Someone sent me a scenario the other day similar to this:
A small dimension table, for example of US State names, has a synthetic primary key (say STATE#) and a unique key on the real-world value (STATE_NAME), and a much larger fact-type table has an indexed foreign key to this lookup table. The distribution of values in the fact table is very skewed. Is it possible for the optimizer to take into account the skew of STATE# in the fact table when it is joined to the dimension table and a predicate is placed on STATE_NAME?
Well, a tricky issue. Read the rest of this entry »
