A DBMS implementation Question

Some small rational dbms use B tree for indexing. I found a problem in the condition that the key for the table is a multiple key. For example , there is a table which is denoted as emp(no, age, name), and (no + age) is the key for indexing.In the B Tree, the values of "no" and "age" are transfered to String and merge together, such as the key value of (1, 20, "abc") is "00010020",and the the B Tree of the table is established by the sequence of the key.

     Now , I have a problem , if I want to query all the people whose
     age are more than 20 , which is described as "select * from emp
     where age > 20", i can't use b tree for indexing, if i use , the
     number of records that i must scan may be very large, how to sovle
     the problem.
     The problem can be expressed as "how to deal with scaning the table
     using non-primary key or non-key field".
     thanks!!

--
Posted via http://dbforums.com