Row Allocation for FastLoad Operations Into Nonpartitioned NoPI Tables
When you use the FastLoad utility to bulk load rows into an nonpartitioned NoPI table, Teradata Database accesses the configuration map and builds a list of system AMPs, including offline AMPs.
A round robin protocol sends the first buffer of rows to its own AMP. The next buffer is assigned to the next highest AMP, and so on. When the highest AMP is reached, Teradata Database wraps the list around and selects the lowest AMP again. For a down AMP, Teradata Database automatically redirects the buffer to the fallback owner of the assigned hash bucket, and the receiver inserts the rows into the fallback subtable. Down AMP recovery occurs when the down AMP comes back online.
To prevent row redistributions for NoPI tables during reconfiguration and restore operations, AMPs build a separate hash map to be used specifically for NoPI tables.