cost and SLA violation rate are to be minimized. In this paper our approach is to obtaining future CPU load information in advance. As CPU is the key factor for the performance of application. So CPU load prediction is one of the most important aspect in the cloud resource provisioning. We propose two different CPU load prediction approaches based on Linear Regression and Support Vector Regression technique, which are suitable for dynamic characteristics of cloud applications and complex cloud computing environment. The proposed approach approximate the short time future CPU utilization based on the history of usage in each host. Planetab real workload has been considered for testing the performance of our proposed approaches. Experimental results show that the proposed techniques can significantly minimize the energy consumption cost and SLA violation rate.