Another question I received had to do with placement of the data warehouse and the potential myriad of other databases (staging, data marts, ODS, etc.) in the data warehouse architecture. The recommended approach by the questioner's management was to place them all on the same DBMS instance -- the same instance that the main operational source system was on!
There are several aspects to this approach that are not recommended. Firstly, mixing operational and data warehouse workloads on the same instance creates impact on each other and limits what you can aspire to with the data warehouse over time -- even if it works initially. Perhaps, but still not recommended, separate partitions on a highly partitionable machine could be used to get the effort started.
Depending on the scalability of the chosen data warehouse technology, it is usually recommended that the various data warehouse databases also habitate physically independent in separate partitions or machines.
You may or may not decide to have a physically separate staging area (landing ground and transformation area for raw operational, pre-warehouse data), but few are able to take raw operational data and move it directly into a warehouse schema (especially a dimensional one) without some interim tables. These tables can simply be non user-accessible tables in the warehouse, but where I can take transformation cycles off the warehouse machine, I would do so. Hence, a physically
Similarly, data marts can reside in the data warehouse instance but here again is an opportunity to take cycles, this time access cycles, off the warehouse and keep the warehouse database to be the database with the accessible large-scale atomic data whose primary function is the distribution point for the data marts.
Read part one of this tip, Backing up the data warehouse.
For more information, check out SearchCRM's Best Web Links on Data Warehousing.
This was first published in May 2002