Archives of the TeradataForum
Message Posted: Fri, 10 Sep 2004 @ 10:51:14 GMT
| Subj: | | Re: Optimizing SQL |
| |
| From: | | Victor Sokovin |
| | Thanks for all the input on optimizing my sql. I changed the table to use a UPI and increased the speed from an hour to one second. Just
thought everyone would want to know the outcome. | |
I wonder which part of that second accounts for the journaling.
My feeling with this type of updates was always that the tables should be in one-to-one relationship (on row level) with each other for the
statement to perform well. The syntax and stats should be clear enough for the optimizer to understand that.
I must admit that it also helps *me* to understand the logic of what I am about to do as well, so I have the sympathy for the optimizer in this
case. UPI clearly helps the optimizer a lot to choose the right plan but it is not always the option to change NUPI to UPI, is it? Did you try to
keep the NUPI and "clarify" the predicates, especially that OR condition? It would be good to have a solution for the NUPI as well, and your data
set seems to be a good test case with such a range in performance. Let's hope you have the energy to play further with it and post the
results.
Regards,
Victor
| 
