面向云计算的虚拟机能耗模型及其应用方法研究

With the energy consumption problem of cloud computing is becoming increasingly prominent, the energy consumption optimization in cloud computing is becoming more and more important. The power consumption model is the basis of energy consumption optimization in cloud computing. Therefore, this project will first study the mathematical relationship between resource usage of key parts (such as CPU, memory and disk) of the cloud server and system energy consumption and the power consumption models of the key parts of the cloud server are proposed. And then a virtual machine power consumption model based on power function is proposed to overcome the shortcomings of traditional CPU energy consumption model which does not deal with virtualization technology. In order to meet the needs of virtual machine energy adaptive modeling for various types of cloud workload (such as CPU intensive, memory intensive, IO intensive), an adaptive energy power model based on the type of workload is proposed. Based on the power consumption model, the project will present a software-based energy consumption estimation method for cloud environment and develop the distributed energy measurement system, propose multi-resource (such as CPU, memory, IO) energy consumption simulation method for cloud computing and by extending the class library of CloudSim, and implement the API for multi-resource energy consumption simulation. Based on the above results, the project will further research on energy optimization algorithm and application technology in cloud computing and big data applications. The research of this project not only has positive effect on the development of the energy consumption model of cloud computing, but also has a certain application value on energy-efficient management of Cloud Data Center.

随着云计算的能耗问题日益突出，其能耗管理优化研究越显重要。能耗模型是云计算能效优化研究的基础，为此，本项目将研究云服务器的关键部件（CPU、内存和磁盘）的资源使用情况与系统能耗的数学关系，给出各关键部件能耗建模方法，并重点针对传统CPU能耗模型对虚拟化技术考虑不足等问题，提出基于幂函数的虚拟机能耗模型；为了实现面向云环境下多种类型负载（CPU密集型、内存密集型、IO密集型）的虚拟机能耗自适应建模，提出基于负载类型的自适应能耗模型。在提出与建立的能耗模型基础上，提出面向云环境的基于模型估算的能耗测算方法，并研发分布式能耗测算系统；提出面向多资源的云计算能耗仿真方法，通过扩展CloudSim类库实现CPU、内存、IO等多资源的能耗仿真API；并研究面向云计算和大数据应用的能耗优化算法与技术。本项目的研究不仅对云计算能耗模型发展有较好推动作用，而且对云数据中心的节能技术有一定应用价值。

随着云计算的能耗问题日益突出，其能耗管理优化研究越显重要。能耗模型是云计算能效优化研究的基础，为此，本项目研究云服务器的关键部件（CPU、内存和磁盘）的资源使用情况与系统能耗的数学关系，针对传统CPU能耗模型对虚拟化技术考虑不足提出了基于幂函数的虚拟机能耗模型CAM，实验结果表明在执行计算密集型负载时CAM在VMs功率估计中产生的误差相比线性和累积模型都更小（平均为4.26%）；然后提出了基于幂指函数的服务器CPU功耗模型，与现有的线性、多项式和幂函数模型相比，该模型能够更准确地估计最新云服务器的CPU功耗；并且，为了利用ANN模型的更优的泛化和非线性建模能力建立更加精准的服务器功耗模型，提出了基于ANN的服务器功耗建模方法，建立基于BP、ENN和LSTM神经网络的服务器功耗预测模型，在不同类型的任务负载下进行模型的分析和评估，功耗预测的精度优于多元线性回归和支持向量回归方法。为了实现面向云环境下多种类型云应用负载（CPU密集型、内存密集型、IO密集型和混合型）的能耗自适应建模，提出基于负载类型感知的自适应云服务器能耗测算方法和基于硬件感知的CPU能耗测算方法，研发了面向云服务器的分布式能耗测试系统。项目提出了面向多资源的云计算能耗仿真方法，并基于CloudSim实现了面向云计算多资源能耗仿真API。项目提出了基于最佳能效的虚拟机调度算法和一种基于能效模型的能耗优化云任务调度算法等。项目研发的云计算能耗建模方法、能耗模型和调度优化技术开展了成果转化应用，取得较好的经济和社会效益。