Query Optimization: Eager and Lazy Aggregation

P-A (Paul) Larson

University of Waterloo
Currently on leave at IBM Santa Theresa


Efficient processing of aggregation queries is essential for decision support applications. This talk describes a class of query transformations, called eager aggregation and lazy aggregation, that allows a query optimizer to move group-by operations up and down the query tree. For some queries, this type of transformation can reduce the processing time by 90% or more. Eager aggregation partially pushes a group-by past a join and begins to produce partial groups early, thereby reducing the number of rows flowing into subsequent operations. The reverse transformation, lazy aggregation, moves a group-by past a join and combines two group-by operations into one. This transformation is typically of interest when an aggregation query references a grouped view (a view containing a group-by). We show that eager/lazy aggregation can be applied if certain (derived) functional dependencies are guaranteed to hold in the join result. The talk will also show that sometimes we have several ways of transforming a query and outline a practical algorithm for finding all valid transformations. This is joint work with Paul Yan, University of Waterloo.

Publications Related to Talk

Please contact Paul Yan (pwyan@bluebox.uwaterloo.ca) if you have specific questions about the talk.