There are several techniques for storing multidimensional data structures based on the OLAP cube, Online Analytical Processing (OLAP): Multidimensional OLAP( MOLAP), Desktop OLAP( DOLAP), Relational OLAP (ROLAP), Hybrid OLAP( HOLAP). In ROLAP, relational OLAP, the data is completely stored in the database server.
Both the detailed data and the aggregations of the OLAP structure are stored in a relational database. For the detail data usually the tables are used, which are filled by the ETL process (extract/transform/load). For the aggregations, additional tables are created.
No user data is stored on the OLAP server. Instead, the OLAP server directly accesses the underlying relational database.
The graphic shows where the data is located in the different storage techniques.
The three storage techniques have different characteristics:
Relational OLAP (ROLAP) is therefore particularly suitable for multidimensional data with a large to very large volume, with few aggregation levels, e.g. for very extensive but simply structured accounting data.