Deployment of Parallel Data Warehouse Joint Design in Teradata Ladjel Bellatreche 1 , Soumia Benkrid 2 , Ahmad Ghazal 3 , Alain Crolotte 3 , and Alfredo Cuzzocrea 4 1 LISI/ENSMA Poitiers University Futuroscope, France bellatreche@ensma.fr 2 National High School for Computer Science (ESI) Algiers, Algeria s benkrid@esi.dz 3 Teradata Corporation, San Diego, CA, U.S.A. (Ahmad.Ghazal, Alain.Crolotte)@teradata.com 4 ICAR-CNR and University of Calabria, Italy cuzzocrea@si.deis.unical.it Abstract. Parallel databases play an important role in delivering per- formance required by data warehousing applications. Such performance can not be accomplished without a good scheme for fragmentation and allocation of data. Both problems are NP-hard problems and optimal solutions are impractical in real life situations. These two problems were addressed extensively in research and there are a lot of heuristic based efficient solutions that provide good results. In these solutions, the frag- mentation and allocation were done separately. In our previous work, we proposed a genetic solution that solves both problems simultane- ously which we refer to as a the joint solution. The joint solution was tested on some known benchmarks using internal simulations of a paral- lel database. These simulations are not reliable since it lacked the real life aspects of parallelism. This paper addresses this issue by applying our joint solution on the Teradata DBMS. We experiment the SSB bench- mark (TPC-H like) on a Teradata appliance running TD 13.10. The re- sults shows a significant improvement over previous results that performs fragmentation and allocation sequentially. 1 Introduction Data warehousing is becoming more complex in terms of applications and data size. The parallelism is one of the relevant solutions to deal with mountains of data managed by warehouse and complex queries. More and more organizations are relying on parallel processing technologies to achieve the performance, scal- ability, and reliability they need [13]. Most of the major commercial database systems support parallelism (Teradata, Oracle, IBM, Microsoft, Sybase, etc.). Rather than relying on a single monolithic processor, parallel systems exploit fast and inexpensive micro processors to achieve high performance.