Archives of the TeradataForum
Message Posted: Thu, 01 Jun 2006 @ 18:35:42 GMT
| Subj: | | Hash Joins - hash table bucket overflow |
| |
| From: | | Robert D Meunier |
The following is a paragraph regarding Hash Joins out of the manual SQL Reference: Statement and Transaction Processing:
If the specified skew allowance is insufficient to correct for data skew, and hash table bucket overflow occurs, then the system matches rows
from the corresponding probe table against build table rows that are already loaded into the memory-resident hash table. After all the probe
partition rows have been processed, the system clears the hash table and moves more rows from the oversized build table partition into memory. The
system then re-reads rows from the probe partition and matches them against the freshly loaded build table rows. This procedure repeats until the
entire build partition has been processed.
Does anyone in the group know how you can tell when a "hash table bucket overflow" has occurred?
Robert Meunier
| 
