Cubes, dimensions, hierarchies, and measures are the essence of the multidimensional navigation of OLAP.
By describing and presenting data in this fashion, users can navigate a complex set of data intuitively. However, describing the data model intuitively does little to speed delivery of the information to the user.
A key tenet of OLAP is that users should see consistent response times for each view, or slice, of the data they request. Because data is usually collected at the detail level only, the information summary usually is computed in advance. These precomputed values, or aggregations, are the basis of the OLAP performance gains.
In the early days of OLAP technology, most vendors assumed that the only possible solution for OLAP applications was a specialized, nonrelational storage model. Later, other vendors discovered that through the use of database structures (star and snowflake schemas), indexing, and storage of aggregates, relational database management systems (RDBMS) could be used for OLAP. These vendors called their technology Relational OLAP (ROLAP). The earlier OLAP vendors then adopted the term multidimensional OLAP (MOLAP).
MOLAP implementations usually outperform ROLAP technology, but have problems with scalability. On the other hand, ROLAP implementations are more scalable and are often attractive to customers because they leverage investments in existing relational database technology.
A recent development has been a hybrid OLAP (HOLAP) solution, which combines the ROLAP and MOLAP architectures to yield a solution with the best features of both: superior performance and extensive scalability. One approach to HOLAP maintains detail records (the largest volumes) in the relational database, while maintaining aggregations in a separate, MOLAP store.