Mastering Data Warehouse Aggregates: Solutions for Star Schema Performance

7.1. Data Warehouse Implementation

Dimensional data warehouses are usually implemented in parts. Known as a data mart, each part is organized around a subset of the business processes that constitute the enterprise. Ideally, data marts are implemented according to a master plan, around a set of conformed dimensions. This ensures that as each data mart is brought on-line, it interoperates with the others. However, data marts are sometimes built without consideration for the rest of the enterprise, or as an analytic adjunct to a normalized repository of enterprise data.

Regardless of the overall approach, these dimensional databases are implemented as a series of data marts. This incremental approach provides an opportunity to delay the consideration of aggregates until after the first project. The scope of the initial project is reduced, allowing the business to become familiar with the data warehouse lifecycle. A subsequent project introduces aggregates to the first data mart. Future implementations include plans to incorporate aggregates, which will also be re-evaluated on a periodic basis.

7.1.1. Incremental Implementation of the Data Warehouse

One of the advantages of the dimensional approach to data warehouse design is its ability to scale from a single subject area to the enterprise level. By planning a series of implementations around a framework of conformed dimensions, the data warehouse is delivered incrementally. This avoids the risks associated with a single project ...