 |
 |
Archives of the TeradataForum
Message Posted: Thu, 17 Mar 2005 @ 13:14:47 GMT
| Subj: | | Re: TPump Update |
| |
| From: | | Victor Sokovin |
| | Therefore if I locked the row with UPI = 5697436052 with a row hash lock, it would also lock the row with a UPI = 1402468761. | |
| | I think this is supported by Bills observations that the larger the Pack factor and/or the larger the number of sessions, the more
blocking. | |
It is definitely good to revisit the subject. This phenomenon is called hash synonyms (not collisions, as somebody in the thread suggested).
Dennis had once provided a good crash course on them:
www.teradataforum.com/teradata/20020219_094825.htm
I did spend some time investigating how often synonyms occur in typical real-life data and my conclusion was that that they were negligible.
One should not forget them entirely, though. Who knows, perhaps with some data they might play a significant role. Also, on older releases the
hashing algorithm might have allowed more synonyms - speculation on my part, I did not verify it. Perhaps Bill could investigate how many synonyms
he has in his R3 table under his R3 hashing algorithm. It would be a good extra metric to describe his case. On the other hand, he has already
mentioned that the target table size (and, therefore, the possibly higher or lower number of potential synonyms) did not play an important
role.
Nice to see the thread has had so many interesting ramifications.
Regards,
Victor
| |
