Reporting and BI applications need to transform their data from raw database form into denormalized structures, dimensional model, or a cube. They also must also apply filtering, aggregation, and business transformations.
Materialized views are being misused in order to skip ETL step. This is a simplistic approach & it does not work since often the developers do not understand the data transformation needs of the business. The cost is paid in the form of a much more expansive and expensive reporting & BI development cycle.
The more cost-effective approach is to use an extended hub-and-spoke model in which the ETL step leads to further spokes in the form of data marts and cubes.
No comments:
Post a Comment