Algebra-Based Approach for Incremental Data Warehouse Partitioning

Abstract

AbstractHorizontal Data Partitioning is an optimization technique well suited to optimize star-join queries in Relational Data Warehouses. Most works focus on a static selection of a fragmentation schema. However, due to the evolution of data warehouses and the ad hoc nature of queries, the development of incremental algorithms for fragmentation schema selection has become a necessity. In this work, we present a Partitioning Algebra containing all operators needed to update a schema when a new query arrives. To identify queries which should trigger a schema update, we introduce the notion of query profiling.