 |
 |
Archives of the TeradataForum
Message Posted: Fri, 12 Dec 2008 @ 11:18:43 GMT
| Subj: | | Re: Rank and Delete? |
| |
| From: | | Victor Sokovin |
| | Sorry to disagree with you, Victor, but MultiLoad Delete is faster than an ordinary SQL delete. It apparently uses a different algorithm to
handle the blocks. I once deleted a billion rows in about 15 minutes (intentionally :-) ). | |
Geoffrey, I left a lot of exits open in my previous posting so it is not really possible to disagree with it. Of course, you are right that
when the query triggers the bulk algorithm or at least some parts of it then ML DELETE would be the fastest option (but then again, as Michael
said, you have the initialization phase of ML, which is only worth waiting on if you delete a significant number of rows) but in the context of
the original posting in this thread it is probably still a long way to find the right query for the job and while they are in this experimental
phase ML may not be the best instrument. Failures will probably have to be handled on DBA level whereas with BTEQ they might be still OK on
application level, and that's also part of the overall performance: how fast you can handle unforeseen situations. Average, real-life
performance, so to speak.
Victor
| |
