基于成本的云计算动态资源分配与故障探针选择方法

发布时间：2018-11-02 09:11

【摘要】：随着云计算越来越受到人们的欢迎,其规模和复杂度日益增大,也给管理带来了新的挑战。成本因素成为云计算管理需要考虑的主要问题。云计算数据中心规模不断扩大,能耗成为了数据中心的主要运营成本。为了满足峰值负载的需求,云计算预备了大量计算资源。大部分的物理服务器在大部分时间并没有被完全利用。虽然如此,这些服务器的运行依然会消耗大量的能耗。根据数据中心负载量的变化,通过动态资源分配的方式提高数据中心的资源利用率同时降低能耗对云计算有重要意义。云计算依赖于底层计算机网络的支持,而逐渐庞大和复杂的网络让网络管理的成本越来越高。传统的基于被动告警事件关联方式的故障诊断在云计算网络中并不适用。主动探测的方法由于其灵活性,可能成为云计算网络故障诊断的解决方案。由于主动探测会对网络性能造成影响,需要限制探测的成本。在成本约束下高效地挑选探测的探针是本文的另一个研究点。本文的贡献包括： (1)为了解决虚拟化云计算数据中心的能耗问题以及资源利用率低的问题,本文设计了一种基于能耗成本的动态资源分配方法DRAMDT。考虑到一台物理服务器上的虚拟机实例间运行截止时间差异较大可能造成服务器长期处于未完全利用状态,造成能耗的浪费,DRAMDT算法利用随时间轮转的虚拟机资源子池对虚拟机实例进行分组。分组内的虚拟机实例才能被放置在相同的物理服务器上。仿真实验验证了DRAMDT算法的有效性和可行性。 (2)基于信息熵的性质,本文证明了一个关于探针信息熵增益的公式。基于这个公式,本文提出了一种计算探针信息熵增益的近似计算方法。有了这种近似计算方法,本文进而提出了一种高效的故障探针选择方法MEAP。MEAP算法的有效性和可行性在本文的仿真实验中得到验证。
[Abstract]:With the increasing popularity of cloud computing, the scale and complexity of cloud computing is increasing, which brings new challenges to management. Cost factor becomes the main problem that cloud computing management needs to consider. Cloud computing data center scale continues to expand, energy consumption has become the main operating cost of the data center. In order to meet the peak load requirements, cloud computing has prepared a large number of computing resources. Most physical servers are not fully utilized most of the time. However, these servers still consume a lot of energy. According to the change of data center load, it is important for cloud computing to improve the resource utilization of data center and reduce energy consumption by dynamic resource allocation. Cloud computing relies on the support of the underlying computer network, and the increasingly large and complex network makes the cost of network management higher and higher. Traditional fault diagnosis based on passive alarm event association is not applicable in cloud computing networks. Because of its flexibility, active detection may become a solution for cloud computing network fault diagnosis. Since active detection has an impact on network performance, it is necessary to limit the cost of detection. Another research point in this paper is to select probes efficiently under cost constraints. The contributions of this paper are as follows: (1) in order to solve the problem of energy consumption and low resource utilization in virtualized cloud computing data centers, a dynamic resource allocation method, DRAMDT. based on energy consumption cost, is designed in this paper. Considering that there is a large cut-off time difference between instances of virtual machines on a physical server, which may cause the server to remain in a state of incomplete utilization for a long time, resulting in a waste of energy consumption, In DRAMDT algorithm, virtual machine instances are grouped by the virtual machine resource subpool with time rotation. A virtual machine instance within a packet can be placed on the same physical server. Simulation results show that DRAMDT algorithm is effective and feasible. (2) based on the properties of information entropy, this paper proves a formula about the information entropy gain of probe. Based on this formula, an approximate method for calculating the information entropy gain of the probe is proposed. With this approximate calculation method, an efficient fault probe selection method, MEAP.MEAP algorithm, is proposed in this paper. The validity and feasibility of the algorithm are verified by the simulation experiments in this paper.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP3

【共引文献】