Physical Database Design

11.7 Forcing the Issue When the Query Optimizer Chooses Wrong 215

time consuming when large amounts of data are involved, but are relatively quick

within a data-free environment.

We mentioned earlier that query execution plans depend on the size and distribu-

tion of data in the database. Thus, one might wonder how a valid simulation of the

database can be set up that is data free, but still provides a fair simulation of the query

execution plans. The answer lies in the fact that the query optimizer uses statistics that

are stored in the system catalogs and these cataloged statistics can be copied over to the

data-free database. As a result, even though the database has no data in it, the query

compiler (and optimizer) sees a database full of data ...