Group Data Warehouse

2005-2006 For a major Austrian bank we take the responsibility to design a new group data warehouse as a base for all Basel II reporting tasks.

On the system architecture side, we redesign the existing SMP based system and run a comparison avaluation between the Unix based system and a new Linux based one. The later shows a very high stability and a great performance boost, so the final decision goes to a clustering (Oracle RAC), linux based soluton, running on standard, extremely cost eficient hardware. We design the underlying IO subsystem from scratch, achieving an significant performance optimizations. At the end the overall performance increase is in the range of 4 to 50 times (measured with production ETL jobs running once on the legacy Unix based system and the same jobs running on the new Linux based cluster solution with the new underlying IO subsystem)

On the software architecture side we introduce new techniques for all ETL related steps. Most of the developing effort is optimized by model driven architecture (code generation based on central managed metadata) , big parts of the system could be implemented by using ETL tools.