The primary OLAP object is the cube, a multidimensional representation of detail and summary data. A cube consists of a data source, dimensions, measures, and partitions. You design cubes based on the analytical requirements of users. A data warehouse can support many different cubes such as a Sales cube, an Inventory cube, and so on.
A cube’s data source identifies and connects to the database containing the data warehouse data that is the source of data for the cube.
Dimensions map data warehouse dimension table information into a hierarchy of levels, such as a Geography dimension with levels of Continent, Country, State-Province, and City. Dimensions can be independently created and shared among cubes for ease of cube construction and to ensure consistency of analysis data summarization. For example, if a shared dimension is used for a product hierarchy in all appropriate cubes, the organization of summarized product information will be consistent among the cubes that use the dimension.
A virtual dimension is a special type of dimension that maps the properties of members of another dimension into a dimension that can then be used in cubes. For example, a virtual dimension of a product’s size property enables a cube to summarize data such as sale quantity by product by size, such as the quantity of shirts sold by style by size. Virtual dimensions and member properties are evaluated as necessary for queries and they require no physical cube storage.
Measures identify the numerical values from the fact table that are summarized for analysis such as price, cost, or quantity sold.
Partitions are the multidimensional storage containers that hold cube data. Each cube contains at least one partition, and a cube’s data can be combined from multiple partitions. Each partition can take its data from a different data source and can be stored in a separate location. A partition’s data can be updated independently of other partitions in a cube. For example, a cube’s data can be divided by time, with a partition for current year’s data, another partition for the previous year’s data, and a third partition for all data prior to the previous year.
A cube’s partitions can be independently stored in different storage modes with different degrees of summarization. Partitions are invisible to the user, to whom the cube appears to be a single object, yet they provide the administrator with a wide variety of options to manage the underlying OLAP data.
Note User-defined partitions are available only if you install Microsoft® SQL Server™ OLAP Services, Enterprise Edition.
Roles enable the management of user access to cube data by mapping Microsoft Windows NT® user group and user accounts to cube access privileges.
A virtual cube is a logical view of portions of one or more cubes. A virtual cube can be used to join relatively unlike cubes that share a common dimension, such as a Sales cube and a Warehouse cube, for special analysis purposes while retaining the separate cubes for simplicity. Dimensions and measures can be selected from the joined cubes to be presented in the virtual cube.