CPU Frequency Scaling Algorithm for Energy-saving in Cloud Data Centers

Abstract

High energy consumption becomes an urgent problem in cloud datacenters. Based on virtualization tech- nologies, the pay-as-you-go resource provision paradigm has become a trend. Specifically, Virtual Machine (VM) is the basic resource unit in data center for resource migration and provisioning. Many researches have been devoted to improve datacenter resource utilization and reduce power consumption by VM placement. As the most important power consumption resource, CPU has a fluctuant frequency range. Based on CPU frequency scaling, a new approach for VMs placement is proposed. The approach is realized in two stages. In the initial stage, we propose a multi- objective heuristic ant colony algorithm, which will find the optimization solution for energy saving as well as service- level agreement (SLA). In the dynamic stage, by using autoregressive prediction and CPU frequency scaling, the proposed approach can adjust the CPU utilization if needed, not depending on whole VM migration. The experiments show that the energy saving algorithms based on CPU frequency scaling are much better than the traditional BFD and FFD algorithms in saving energy and satisfying SLA.