基于度量波动时间框架的流立方体研究

摘要

流数据存在于很多动态环境中,一般具有多维属性,它能够实时描述系统状态,蕴含着大量信息.为了能近实时地对流数据进行分析,引入流立方体来对流数据进行建模,利用增量更新保证了立方体快速刷新.流数据的规模很大,为了节省存储空间,利用度量的波动性质提出一种新的时间框架.该时间框架能够在保持历史数据有效信息的前提下,缩减物化单元,减小立方体存储代价.%Data streams exist in many dynamic environments, usually they have the property of multiple dimensions.Filled with a great amount of information, data streams can describe system status in time.To analyze them approximately timely, we model the data stream with stream cube,and use incremental update to make the fast cube refreshing guaranteed.Since the size of data stream is so big, in order to minimize the memory cost, we propose a new time frame which takes the fluctuation property of measure into consideration.The time frame can reduce the memory cost of cube materialization on the premise of keeping valid and useful information of the history stream data.