Enterprise Data Management for Transaction and Analytical Processing - Benchmarking Transaction and Analytical Processing Systems - page 36

a

b

Fig. 2.12 Example ROLAP structures for the sales order example. ( a ) Star schema. ( b )Snowflake

schema

Analytical relational database schemas follow the dimensional modeling app-
roach and the according relational database schemas are the star schema and the
snowflake schema [ 54 ]. Figure 2.12 shows the star and snowflake schemas accord-
ing to the sales order example. The star schema (see Fig. 2.12 a) is composed of a
central fact table and multiple dimension tables. The fact table contains data from
business transactions or a snapshot summary of that data. It is connected through a
many-to-one relationship to each dimension table.
To speed up joins between the dimensions and the fact table, which is relevant
for every OLAP query, join and bitmap indexes are used [ 123 ]. The snowflake
schema is an extension to the star schema (see Fig. 2.12 b) to model hierarchies
explicitly, thus normalizing the dimensions. Thereby, redundant data storage within
the dimensions is reduced, but additional joins are needed if data from the outer
dimensions is required.
Even normalized schemas, for example, in third normal form (3NF) [ 39 ]are
employed in some analytical use cases, backing away from the dimensional model.
The different analytical database schemas are used in different kinds of data stores
that exist in the OLAP environment to cater for specific reporting needs. Martyn
[ 137 ] states that it is reasonable to use the star schema in a data mart, the snowflake
schema in a data warehouse and the 3NF schema in an ODS.