Everything should be as simple as possible, until no longer simple so far. I want to understand, fully understand the concept is to do one of the following prerequisites number of bins engineers.
First, the fact table for the measure
1. should try to come from the same underlying measurement results, a business process model stored in one dimension.
2. Each row in the table corresponds to the fact that a measure of the event. The data in each row is a certain level of detail data, referred to as granularity.
A measure of the physical world each event and the corresponding fact table rows with one to one relationship, this idea is the basic principle of dimensional modeling.
3. A fact table contains a plurality of external keys, foreign keys assembled in a primary key in the primary key table fact often referred to key combinations.
Second, the dimension table for the described environment
1. The quality of the data warehouse is directly dependent on the dimension setting properties.
2. Each dimension table is defined by a single primary key, as the reference basis for the integrity of the connection during operation the fact table.
3. dimension table usually expressed as a hierarchy.
4. Follow dimension table simplicity and accessibility, not required to meet the third paradigm, but also to minimize the snow mode.
Third, the connection dimension tables and fact tables
1. Each fact table contains the business process, the fact table around a plurality of dimension tables, a star connection.
2. The particle size of the smallest data or the largest dimension having atomic data.