>I agree that individual fields have relatively few possible values -
> hopefully, when those are combined in a multi-column index, he will
> have a greater number of unique combinations, gaining more out of the
> index. That's why I suggested putting stype and Is_id as the first
> two fields in the index (though I guess I did not mention that!).
>
> stype had 6 values, Is_id had 5, so he may have up to about 30
> combinations as the first two fields, which should be enough to help a
> lot.
ls_id is evenly distributed but stype is not. But thought is interesting.
And one question. I mentioned that update:select ratio is 3:1. There are
no deletes and inserts. Also update does not touch ls_id and stype fields.
Is there index updating overhead then?
Mindaugas