Vernon Wu wrote:
>
> The followings are the analyise outcomes after I did some modifications with the query. My finding is that the new query
> does improve the performance according to the plan. The actual time is reversed might due to the fact the test data is
> very small (the machine is a very old one by the way). The userid is the key for all tables and the gender is indexed. Do I
> also index the country and province to improve the preformance?
You start by using a dataset of realistic size. Sorry, but if the actual
time is < 10.00 ms it is rather pointless to optimize further since
chance is going to be the biggest factor. And the IN/EXISTS difference
is dependent on dataset size.
Jochem