 |
 |
Archives of the TeradataForum
Message Posted: Fri, 28 Aug 2009 @ 10:22:10 GMT
| Subj: | | Re: Parsing of SQL in teradata |
| |
| From: | | Victor Sokovin |
| | My question was exactly that when I write a sql (before I actually do the explain),do I need to worry about the order of where and join
clauses ? Or anyhow I put it the teradata v2r6 optimizer is 'smart' enough to give me a low cost way of executing the same? | |
You may worry about the *number* of joins and clauses rather than their order. Try to avoid unnecessary or fancy ones because they increase the
number of branches on the decision tree the optimizer has to go through. You potentially will also need to collect more statistics for all those
nice-to-haves to really work. Whether the optimizer will find the best path is a question but you always lose the time on that extra work the
optimizer has to perform, and regular statistics collection is not an easy job either. Also, when queries get really complicated execution plans
tend to become less stable. They may change depending on demographics. Sometimes you may prefer stability and less surprises in this department
even at the price of a somewhat suboptimal performance.
Victor
| |
