Re: what made the difference?

>I don't have an execution plan from prior index (col1,col2,col3) but I

>remember that it wasn't FTS for sure. Actually both execution paths
are
>the same, but for some reason later version is much faster.

If it happens to be an index range scan, with the useless column at the beginning it has to scan the whole index, but with col2 at the beginning it only has to scan a little bit of it.

Read what Dan and David wrote and follow their advice!

I'm also working with a packaged app that has such issues. It's amazing how often a company will say something like "We will _never_ have more than one warehouse." Then code a bunch of customizations based on that. Then get some more warehouses. So be sure and have a special doc that says everything about how the app is customized and any special indexing.

jg

--
@home.com is bogus.
http://www.atomicmpc.com.au/article.asp?CIID=21276