云集群中的计算资源热点问题研究

发布时间：2018-03-03 11:05

本文选题：云计算　切入点：热点　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：云计算以便捷的按需服务的资源使用方式受到了各界的关注。云计算集群通过虚拟化技术将其上的资源提供给各个应用。但是云集群规模的增加以及应用的多样性都给云计算的资源管理带来了挑战。云计算的一个最主要的目标就是为用户提供稳定的物理资源支持。而云集群中的服务器资源热点的问题严重阻碍了这一目标。资源热点问题指的是服务器上的某些资源的使用量超过限制,进而干扰了其上的虚拟机正常运行的问题。为了保证应用的稳定运行,云计算平台必须能够及时发现集群中的热点服务器,并设法让其回到正常状态。因此,本文针对该问题展开了研究,主要研究发现热点的方案和解决热点的方案。本文的主要工作如下:1.本文设计并实现了一个云平台监控系统。该系统能够收集服务器和虚拟机的资源使用情况,并在发生资源热点时通知用户。2.基于最小化迁移时间的考虑,本文提出了 MCMS(Min-Cost-Min-Server)算法来计算最佳的虚拟机迁移方案。该算法通过迁移热点服务器上的虚拟机来解决热点问题。同时,在完成热点消除后,该算法尝试迁移资源使用量较少的物理机上的所有虚拟机来减少集群的服务器数量,减少能耗。3.本文设计并实现了热点消除系统。通过结合监控系统和MCMS算法,该系统实现自动发现热点并消除服务器热点问题。4.本文通过搭建真实环境验证了热点消除系统的可行性。同时,本文还通过仿真实验测试MCMS算法的性能。与其他两种常规算法对比,MCMS算法能够以最少的迁移时间解决服务器热点问题。并且,MCMS算法能够以更少的迁移时间来解决服务器热点问题。并且,MCMS算法能够使云集群中的服务器数量明显的减少。
[Abstract]:Cloud computing has attracted much attention from all walks of life because of its convenient on demand resource usage. Cloud computing cluster provides its resources to various applications through virtualization technology. But the increase of cloud cluster size and the diversity of applications. Sex brings challenges to the resource management of cloud computing. One of the main goals of cloud computing is to provide users with stable physical resource support. However, the problem of server resource hot spots in the cluster seriously hinders this goal. Resource hotspots refer to the use of certain resources on the server in excess of the limit, In order to ensure the stable operation of the application, the cloud computing platform must be able to discover the hot servers in the cluster in time and try to get them back to normal state. In this paper, the problem is studied. The main work of this paper is as follows: 1. This paper designs and implements a cloud platform monitoring system. This system can collect the resources of server and virtual machine. And notify user. 2 when resource hotspots occur. Based on the consideration of minimizing migration time, In this paper, we propose the MCMS Min-Cost-Min-Server algorithm to calculate the best migration scheme of virtual machine. This algorithm solves the hot spot problem by migrating the virtual machine on the hot spot server. At the same time, after the hot spot is eliminated, The algorithm attempts to migrate all virtual machines on physical machines with less resource usage to reduce the number of servers in the cluster and reduce energy consumption. 3. This paper designs and implements a hot spot elimination system. This system can automatically discover hot spots and eliminate server hot spots. 4. This paper verifies the feasibility of hot spot elimination system by setting up real environment. At the same time, The performance of the MCMS algorithm is tested by simulation experiments. Compared with the other two conventional algorithms, the MCMS algorithm can solve the hot spot problem of the server with the least migration time, and the algorithm can solve the problem with less migration time. And the MCMS algorithm can significantly reduce the number of servers in the cluster.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP302;TP368.5

【参考文献】