Incremental Maintenance for Non-Distributive Aggregate Functions

Incremental view maintenance is a well-known topic that has been addressed in the literature as well as implemented in database products. Yet, incremental refresh has been studied in depth only for a subset of the aggregate functions. In this paper we propose a general in-cremental maintenance mechanism that applies to all aggregate functions, including those that are not distributive over all operations. This class of functions is of great interest, and includes MIN/MAX, STDDEV, correlation, regression , XML constructor, and user deened functions. We optimize the maintenance of such views in two ways. First, by only re-computing the set of aaected groups. Second, we extend the incremental infrastructure with work areas to support the maintenance of functions that are algebraic. We further optimize computation when multiple dissimilar aggregate functions are computed in the same view, and for special cases such as the maintenance of MIN/MAX, which are incrementally main-tainable over insertions. We also address the important problem of incremental maintenance of views containing super-aggregates, including materialized OLAP cubes. We have implemented our algorithm on a prototype version of IBM DB2 UDB, and an experimental evaluation proves the validity of our approach. Work done while the author was visiting IBM Almaden Research Center. Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the VLDB copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Very Large Data Base Endowment. To copy otherwise, or to republish, requires a fee and/or special permission from the Endowment.