Data redistribution

Data redistribution is a database administration operation that can be performed to primarily move data within a partitioned database environment when partitions are added or removed. The goal of this operation is typically to balance the usage of storage space, improve database system performance, or satisfy other system requirements.

Data redistribution within a partitioned database is done for one of the following reasons:

• To rebalance data whenever a new database partition is added to the database environment or an existing database partition is removed.
• To introduce user-specific data distribution across partitions.
• To secure sensitive data by isolating it within a particular partition.

Data redistribution is performed by connecting to a database at the catalog database partition and beginning a data redistribution operation for a specific partition group by using one of the supported interfaces. Data redistribution relies on the existence of distribution key definitions for the tables within the partition group. The distribution key value for a row of data within the table is used to determine on which partition the row of data will be stored. A distribution key is generated automatically when a table is created in a multi-partition database partition group. A distribution key can also be explicitly defined by using the CREATE TABLE or ALTER TABLE statements. By default during data redistribution, for each table within a specified database partition group, table data is divided and redistributed evenly among the database partitions. Other distributions, such as a skewed distribution, can be achieved by specifying an input distribution map which defines how the data is to be distributed. Distribution maps can be generated during a data redistribution operation for future use or can be created manually.