Archives of the TeradataForum
Message Posted: Tue, 01 Aug 2006 @ 18:53:14 GMT
| Subj: | | Re: Parallel execution of Queries |
| |
| From: | | Victor Sokovin |
| | Consider the following two scenarios: | |
| | 1) There is a single session of a user. The user executes two queries in a parallel fashion. | |
| | 2) There are two sessions of the same user. Two queries are executed separately in the two sessions. These queries are
submitted with, say, a very negligible time difference. | |
I read this as 1) is one transaction and 2) - two transactions.
| | Which one of the two scenarios would return the results faster? | |
In general, it probably depends on how "interwoven" the statements are. In scenario 1 they form one transaction and might need to wait on each
other before TD can decide whether to commit or rollback the whole transaction.
| | Also, which one will consume fewer resources? | |
1) does not need to prepare for two rollbacks so it is likely to consume less but who knows what happens if the execution plans change because
the statements are in one transaction.
| | Will there be any difference in the spool space consumed? | |
In general, when the plans are more or less the same then probably not. If the plans change, hard to predict what happens.
Sorry for the vague answers.
There is one more obvious case: when you insert data into an empty table. 1) should be the winner then because the rollback is so easy.
Regards,
Victor
| 
