15.00 - Hash-Based Table Partitioning to AMPs - Teradata Database

Teradata Database Design

Teradata Database
User Guide

Hash-Based Table Partitioning to AMPs

All Teradata Database tables are partitioned, or distributed, horizontally by default. This means that the individual rows of each Teradata Database table are assigned to different AMPs in a way that balances their distribution equally in order to facilitate parallel processing.

Teradata Database uses the hash bucket portion of the computed rowhash value for each row to determine which AMP it should be assigned to (see “Teradata Database Hashing Algorithm” on page 225 and “Row Allocation for Teradata Parallel Data Pump” on page 237 for information on how this is done for nonpartitioned and column‑partitioned NoPI tables). If the primary index for a table is row‑partitioned, then further processing is required to compute which row partition it belongs to on its assigned AMP (see “Row‑Partitioned Row Allocation” on page 242). See “Row Allocation for FastLoad Operations Into Nonpartitioned NoPI Tables” on page 238 for a description of how Teradata Database FastLoads rows into nonpartitioned NoPI tables.