分布式文件系统数据块聚类存储节能策略

发布时间：2018-04-11 02:20

本文选题：云计算 + 分布式文件系统　；参考：《计算机应用》2015年02期

【摘要】：针对分布式文件系统中由于数据块随机放置而导致的服务器利用率低、能耗管理复杂的问题,建立了数据块访问特征向量模型描述用户对数据块的随机访问,运用K-means算法对数据块进行聚类计算,根据计算结果将数据节点划分为多个区域以存储不同聚类簇的数据块,在系统负载较低时进行数据块动态重配置,关闭不必要节点达到节能的目的。为使得策略适用于对能耗和资源利用率有不同要求的场景,算法中聚类簇间隔参数可灵活设置。实验通过和冷热区划分算法进行比较表明:按照聚类结果进行数据块重配置后,能耗节省效率优于冷热区划分算法,节省能耗35%~38%。
[Abstract]:The low utilization rate for the distributed file system by data blocks randomly placed in the server, the management of energy consumption complex problems, establish a data block access feature vector model to describe the random access to user data blocks, K-means clustering algorithm is used to calculate the data block according to the calculation results, the data node is divided into a plurality of areas to a block of data stored in different clusters, the data block dynamic reconfiguration in the system load is low, turn off unnecessary nodes to achieve the purpose of saving energy. In order to make the strategies applicable to the use of resources and energy consumption rate have different requirements of the scene, algorithm of cluster interval parameters can be set flexibly. Compared by experiment and hot and cold division algorithm: data block reconfiguration according to the clustering results, energy saving efficiency is better than that of hot and cold region partition algorithm, energy saving 35%~38%.

【作者单位】：新疆大学信息科学与工程学院;新疆大学软件学院;
【基金】：国家自然科学基金资助项目(61462079,61262088,61363083) 新疆维吾尔族自治区自然科学基金资助项目(2013211A011)
【分类号】：TP333

【参考文献】