Entrepôts, Représentation et Ingénierie des Connaissances
Publications du laboratoire

Recherche approfondie

par Année
par Auteur
par Thème
par Type
- Intentional Data Placement Policy for Improving OLAP Cube Construction on Hadoop Clusters hal link

Auteur(s): Arres B., Kabachi N.(Corresp.), Boussaid O.(Corresp.), Bentayeb F.(Corresp.)

(Affiches/Poster) 30ème édition de la Conférence Base Données Avancées - BDA 2014 (Autrans, FR), 2014-10-14

Ref HAL: hal-01166222_v1

In the recent past, we have witnessed dramatic increases in the volume of data literally in every area: business, science,and daily life to name a few. The Hadoop framework - an open source project based on the MapReduce paradigm - isa popular choice for big data analytics. However, the performance gained from Hadoop’s features is currently limitedby its default block placement policy, which does not take any data characteristics into account. Indeed, the efficiencyof many operations can be improved by a careful data placement, including indexing, grouping, aggregation and joins.In our work we propose a data warehouse partitioning strategy to improve query gain performances. We investigatethe performance gain for OLAP cube construction with and without data organization on a Hadoop cluster. And this,by varying the number of nodes and data warehouse size.Our experiments suggest that a good data placement on acluster during the implementation of the data warehousecan significantly increase the OLAP cube construction andquerying performances. In the next step, we will extendthe experiments to study the effects of other configurationparameters on collocation data in the context of paralleldata warehousing, such as partitions size, replication factorand OLAP query complexity. We plan also to study an in-telligent system for warehouses data placement on clustersby integrating Multi-Agent System (MAS) and IntelligentAgents to the process.