Partitioning stores related groups of data in physical proximity to improve the performance of queries that are likely to require that data. To satisfy a query, Teradata Database can employ an optimization called “partition elimination” to limit data searches to only those partitions containing data relevant to the query.
A row-partitioned table assigns rows to a particular partition within an AMP based on a user-defined partitioning expression that defines how the rows should be grouped for storage (horizontal partitioning). The partitioning expression is defined when a table is created or altered.
The rows of a row-partitioned table are assigned to an appropriate partition based on the value of the partitioning expression.
Tables without primary indexes can also be partitioned by column (vertical partitioning). Whereas row partitioning allows sets of rows to be stored in separate partitions based on a partitioning expression, column partitioning allows sets of columns (including just a single column) to be stored in separate partitions. Like row partitioning, column partitioning can improve performance for some types of queries by allowing for partition elimination, whereby only the column data relevant to a particular query is searched during the processing of that query.