Rows are distributed across the AMPS using a hashing algorithm that computes a row hash value based on the primary index. The row hash is a 32-bit value. Depending on the system setting for the hash bucket size, either the higher-order 16 bits or the higher-order 20 bits of a hash value determine an associated hash bucket.
Normally, the hash bucket size is 20 bits for new systems. If you are upgrading from an older release, the hash bucket size may be 16 bits. If a 20-bit hash bucket size is more appropriate for the size of a system, you can use the DBS Control Utility to change the system setting for the hash bucket size.
|IF the hash bucket size is …||THEN the number of hash buckets is …|
The hash buckets are distributed as evenly as possible among the AMPs on a system.
Teradata Database maintains a hash map—an index of which hash buckets live on which AMPs—that it uses to determine whether rows belong to an AMP based on their row hash values. Row assignment is performed in a manner that ensures as equal a distribution as possible among all the AMPs on a system.
For details about hash bucket size, see “DBS Control Utility” in Utilities , B035-1102 .