Definitions for the Six Database Hash Maps - Teradata Database

Definitions for the Six Database Hash Maps

The 6 hash maps maintained on each node in a Teradata system are defined as follows.

The current configuration primary map

This map maintains the hash bucket-AMP correspondences for the primary copies of all rows on the system.

The following graphic is a logical diagram of the primary hash map for a system with 65,536 hash buckets. Physical hash maps do not look like this.

The current configuration fallback map

This map maintains the hash bucket-to-AMP correspondences for the fallback copies of all rows from tables defined with fallback. The fallback hash map is not identical to the current configuration primary hash map.

The reconfiguration primary map

This map indicates which primary rows migrate to which AMPs during a reconfiguration.

The reconfiguration fallback map

This map indicates which fallback row copies migrate to which AMPs during a reconfiguration.

The NoPI hash map

The bit map hash map

This map is used to optimize hash‑related queries on the AMPs (see “NUSI Bit Mapping” on page 428).

Another hash map, called the Open PDE hash map, is maintained by the Parallel Database Extensions system to indicate the destination for Parser messages that are addressed by hash bucket number.

The increased size of the hash maps for systems with 1,048,576 hash buckets increases the amount of memory consumed by them, as indicated by the following table.

Hash Map	Size for 16-Bit Bucket	Size for 20-Bit Bucket
DBS primary	128 KB	2 MB
DBS fallback	128 KB	2 MB
DBS reconfiguration primary	128 KB	2 MB
DBS reconfiguration fallback	128 KB	2 MB
DBS NoPI	128 KB	2 MB
DBS bitmap	64 KB	1 MB
Open PDE	4 MB	4 MB

These changes result in a total increase of memory consumption of approximately 8.5 MB for a system configured with 20‑bit hash buckets. If the system is configured with 16-bit hash buckets, increased memory consumption is restricted to the 4 MB used for the Open PDE hash map.