 |
 |
Archives of the TeradataForum
Message Posted: Wed, 24 Jun 2009 @ 15:57:31 GMT
| Subj: | | Re: Indexing Intersection (association) tables |
| |
| From: | | Victor Sokovin |
| | I am looking for some feedback on indexing for intersection tables. Something like a, Student_Enrollment table, that is an intersection
between a Student (PI - StudentID) table and a Course (PI - ClassID) table. Would I be better off with a combined Unique index of (StudentID,
ClassID) OR a Non Unique PI of Student ID OR Course ID alone (to facilitate collocation/same amp join with either of these two tables) for the
Student_Enrollment table. What are the factors to consider? | |
I understand you mean what we call "join" of tables, not the "intersection", because intersecting students with courses should just produce an
empty set, no?
If you really need the help of (N)PI for such a join then the StudentID NPI on the Enrollment table might be an idea as the Student table is
potentially larger than the Course table (unless we study the education system during the financial crisis) so the benefit might be larger than
otherwise. But the best thing is to experiment with the data. Perhaps you don't want to "waste" the NPI just on this particular join, which might
be easy anyway, and "save" it for a rainy day with a harder join.
Victor
| |
