How to drop/add few colums in cheap and faster way from few huge partition table

Hi All,

  Our application support need to change few column from number (5,2) to NUMBER(10,7)  in few of the large partition table (there have more  1,000,000,000 rows on each tables).

  Just wonder any good recommendation to implement this change in efficient way. THis is very critical application and request downtime is very minimial.

  We're testing current approach in testing environment:

  #alter table tab1 set unused (col1,col2,col3)

  #alter table tab1 drop unused colums checkpoint 1000;

  #alter table tab1 add (col1 number(10,7), col2 number(10,7), col3(10,7)

  But still very slow, it took more than 10 hours and still running in step2, i can see the process still running (checking used undo segment, i can see the process is still running)