Power Metering Model And Energy Efficiency Optimizing Mechanism In Cloud Computing

In large scaled cloud computing, the massive servers running as a parallel and distributed system consume a large amount of electrical power, which makes energy consumption become a major operational cost. Reducing the energy consumption of cloud not only saves costs for providers to obtain attractive economic benefit, but also helps alleviate worldwide energy crisis and reduces emission of greenhouse gas. It becomes more and more important to reduce power of infrastructure under the premise of guaranteeing Quality of Service (QoS). Virtualization technology used in cloud computing platforms can improve energy efficiency and reduce costs. It raises certain additional challenges of power management because the granularity of power monitoring and resource scheduling will be individual virtual machine (VM).Based on the above background this paper researches the characteristic of energy consumption and related technologies in cloud computing and virtualized environment. We present an energy efficiency management architecture called EECLOUD which mainly consists of an online power metering model and VM allocation and dynamic VM placement strategies. The main research in this paper includes the following three parts:a) Analysis of current related work in energy management of cloud computing and the requirements of energy-saving method in virtualization cluster and distributed environment. Design of energy management structure which describes overall solution of energy monitoring, management and optimization for cloud environment.b) Proposal of power metering model for virtual machine to achieve equivalent visibility of power in virtualized platform. Power model uses running performance metrics of processor, memory and disk device as input to estimate the power consumption value of individual virtual machine or physical server. The power model provides support for making energy management strategies and decision.c) Research on energy efficiency optimizing mechanism for reasonable resource allocation and virtual machine placement to ensure energy-saving as well as performance requirements. Establishment of an energy-aware utility function which allocates VMs by making trade-off between job runtime and allocated virtual machines. Development of VM placing strategy, which takes both VM runtime duration and resource capacity into account and minimizes global cluster energy. Proposal of a Dynamic VM Bin-packing algorithm to place VMs to the least amount of physical server.