A Management Mechanism Of Virtual Machine Cluster Based On XenServer

In today's Internet field,the technology of virtualization is developing in full swing.The emergence of virtualization technology solves the issue that the low utilization of computing hardware,it improves the efficiency of resource utilization by abstracting a single physical resource into several virtual resources.Especially in terms of servers,virtualization technology turns a single physical server into several logical virtual computing units that provides efficient computing services for a large number of data center.However,there are still some researches need to do in the field of virtualization technology at present.For example,how to efficiently allocate and utilize computing resources on a single physical node,how to flexiblly arrange virtual machines in multiple physical nodes so that the server cluster can load balance and meet a variety of computing needs,and how to reset the resource of virtual machine according to the usage status of the virtual machine in a period of time.In addition,at present,the main provider of cloud hosting services provides command line window to users,the command line window is difficult for some computer professionals,not to mention common users.The main work of this thesis is that through the analysis of the existing virtual machine management mechanisms,combining with the practice background of virtual digital community,put forward the virtual machine cluster management mechanism and strategies.Based on these mechanisms and strategies,the thesis develops a virtual machine cluster management system to provid virtualization services for users by automatically scheduling virtual machines.Based on the XenServer service platform,this thesis uses iSCSI protocol to create shared SR and privated SR to store users' virtual machine image files.At the initial time of deploying virtual machines,the thesis takes the differences of hosts' performance into consideration,and uses the algorithm that depends on probability distribution to staticly deploy virtual machines.In this thesis,the system sets two thresholds for CPU utilization and memory utilization,and forecasts the CPU utilization by gray forecasting model to locate hot mechine.After determining the hot spot,the system calculates every VM's Minimum Migration Factor to locate VM that will be migrated.In the aspect of locating target physical machine,the thesis chooses Maximum Left offset Priority algorithm.Besides,the thesis uses K-means algorithm to resize VM's virtual resources so that maximize the utilization of hardware resources.In the aspect of system implementation,this thesis uses XAPI and JavaFX graphics frame as a major technology,and integrates mainstream development framework and technology such as Spring,Mybatis,Thrift and RabbitMQ.Finally,the thesis tests the system function and algorithm performance,the result shows that the function of the system fully meets the requirements and the performance of the algorithm is more suitable for virtual digital community.