| Cloud computing improves the utilization rate of physical resources.It enables the single physical platform with rich system environments and application services,by optimized in-tegration and allocation of physical resources through virtualization technology.However,cloud service platform facing serious challenges of excessive consumption of power,com-puting,storage,bandwidth and other resources.Because,the platform often provides the services at all-weather real-time response via centralized hardware devices,high-density application software and massive complex computing tasks.According to the initiative poli-cies of "green,low carbon" of China,the primary goal of green cloud computing lie in the fact to improve resource utilization(including networks,servers,storage,applications,ser-vices),reduce the amount of used physical equipment and reduce energy consumption of traditional cloud platform.Virtual machine is the core service providing facility of cloud computing platform.Therefore,strategy of virtual machine deployment is the key to reduce the amount of physical equipment and improve physical resource utilization.Even though scheduling and allocation of virtual resources is at the topmost importance,the core idea of traditional load balancing and scheduling strategy is to avoid too heavy or light workload of the physical server.It does not describe the resources utilization from the perspective of how much resources that virtual machine takes from physical machine in a fine-grained level.That result in the same physical machine often hosts virtual machines with both heavy and light resource utilization.If the virtual machines are classified according to utilization,and selectively migrate some virtual machines then shutdown certain physical machines,it can effectively reduce the overall energy consumption of a cloud platform.Therefore,on the premise of ensuring Service Level Agreement,this dissertation proposed an virtual re-source utilization based energy model to systematically study the evaluation,selection and deployment optimization of the cloud platform in order to provide theoretical and technical support for the construction of low-power cloud computing platforms.The authors’ main contributions are summarized as follows.1.Based on the analytic hierarchy process,propose a virtual machine evaluation model for cloud computing platform.This model can effectively evaluate the impact of virtu-al machine characteristics and resource consumption towards the physical machine energy consumption.Thus solved the problem of screening virtual machine with light resource utilization.First,based on the application requirements,the appropriate quantitative indi-cators are established by references the parameters consist of status attributes such as work characteristics,usage and utilization of the virtual machine.Second,based on quantitative indicators,a virtual machine evaluation system has been built to determine the judgment ma-trix between the various features,and then the consistency of the judgment matrix has been verified.Finally,based on frequent itemset theory,a metric model is proposed for in-depth study of running status of the virtual machine of evaluation index system.Experimental results show that the virtual machine evaluation model achieves a quantitative description of the characteristics of virtual machine features and resource utilization,and it is able to distinguish virtual machines with light utilization effectively.2.Based on collaborative filtering method,proposed a virtual machine recommendation model.The model is capable for the classification of massive virtual machine instances on cloud platform.It resolved the problem of the determination of suitable migrate virtual ma-chines by extracting features from utilization and energy consumption.First,an attribute matrix of the virtual machine has been constructed for the calculation of similarity between the suitable migrate virtual machine and other virtual machines.Secondly,by analyzing the similarity calculation results and k-means clustering analysis,the hidden feature parameter-s have been mined.Finally,based on the average weighted strategy,the preferred feature parameters are summarized then eventually formed the recommended collection of virtual machines.The results showed that,compared with the traditional item rating based recom-mendation algorithm,recommend accuracy of the proposed method has been increased by 10%maximally.3.Based on optimized ant colony algorithm,a virtual machine deployment optimization model has proposed.The model is able to determine the physical machine which suitable for hosting migrate virtual machines.Thus solved the problem of redundant power consump-tion according to idle resource waste of the physical machine.First,based on the utilization parameters of the virtual machine,the idle resources and energy consumption model have been proposed.The models are dedicated to quantify features of virtual resource utilization and energy consumption of physical machines.Secondly,a multi-objective optimization s-trategy has been built for virtual machine deployment in cloud environments.Finally,the optimal deployment of virtual machines schemes has been determined based on feature met-rics,multi-objective optimization and ant colony algorithm.Experimental results show that compared with traditional genetic algorithms based MGGA model,the convergence rate is increased by 16%,and the optimized highest average energy consumption is reduced by 18%. |
