Implementation and Selection of Derived Partitioning for Optimizing the Performances of Relational Data Warehouses

摘要

Many works have been done to optimize the performances of relational data warehouses. Three main techniques are available in relational DBMS or in specific systems : indexes including join indexes, materialized views, partitioning. Each one can be used alone, or combined with the others. In this paper we will consider more precisely the partitioning technique. In general, only the direct partitioning of a relation (i.e. according to the attributes of this relation) is possible. However, in the context of a warehouse, it would be interesting to be able to implement the derived partitioning which consists in partitioning a relation according to the attributes of another relation referenced by the first. We show how to implement a derived partitioning with a relational DBMS and we establish its interest through several experiments.