Database designers and administrators have been partitioning tables since long before Oracle8 hit the scene. Generally, table partitioning within a single database is done to improve performance and simplify administration tasks, while table partitioning between databases is meant to facilitate data distribution. For example, sales data might be partitioned by region and each partition hosted in a database housed at its respective regional sales office. Whereas a central data warehouse might gather sales data from each office for reporting and decision-support queries, it might be perfectly reasonable for the operational sales data to be distributed across multiple sites.
Partitioning by sets of rows such as in the sales data example, in which the value of the sales office column determines where the data resides, is known as horizontal partitioning. Partitioning may also be accomplished by splitting up sets of columns, in which case it is called vertical partitioning. For example, sensitive data such as salary information and social security numbers may be split off from the employee table into a separate table with restricted access. When partitioning vertically, primary key columns must be included in the set of columns for every partition. Therefore, unlike horizontal partitioning, where each partition contains nonoverlapping subsets of data, vertical partitioning mandates that some data be duplicated in each partition.
While both vertical and horizontal partitioning may be accomplished manually within and between Oracle databases, the Partitioning Option introduced in Oracle8 specifically deals with horizontal partitioning within a single database.