一种大时间窗口StreamCube体系结构

An Architecture of Big-Windowed Stream Cube

目前,在大时间窗口的实时Cube查询中,若大时间窗口数据方切片容量超过DSMS的内存限制,则无法在DSMS保存的StreamCube中得到完整正确结果。提出一种采用混合数据库模式(DSMS和DBMS)存储StreamCube体系结构HDS-Cube(Hybrid Database Stream Cube),即时间片流数据方将StreamCube切分为按时间等分的数据方切片,并按时间规则分别存放在两种数据库中。StreamCube查询过程中读取两种数据库中的数据方时间切片,得到正确结果。实验结果表明,该体系结构能够高效的支持实时联机在线分析。

Currendy, big-windowed Cube queries on Stream Cube that implemented on DSMS could not be responded completely and exactly because memory capability on DSMS limits the capability of big-windowed Cube slices. We proposed architecture of big-windowed StreamCube, Hybrid Database StreamCube, which uses hybrid database pattern （DSMS ＆ DBMS）. Time-Sliced StreamCube partitioned StreamCube into Cube slices by time dimension. Slices will be stored respectively in two types of database according to time rule. When querying in StreamCube, system can get the exact results from slices in hybrid database. Through extensive experiments, it is shown that HDS-Cube can support real-time OLAP efficiently.