基于数据挖掘的数据中心能耗分析系统研究与开发
发布时间：2019-02-28 14:36
【摘要】：我们生活在一个数据信息的时代,人们的日常生活已经离不开数据和信息了,并且随着时间的推移,数据开始呈现出爆炸式增长的趋势。在数据急速增长的背后,完成对这些数据存储及加工的重任就交给了数据中心,数据中心为了应对数据大规模增长的趋势,它内部的IT设备及其它辅助设备的规模也会逐渐扩充,同时也会建造新的数据中心来保障新的需求。据估计,数据中心一年的电能消耗占全球总电能消耗的1.5%左右,相当于26个核电站一年的发电量,并且这个数字在未来还会增长。如果不能及时对数据中心的能耗加以管理、未采取合适的措施降低数据中心的能耗,那么就可能会出现能源紧缺,数据中心也不能及时的完成用户请求,这些都会影响到人们的日常生活的各个方面。 为了将数据中心变成“绿色”数据中心,我们需要研究并发现数据中心能耗的因素,通过合理的改善这些耗能因素降低数据中心的整体能耗。这些因素可能是环境因素,也可能是设备因素。本文针对数据中心能耗数据的分析开展了相关的研究工作。主要的工作包括： 1)通过数据挖掘聚类算法对数据中心内部设备的能耗进行聚类。由于同一种类型的设备能耗较为接近而不同设备类型的能耗差异较大,所以可以通过聚类结果发现某些异常耗能的设备,对这些设备加以改造改善其能耗。 2)通过数据挖掘分类和预测算法对数据中心的历史数据进行分类并对未来进行预测。这里提出了基于数据中心的历史数据的分析对未来一段时间内的能耗或业务请求量等的预测,可以通过预测的结果,控制数据中心内部一些设备的状态(如开启或关闭),通过这种手段控制数据中心的能耗。 3)建设开发能耗分析系统,使得算法可以运行在系统之上,并得以实际的应用。使用人员通过系统的使用,方便他的操作,系统也增强了人机的交互性。目前,系统的一期已经建设完毕,系统的功能包括统计分析,聚类分析、分类和预测分析三大模块。每个模块可以应用于不同的场景对能耗数据进行分析。
[Abstract]:We live in an era of data and information, people's daily life has become inseparable from data and information, and with the passage of time, the data began to show an explosive growth trend. Behind the rapid growth of data, the task of storing and processing these data is entrusted to the data center, which in order to cope with the trend of large-scale data growth, Its internal IT equipment and other ancillary equipment will also grow in size, and new data centers will be built to support new needs. It is estimated that data centers consume about 1.5 percent of the world's electricity a year, the equivalent of 26 nuclear power plants a year, and that number will grow in the future. If energy consumption in the data center is not managed in a timely manner and appropriate measures are not taken to reduce the energy consumption in the data center, there may be energy shortages and the data center will not be able to complete user requests in a timely manner. These will affect every aspect of people's daily life. In order to turn the data center into a "green" data center, we need to study and identify the energy consumption factors in the data center and reduce the overall energy consumption of the data center by reasonably improving these energy consumption factors. These factors may be environmental factors or equipment factors. In this paper, the energy consumption data analysis of the data center carried out related research work. The main work includes: 1) clustering energy consumption of internal equipments in data center through data mining clustering algorithm. Because the energy consumption of the same type of equipment is close and the energy consumption of different equipment types is quite different, we can find out some equipment with abnormal energy consumption by clustering results, and improve the energy consumption of these devices. 2) classify the historical data of data center and forecast the future by data mining classification and prediction algorithm. Here is the analysis of the historical data based on the data center, which can control the status of some devices (such as on or off) in the data center by forecasting the energy consumption or the quantity of business requests, etc., in the coming period, and the result of the prediction can be used to control the status of some devices in the data center. Control the energy consumption of the data center in this way. 3) build and develop the energy consumption analysis system, so that the algorithm can run on the system, and can be applied in practice. User through the use of the system, convenient for his operation, the system also enhanced the human-computer interaction. At present, the first phase of the system has been completed. The functions of the system include statistical analysis, cluster analysis, classification and prediction analysis. Each module can be applied to different scenarios to analyze energy consumption data.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.13;TP308
本文编号：2431886
[Abstract]:We live in an era of data and information, people's daily life has become inseparable from data and information, and with the passage of time, the data began to show an explosive growth trend. Behind the rapid growth of data, the task of storing and processing these data is entrusted to the data center, which in order to cope with the trend of large-scale data growth, Its internal IT equipment and other ancillary equipment will also grow in size, and new data centers will be built to support new needs. It is estimated that data centers consume about 1.5 percent of the world's electricity a year, the equivalent of 26 nuclear power plants a year, and that number will grow in the future. If energy consumption in the data center is not managed in a timely manner and appropriate measures are not taken to reduce the energy consumption in the data center, there may be energy shortages and the data center will not be able to complete user requests in a timely manner. These will affect every aspect of people's daily life. In order to turn the data center into a "green" data center, we need to study and identify the energy consumption factors in the data center and reduce the overall energy consumption of the data center by reasonably improving these energy consumption factors. These factors may be environmental factors or equipment factors. In this paper, the energy consumption data analysis of the data center carried out related research work. The main work includes: 1) clustering energy consumption of internal equipments in data center through data mining clustering algorithm. Because the energy consumption of the same type of equipment is close and the energy consumption of different equipment types is quite different, we can find out some equipment with abnormal energy consumption by clustering results, and improve the energy consumption of these devices. 2) classify the historical data of data center and forecast the future by data mining classification and prediction algorithm. Here is the analysis of the historical data based on the data center, which can control the status of some devices (such as on or off) in the data center by forecasting the energy consumption or the quantity of business requests, etc., in the coming period, and the result of the prediction can be used to control the status of some devices in the data center. Control the energy consumption of the data center in this way. 3) build and develop the energy consumption analysis system, so that the algorithm can run on the system, and can be applied in practice. User through the use of the system, convenient for his operation, the system also enhanced the human-computer interaction. At present, the first phase of the system has been completed. The functions of the system include statistical analysis, cluster analysis, classification and prediction analysis. Each module can be applied to different scenarios to analyze energy consumption data.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.13;TP308
【参考文献】
相关期刊论文 前8条
1 周雄;董威;;Apriori模型在电力企业决策支持中的应用[J];电力建设;2007年12期
2 温鹏,万永华,汪同庆,郑俊杰,杨立常;电力系统短期负荷离线预测研究[J];电力系统自动化;1998年10期
3 姚李孝,姚金雄,李宝庆,万诗新;基于竞争分类的神经网络短期电力负荷预测[J];电网技术;2004年10期
4 林闯;田源;姚敏;;绿色网络和绿色评价:节能机制、模型和评价[J];计算机学报;2011年04期
5 陈德军;胡华成;周祖德;;基于径向基函数的混合神经网络模型研究[J];武汉理工大学学报;2007年02期
6 舒正渝;;浅谈数据挖掘技术及其应用[J];中国西部科技;2010年05期
7 邓宏贵;罗安;曹建;;电力负荷预估的神经网络正则化及其应用[J];小型微型计算机系统;2006年08期
8 谢宏,程浩忠,张国立,牛东晓,杨镜非;基于粗糙集理论建立短期电力负荷神经网络预测模型[J];中国电机工程学报;2003年11期
，本文编号：2431886
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2431886.html
