基于数据挖掘的服务器性能优化技术研究

发布时间：2018-06-24 11:46

本文选题：关联规则 + Apriori　；参考：《吉林农业大学》2013年硕士论文

【摘要】：随着计算机技术的发展,互联网提供各式各样的服务,而这些服务程序在服务器上运行。有些服务器运行着多种服务,但是这些服务在一台服务器上运行是否合理,是否能提供稳定的服务值得思考。然而有些服务需要很多台服务器来支撑,如google搜索引擎需要上万台服务器来处理用户的检索请求。为了应对用户高峰,很多服务采用多服务器来应对高峰来临时提供稳定的服务,这也带来了一个值得思考的问题,在用户较少时很多服务器资源也随着浪费。为了解决上述问题,我们采用数据挖掘技术进行分析。在研究多个服务在一台服务器上运行是否合理的问题时,我们引入关联关系分析来找出每个服务对服务器资源的使用情况,针对需要挖掘的数据维度较小的特点,提出基于字典树和倒排索引技术优化的Apriori算法,在研究集群服务器优化时,提出了一种动态加权负载均衡算法,以避免请求集中在某几个服务器的现象。为了提供稳定的服务,我们引入了基于规则的故障预警模块,该模块可以监测服务器状态,并且根据规则分类服务器当前的危险级别,当达到指定危险级别时通知管理员,并及时处理以避免服务不稳定或服务不可用的情况。本文最后通过实际数据分析提出一些在服务器空闲时有效提高服务器资源利用率的方法,包括单服务器和集群服务器的情况个给出建议。
[Abstract]:With the development of computer technology, the Internet provides a variety of services, and these service programs run on the server. Some servers run multiple services, but it is worth thinking about whether they work on a single server and provide stable services. However, some services need many servers to support, such as the google search engine requires tens of thousands of servers to handle users' retrieval requests. In order to cope with the peak of users, many services use multi-servers to cope with the peak to provide stable services, which also brings a problem worth thinking, when the number of users is less, a lot of server resources are also wasted. In order to solve the above problems, we use data mining technology to analyze. When we study whether it is reasonable for multiple services to run on a single server, we introduce the relational analysis to find out the usage of server resources by each service, aiming at the small dimension of data to be mined. This paper presents a Apriori algorithm based on dictionary tree and inverted index technology. In order to avoid the phenomenon that requests are concentrated on several servers, a dynamic weighted load balancing algorithm is proposed in the research of cluster server optimization. In order to provide stable service, we introduced a rule-based fault warning module, which can monitor the status of the server and classify the server's current danger level according to the rules, and notify the administrator when the specified danger level has been reached. And timely handling to avoid service instability or service unavailable situation. Finally, this paper puts forward some methods to improve the utilization of server resources when the server is idle, including the case of single server and cluster server.
【学位授予单位】：吉林农业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.13

【参考文献】