摘要
为解决传统行存储结构导致OLAP聚集计算效率低下的问题,设计了基于维存储的OLAP数据存取技术.首先,将OLAP事实表中的维属性集和度量属性集定义为2个列族,每张维表的所有属性定义为1个列族.对维表进行二进制编码,生成维层次编码,从而保持了维的层次语义特性.以(维层次编码,度量值)对形式按列组织数据,消除查询时维表与事实表的复杂连接操作运算.然后,采用自底向上方法构建B+树,对维层次编码进行索引,加快了数据读取效率.通过增删事实表和维层次编码-度量表中相应的列,实现维和度量的增加和删除.性能分析结果表明,这种OLAP数据存取技术具有良好的可扩展性,能高效地管理和存取OLAP海量多维数据,有效支持上层OLAP聚集计算.
To improve the efficiency of OLAP( online analytical processing) aggregation in tradition- al row-stores, a dimension-stores based OLAP data access technology is designed. First, the dimen- sion attributes and measure attributes of the OLAP fact table are defined as two column families, and all attributes in a dimension table are defined as one column family. Dimension tables are encoded by using binary digit to generate dimension hierarchy codes, and thus the hierarchy semantics of dimen- sion is maintained. The data are organized as (dimension hierarchy code, measure value) to elimi- nate the complex join operations between dimension tables and fact table in query. Then, a B + tree index is bottom-up built to index the dimension hierarchy codes, which accelerates the data access efficiency. Adding and deleting the corresponding columns in fact table and dimension hierarchy code-measure tables can realize the addition and deletion of dimensions and measures. The performance analysis results show that this OLAP data access technology has good expandability. It can effi- ciently manage and access massive OLAP multidimensional data, and effectively supports the upper OLAP aggregation.
出处
《东南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2012年第5期797-802,共6页
Journal of Southeast University:Natural Science Edition
基金
国家重点基础研究发展计划(973计划)资助项目(2010CB328104)
国家自然科学基金资助项目(61070161
60903161
60903162
61003257)
"十一五"国家科技支撑计划资助项目(2010BAI88B03
2011BAK21B02)
高等学校博士点学科专项科研基金资助项目(20110092130002)
江苏省自然科学基金资助项目(BK2008030)
东南大学江苏省网络与信息安全重点实验室资助项目(BM2003201)
东南大学教育部计算机网络与信息集成重点实验室资助项目(93K-9)
浙江师范大学计算机软件与理论省级重中之重学科开放基金资助项
关键词
OLAP
HBASE
维存储
B+树
OLAP( online analytical processing)
HBase
dimensional-stores
B + tree
作者简介
宋爱波(1970-),男,博士,副教授,absong@seu.edu.cn