基于Xen的多虚拟机资源调度技术研究计算机科学与技术专业论文.docxVIP

哈尔滨工业大学工学硕士学位论文 哈尔滨工业大学工学硕士学位论文 - - II - Abstract Recently, cloud computing technology has become the focus of concern in both academia and industry. Services offered by large data center, especially Web applications, and various levels of virtualization technology make up the infrastructure of cloud computing. Typically a data centers has a large number of servers, but computing power of these servers are underutilized due to poor planning which results in not only a waste of resources, but also large amounts of additional carbon emissions production that affects the environment. Modern data center is usually composed of many multiple virtual machines systems. Virtual hosts can share one or a group of physical machines. Compared with traditional distributed computing systems, the scheduling method in distributed system based on multiple virtual machines includes not only the scheduling of jobs to run on the host, but also the scheduling of virtual resources. These virtual resources can be dynamically allocated at runtime, so how to make the deployment to make full use resources is quite worthy of study. This paper mainly focuses on the CPU and memory resources. Different regulation strategies are designed based on the sensitivity of QoS. Due to the high sensitivity impacted on QoS by CPU, real-time attention to the relationship between QoS and CPU is required. This paper supposes a design of system including a local utility function based on economic principles and a global utility function based on optimization theory. This change of sensitivity in memory’s impact on QoS is large, so resource utilization is needed to monitor. This paper designs a local regulator based on control theory which is used when the resource is sufficient and the balance regulator based on economic theory which is used when it is in shortage of memory. Experimental results show that both of the two regulation systems can significantly improve the performance of Web applications. Keyword