Divide and Aggregate: Caching Multidimensional Objects

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

计算机科学;社会科学（总论）

Wolfgang Lehner ; Jens Albrecht ; Wolfgang Hümmer

Others  :  http://CEUR-WS.org/Vol-28/paper2.pdf
PID  :  21264

来源: CEUR

【 摘 要 】

A common optimization technique in the OLAP application domain is the use of summary aggregate data. For an appropriate support of analysis hot spots, we propose a query based aggregate cache of 'multidimensional objects'. The major drawback of previous query caching methods is that new queries must be completely subsumed by a single cached object in order to utilize it. Our solution is the introduction of 'set-derivability' which allows the combination of several aggregates to derive a single query. The set of cached multidimensional objects is dynamically determined using both users access patterns and semantical knowledge of dimensional structures. Experimental results show that average cost reductions of over 50% are easily reached while spending only 10% of additional storage for summary data.