Research On A Service Relevance Based Virtual Machine Migration Technology In Cloud Computing

Cloud computing has received extensive attention in recent years as an emerging method of computing resource utilization because it can achieve a win-win situation between cloud providers and users.The complexity of resources in the cloud computing center and the variability of user requirements make servers often overloaded or underloaded,which seriously affects the execution efficiency and user experience of cloud computing systems.Therefore,how to deal with load imbalance becomes an urgent problem.The virtual machine migration technology becomes a research hotspot at domestic and foreign because it can redistribute resources through virtual machine remapping to achieve the purpose of balancing load.Based on the summary and analysis of the existing virtual machine migration technology,this thesis proposes a service relevance based virtual machine migration technology considering the relevance of application services in virtual machines.In the process of virtual machine placement,the technology constructs a virtual machine migration group and determines a target server by defining the service correlation factor,thereby effectively reducing communication overhead between servers.In the process of virtual machine migration,the technology realizes memory migration of virtual machines based on Post-Copy method according to setting service priority,effectively reducing the occurrence of page fault and improving virtual machine migration efficiency.The main work of this thesis is as follows:(1)Proposing a service relevance based virtual machine placement method.This method mainly solves the following two problems: the selection problem of the virtual machine and the selection problem of the target server.Firstly,according to the problem analysis,the virtual machine placement is abstracted into NP complete problem.Then,the service correlation factor is defined according to the communication relationship between services.Based on this,the service relevance based virtual machine selection strategy is designed.The strategy combines multiple virtual machines with tight communication to form a virtual machine migration group as a migration unit.Next,using the service correlation and greedy ideas to optimize the cost model VMMigAware to determine the optimization goal of the problem and finally find the minimum communication cost target server that meets the conditions.Finally,verifying the effectiveness of the placement method in balancing system load and reducing communication cost by CloudSim.Compared with RR,LC,AG and AppAware algorithms,the method can reduce the communication cost between servers while balancing the system load.(2)Proposing a service priority based Post-Copy memory migration method.Firstly,the problem of memory migration in this thesis is analyzed and the main goal of memory migration is to reduce the time of outbound and page fault.In order to reduce the memory migration time and speed up the load balancing process,this thesis divides the overall process of the memory migration method into four phases: migration minimum working set,memory pages move out,memory pages move in and request memory pages on demand.In the memory move-in and move-out phase,a service priority based memory pages push algorithm is proposed.The algorithm sets the memory pages push order based on the service priority and updates the priority in real time according to the page fault to reduce the occurrence of page fault.Compared with the traditional memory migration method,the method proposed in this thesis can speed up the virtual machine migration and reduce the occurrence of page fault when the performance impact is small.(3)Designing an efficient autonomous virtual machine migration framework.The framework is divided into three modules: adaptive resource monitoring module,migration scheduling module and migration execution module.The three modules coordinate with each other to promote virtual machines migration efficiently.Then,realizing a cloud computing prototype system by service relevance based virtual machine migration technology.Finally,testing the technology of virtual machine migration whether can perform autonomous virtual machine migration under heavy load,whether can balance the system load,whether can speed up the virtual machine migration and whether can reduce the communication cost between servers.Compared with the default method,the final result shows that the service relevance based virtual machine migration technology proposed in this thesis can solve the server load imbalance problem autonomously,quickly and efficiently and reduce the communication cost between servers.