An Online Energy Saving Resource Optimization Methodology for Data Center

In order to reduce energy consumption of data centers while employing infrastructure resource effectively, a comprehensive resource management method using an improved online energy saving mapping algorithm for virtual machines of data centers is proposed. An intelligent feedback management framework is built for online resource optimization. We propose reinforcement learning and threshold based virtual machine migration to find the lowest energy consumption mapping between virtual machines and host of data centers dynamically and avoid the resources contention in physical machines. Our experimental results shows that the proposed effective online energy saving resource optimization methodology can reduce data center energy consumption by at least 30.3% compared to traditional random placement and save at least 10% running time compared to the offline Q-Learning scheduling algorithm.