云数据中心计算资源监控系统的设计与实现

发布时间：2018-01-06 17:10

本文关键词：云数据中心计算资源监控系统的设计与实现　出处：《电子科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：近年来，，云计算发展迅速，是当今计算机研究的前沿课题，而对于云数据中心的监控也逐渐受到人们的关注。传统的监控方式由于实时性，安全性，扩展性以及访问量过大导致过载等问题不适合当前的云监控。本文结合云计算的特点，着重对云数据中心的计算资源，当前主要是CPU，内存和网络的资源情况，进行监控研究和实现。在分析了网络监控系统的相关理论及体系结构的基础上，设计了C/S和B/S混合结构的云监控系统，其核心思想为利用C/S的单机处理能力，对本节点上的指定信息进行获取，发送给服务端；服务端负责某个局域网内的监控，接受客户端的信息，存入数据库，并封装转发给SuperServer；SuperServer负责整个云监控，接受数据并进行相关分析和处理。也利用了B/S的远程管理，使得管理员能通过网页界面远程查看和管理云内服务器的逻辑结构，系统信息等。本文还对某云系统的用户请求信息进行了展示，并用排队论M/M/C模型分析此情况下的系统排队情况，包括用检验和极大似然拟合得出输入参数，计算出排队队长等参数并给出要求拒绝率低于1%时的最少服务器数量。最后研究服务器功耗和处理器利用率之间的关系，得到功率函数，利用监控到的利用率信息计算出服务器的功耗并和用功耗测试仪监控到的进行比对和分析，研究表明误差范围低于5%，该方法有较大可行性。通过云数据中心计算资源监控系统的研究与设计，一方面探讨研究了云监控系统实现过程中遇到的问题；另一方面能够对未来商业化的云监控提供一定的借鉴和帮助。
[Abstract]:In recent years, cloud computing is developing rapidly, which is the frontier of computer research, and the monitoring of cloud data centers has gradually attracted people's attention. Traditional monitoring methods because of real-time, security. Problems such as extensibility and overloading due to heavy traffic are not suitable for current cloud monitoring. Combined with the characteristics of cloud computing, this paper focuses on the cloud data center computing resources, mainly CPU, memory and network resources. On the basis of analyzing the related theory and architecture of the network monitoring system, the cloud monitoring system with C / S and B / S hybrid structure is designed. The core idea is to obtain the specified information on this node and send it to the server by using the single processing ability of C / S; The server is responsible for monitoring the local area network, receiving the information from the client, storing the information in the database, and encapsulating and forwarding it to SuperServer. SuperServer is responsible for monitoring the entire cloud, receiving data, and analyzing and processing it. It also takes advantage of the remote management of B / S. It enables the administrator to remotely view and manage the logical structure and system information of the cloud server through the web interface. In this paper, the user request information of a cloud system is also displayed, and the queueing condition of the system is analyzed by using the queue theory M / M / C model, including the input parameters obtained by test and maximum likelihood fitting. The queue length and other parameters are calculated, and the minimum number of servers when the rejection rate is less than 1 is given. Finally, the relationship between server power consumption and processor utilization is studied, and the power function is obtained. The power consumption of the server is calculated and compared with that monitored by the power consumption tester. The results show that the error range is less than 5 and the method is feasible. Through the research and design of cloud data center computing resource monitoring system, on the one hand, the problems encountered in the implementation of cloud monitoring system are discussed. On the other hand, it can provide some reference and help for future commercial cloud monitoring.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP308;TP277

【相似文献】