大规模云中基于用户体验和收益优化的能效资源提供技术研究

随着云计算规模的扩大、云服务竞争日益激烈和IT能耗持续上升，围绕用户体验、节能减排和收益模型等问题的云资源管理策略研究具有重要的理论及实际意义。本项目研究大规模云中基于用户体验和收益优化的能效资源提供技术。针对大规模云中应用的多样性，提出组合策略实现差异化资源提供。研究基于用户体验矩阵和云收益矩阵的云服务选择策略，采用层次分析法和效用函数，设计基于用户占优和云服务商占优的云服务选择算法，以实现协同优化。分析能效和用户体验之间的关系，运用能量效益函数实现用户QoS满意度和能耗平衡的虚拟机提供方法。运用对偶分解技术，提出基于云效用优化的资源提供算法，实现云效用优化及用户公平性。研究基于DBSCAN密度聚类的能效虚拟机提供方法，减少虚拟机创建开销。从碳排放、能耗、收益、QoS等方面研究基于收益优化及能耗感知的虚拟资源提供技术。构建云实验床，对所提出的方法进行验证和测试，并提供云平台示范应用环境。

With the expansion of the scale of cloud computing, increasing competition of cloud services and continued growth in IT energy consumption, cloud resource management strategies around user experience, energy conservation and revenue models have important theoretical and practical values. The project aims to study user experience and revenue optimization based energy efficient resources provisioning in large scale cloud computing. For large scale cloud application diversity, a combination of strategies is proposed to provide differentiated resources. Cloud service selection strategy based on user experience matrix and cloud revenue matrix is studied. By adopting analytic hierarchy process and utility function, user dominant and cloud provider dominant cloud services selection algorithms are designed to obtain collaborative optimization. Analyzing the relationship between energy efficiency and user experience, the virtual machine provisioning method is studied to balance the user QoS satisfaction and energy consumption by using the energy utility function, which aims to optimize energy utility. By using dual decomposition technique, cloud utility optimization based user experience oriented resource provisioning algorithms are proposed to achieve cloud utility optimization and user fairness. DBSCAN density clustering based energy efficient virtual machines provisioning method is studied to reduce the overhead of virtual machine creation and reconfiguration. From the perspectives of carbon emissions, energy consumption, income, QoS, etc, revenue optimization based and energy aware virtual resources provisioning scheme is studied. The cloud environment and test bed is built for testing and analyzing the proposed method, and cloud demonstration application environment is provided.