optimizing the sequence of index fields

We're having poor performance when tracing is turned on for an index whose first field has thousands of identical values. It seems as if a multiple-field index will be more efficient if the values of the first field are highly varied - but I never can tell with Informix! Is there a reliable way to sequence the fields in an index for efficiency? -:- "You have SECONDS to live, Cynicalman! Use them wisely!" "OK, I'll get my socks out of the dryer." -- Col. G. L. Sicherman sicherman@lucent.com