Consider the following example. Empirical testing indicates that one of the simplest relationships is that between the time required to create an in-place aggregate join index and the number of rows in the base tables. Even this simple relationship is complicated by the fact that the resulting creation time is likely to be more closely related to the number of rows that qualify for the join rather than the total number of rows.
The following table indicates a measured relationship between the elapsed creation time for a join index and the number of qualified rows.
| Elapsed Time (seconds) | Number of Qualified Rows x 106 |
|---|---|
| 133 | 90 |
| 415 | 00 |
| 2,344 | 900 |
These numbers, with the total number of rows in each test and a least squares linear regression of number of qualifying rows on elapsed time, are graphed in the following figure: