Exploiting hierarchical clustering in evaluating multidimensional aggregation queries

Multidimensional aggregation queries constitute the single most important class of queries for data warehousing applications and decision support systems. The bottleneck in the evaluation of these queries is the join of the usually huge fact table with the restricted dimension tables (star-join). Recently, a multidimensional hierarchical clustering schema for star schemas is suggested. Subsequently, query evaluation plans for multidimensional queries appeared that essentially implement a star join as a multidimensional range restriction. We present a number of transformations for such plans. The transformations place grouping/aggregation operations before joins and safely prune aggregated tuples. They can be applied at no or minimal extra I/O cost. We show how these transformations can be used to construct a new evaluation plan for grouping/aggregation queries over multidimensional hierarchically clustered schemas. The new plan improves previous results by grouping and aggregating tuples and by excluding aggregated tuples from further consideration at an early stage of the computation of a query.