What you get: efficient sorted iteration through the keys, possibly limited to 
a given range.  For example, if you store a list of people keyed by their 
birthdate, and then want to find everyone born in a certain year, in order of 
birth, you can do that very fast.

Also in the BTree package is a TSTree, which has similar properties for String 
keys.  So as well as keeping them sorted, you can do efficient lookups of all 
the keys with a given prefix.  One other neat trick TSTree can do is a certain 
amount of fuzzy matching (eg find all keys with an edit distance of 3 from 
'foo') which makes it especially useful for spell checking and similar 
applications.

Avi