摘要
数据流管理系统计算聚集查询结果保存在内存中形成流数据方(StreamCube),提供快速、精确的在线OLAP查询。有限的内存空间需要一种有效的存储方法来存储更大时间窗口的流数据方。提出一种基于QC-Tree结构的流数据方StreamQCTree生成、裁剪及查询方法。将QC-Tree结构中上界集划分为基本上界类和附加上界类;并分析附加上界类的成本计算模型;根据该模型在固定存储空间下,采用动态选择物化结点的方案选择物化部分附加上界类,使对StreamQCTree的平均查询响应时间最小。实验表明,StreamQCTree能够有效地访问数据方且获得较好的压缩效果。
StreamCube,which responses OLAP queries fast and accurately,is in-memory and composed of Group-bys from DSMS.Becasue of limited capacity of memory,it needs an efficient structure to keep more information of StreamCube with more large time window.This paper presents a QCTree-based structure,StreamQCTree,with constructing,pruning and search algorithm.The upper bounds in QC-Tree are partitioned into two classes：Basic Upper Bounds（BUB） and Addition Upper Bounds（AUB）,and cost model of AUB is analyzed.Using the cost model,a dynamic select approach is put forward to choose the AUBs with high cost-benefit in fixed memory,which gains less average response time for all queries in StreamQCTree.Experiments show that StreamQCTree performs well in compressing StreamCube and make queries efficiently.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第19期140-143,185,共5页
Computer Engineering and Applications
基金
国家高技术研究发展计划(863)No.2007AA01Z474
No.2007AA010502~~
