Major query performance differnce between oracle 8 and 9

Question

Hello, I have the following query
   select distinct UPPER(rf.module), ruf.rpt_seq
     from role_func rf, rd_url_func ruf
    where role_name = 'ADMIN'
      and UPPER(rf.module) not in (select UPPER(ruf.module)
                                     from role_func rf2, rd_url_func ruf2
                                    where UPPER(ruf2.module)        = UPPER(rf.module)
                                      and UPPER(rf2.module(+))      = UPPER(ruf2.module)
                                      and rf2.url(+)                = ruf2.url
                                      and rf2.role_name(+)          = 'ADMIN'
                                      and rf2.url        is null)
      and UPPER(rf.module) = UPPER(ruf.module)
      and ruf.rpt_seq = (SELECT min(rpt_seq)
                           FROM rd_url_func ruf3
                          WHERE ruf3.module = ruf.module)
    order by ruf.rpt_seq; Now on Oracle 8, this executes almost instantly. On oracle 9, however, this takes a very long time (around 30 seconds?) In both databases I have the same data. Also none of the tables are extremely large- they both contain only about 400 rows. Any suggestions on what could be causing this difference, or at least how I can find out the problem?
Thanks

Answer 1

ok does this help:
   explain plan for select distinct UPPER(rf.module), ruf.rpt_seq
     from role_func rf, rd_url_func ruf
    where role_name = 'ADMIN'
      and UPPER(rf.module) not in (select UPPER(ruf.module)
                                     from role_func rf2, rd_url_func ruf2
                                    where UPPER(ruf2.module)        = UPPER(rf.module)
                                      and UPPER(rf2.module(+))      = UPPER(ruf2.module)
                                      and rf2.url(+)                = ruf2.url
                                      and rf2.role_name(+)          = 'ADMIN'
                                      and rf2.url        is null)
      and UPPER(rf.module) = UPPER(ruf.module)
      and ruf.rpt_seq = (SELECT min(rpt_seq)
                           FROM rd_url_func ruf3
                          WHERE ruf3.module = ruf.module)
    order by ruf.rpt_seq;
select
  substr (lpad(' ', level-1) || operation || ' (' || options || ')',1,30 ) "Operation",
  object_name                                                              "Object"
from
  plan_table
start with id = 0
connect by prior id=parent_id;this is from oracle 8 where it executes fast:
Operation     Object
SELECT STATEMENT ()     
SORT (UNIQUE)     
  FILTER ()     
   HASH JOIN ()     
    TABLE ACCESS (FULL)     RD_URL_FUNC
    TABLE ACCESS (FULL)     ROLE_FUNC
   FILTER ()     
    FILTER ()     
     NESTED LOOPS (OUTER)     
      TABLE ACCESS (FULL)     RD_URL_FUNC
      TABLE ACCESS (BY INDEX R     ROLE_FUNC
       INDEX (UNIQUE SCAN)     RFUN_PK
   SORT (AGGREGATE)     
    TABLE ACCESS (FULL)     RD_URL_FUNCthis is from oracle 9 where it executes slow
Operation     Object
SELECT STATEMENT ()     
SORT (UNIQUE)     
  FILTER ()     
   SORT (GROUP BY)     
    FILTER ()     
     HASH JOIN ()     
      TABLE ACCESS (FULL)     ROLE_FUNC
      HASH JOIN ()     
       TABLE ACCESS (FULL)     RD_URL_FUNC
       TABLE ACCESS (FULL)     RD_URL_FUNC
     FILTER ()     
      FILTER ()     
       NESTED LOOPS (OUTER)     
        TABLE ACCESS (FULL)     RD_URL_FUNC
        TABLE ACCESS (BY INDEX     ROLE_FUNC
         INDEX (UNIQUE SCAN)     RFUN_PKcan someone help interpret the difference between the execution plan, and how to make oracle use the first one on oracle 9?