Processing Relational OLAP Queries with UB-Trees and Multidimensional Hierarchical Clustering

2nd Intl. Workshop DMDW'2000 at CAiSE*00

计算机科学;社会科学（总论）

Volker Markl ; Rudolf Bayer

Others  :  http://CEUR-WS.org/Vol-28/paper1.pdf
PID  :  21263

来源: CEUR

【 摘 要 】

Multidimensional access methods like the UB-Tree can be used to accelerate almost any queryprocessing operation, if proper query processingalgorithms are used: Relational queries or SQLqueries consist of restrictions, projections,ordering, grouping and aggregation, and joinoperations. In the presence of multidimensionalrestrictions or sorting, multidimensional rangequery or Tetris algorithms efficiently processthese operations. In addition, these algorithmsalso efficiently support queries that generatesome hierarchical restrictions (for instance byfollowing 1:n foreign key relationships). In thispaper we investigate the impacts on queryprocessing in RDBMS when using UB-Trees andmultidimensional hierarchical clustering forphysical data organization. We illustrate thebenefits by performance measurements ofqueries for a star schema from a real worldapplication of a SAP business informationwarehouse. The performance results reported inthis paper were measured with our prototypeimplementation of UB-Trees on top of Oracle 8.We compare the performance of UB-Trees tonative query processing techniques of Oracle,namely access via an index organized table,which essentially stores a relation in a clusteredB*-Tree, and access via a full table scan of anentire relation. In addition we measure theperformance of the intersection of multiplebitmap indexes to answer multidimensional range queries.