![]() Previous |
![]() Next |
A relational database typically stores data values in tables that represent third normal form data. In this type of implementation, the values of key columns of a relational database table are unique values of a single level of data. For example, at one level in the relational database you might have a table with a key column named City that contains the names of cities and at the next highest level in the database a table with a key column named state that contains the names of states, and so on and so on.
In an analytic workspace the objects that hold the data to analyze are arrays called variables. The keys into variables are stored in other objects which act as the dimensions of the variables. To support performant OLAP analysis, values from multiple levels are stored within a single dimension called a hierarchical dimension. For example, an analytic workspace might have a hierarchical dimension named geog that had as values the names of both cities and states.
The objects that store values that relate values of two or more dimensions are called relations. Thus the one-to-many relationship between values of different levels in a hierarchical dimension are stored in an analytic workspace. For example, the relationship between the city and state values in a hierarchical geog dimension would be stored in an analytic workspace relation typically called a parentrel relation. (See "Parentrel Relation" for more information.)
Additional analytic workspace objects are typically defined to keep additional information about the hierarchical dimension. Several important OLAP DML commands and functions (such as the LIMIT command) presume the existences of these objects in your analytic workspace as the name of these objects is one argument in the syntax of the statement.
The data objects that you define using the OLAP DML are multidimensional objects that are stored in an analytic workspace. When you use OLAP DML statements to perform operations against these multidimensional data objects, those operations apply all at once to entire set of values contained by these objects.