What can partitioning do for your data warehouses and data marts?
Authors:
- L. BellatrecheLBL. Bellatreche
- K. KarlapalemKKK. Karlapalem
- M. MohaniaMMM. Mohania
- M. SchneiderMSM. Schneider
- 7 November 2002
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 437-445
Abstract
Efficient query processing is a critical requirement for data warehousing systems as decision support applications often require minimum response times to answer complex, ad-hoc queries having aggregations, multi-ways joins over vast repositories of data. This can be achieved by fragmenting warehouse data. The data fragmentation concept in the context of distributed databases aims to reduce query execution time and facilitates the parallel execution of queries. In this paper, we propose a methodology for applying the fragmentation technique in a data warehouse star schema to reduce the total query execution cost. We present an algorithm for fragmenting the tables of a star schema. During the fragmentation process, we observe that the choice of the dimension tables used in fragmenting the fact table plays an important role on overall performance. Therefore, we develop a greedy algorithm in selecting "best" dimension tables. We propose an analytical cost model for executing a set of OLAP queries on a fragmented star schema. Finally, we conduct some experiments to evaluate the utility of the fragmentation for efficiently executing OLAP queries.Keywords
This publication has 8 references indexed in Scilit:
- OLAP query processing for partitioned data warehouses2003
- A case for parallelism in data warehousing and OLAP2002
- Index selection for OLAP2002
- Algorithms and Support for Horizontal Class Partitioning in Object-Oriented DatabasesDistributed and Parallel Databases, 2000
- A horizontal fragmentation algorithm for the fact relation in a distributed data warehouse1999
- Index merging1999
- Faster joins, self-joins and multi-way joins using join indicesData & Knowledge Engineering, 1998
- Improved query performance with variant indexes1997