Tunning the sql

Question

Below is the table structure & index's for the table and Total record count for the table is: 40450708
here, I am selecting distinct of ELIM_NBR from table using below query and it contains only: 1641 records
SELECT DISTINCT ELIM_NBR FROM ELIM_RULE_REF_TBL;while we run the above sql, its taking too long (more than 5 min), could you please help me - how to tune the query to get the distinct of ELIM_NBR from the table...
-- Create table
create table ELIM_RULE_REF_TBL
  ENT             VARCHAR2(4) not null,
  SUB_ENT         VARCHAR2(3) not null,
  SRC_ACCT        VARCHAR2(9) not null,
  EFF_FROM_DATE   DATE not null,
  EFF_TO_DATE     DATE not null,
  TARG_ENT        VARCHAR2(4) not null,
  TARG_SUB_ENT    VARCHAR2(3) not null,
  ELIM_DESCR      VARCHAR2(115),
  ELIM_NBR        VARCHAR2(5),
  ELIM_RESID_ACCT VARCHAR2(9),
  ENT_LVL         NUMBER
tablespace INS_DATA
  pctfree 10
  initrans 1
  maxtrans 255
  storage
    initial 128M
    next 50M
    minextents 1
    maxextents unlimited
    pctincrease 0
-- Create/Recreate indexes
create index ELIMREF_IDX_01 on ELIM_RULE_REF_TBL (TARG_ENT,TARG_SUB_ENT,SRC_ACCT,ELIM_RESID_ACCT,EFF_FROM_DATE,EFF_TO_DATE)
  tablespace INS_IDX
  pctfree 10
  initrans 2
  maxtrans 255
  storage
    initial 50M
    next 50M
    minextents 1
    maxextents unlimited
    pctincrease 0
create index ELIMREF_IDX_02 on ELIM_RULE_REF_TBL (TARG_ENT,TARG_SUB_ENT,ELIM_NBR,ENT,SUB_ENT)
  tablespace INS_IDX
  pctfree 10
  initrans 2
  maxtrans 255
  storage
    initial 50M
    next 50M
    minextents 1
    maxextents unlimited
    pctincrease 0
create unique index ELIMREF_IDX_PK on ELIM_RULE_REF_TBL (ENT,SUB_ENT,SRC_ACCT,EFF_FROM_DATE,EFF_TO_DATE,TARG_ENT,TARG_SUB_ENT)
  tablespace INS_IDX
  pctfree 10
  initrans 2
  maxtrans 255
  storage
    initial 50M
    next 50M
    minextents 1
    maxextents unlimited
    pctincrease 0
  );

Answer 1

If it is so crucial for you to get this particular query processed as fast as possible there are a couple of things you can evaluate:
* Follow this thread When your query takes too long ... to generate a proper explain plan output and run tkprof on a sql trace of this statement. Post the results here. This provides information where potentially most of time is spent to process the query.
* Create an (additional) index that consists just of the ELIM_NBR, provided that you only query this column and restrict only on that column in the WHERE clause. This builds a smaller index than the currently used index ELIMREF_IDX_02 which should speed up the full index scan.
* Since a sort unique operation is required you could think of increasing your sort memory usage. You haven't mentioned which database version you're running, but if you're already on 9i and have PGA_AGGREGATE_TARGET defined you could switch your session to "WORKAREA_SIZE_POLICY = manual" and setting SORT_AREA_SIZE to a large value, like 100000000 to allow a maximum of 100M for in-memory sorting, depending on your hardware, environment and concurrent usage of the system.
* If you have the suitable license and hardware you could try to increase the speed of the index fast full scan by using the parallel query option:
SELECT /*+ PARALLEL_INDEX(ELIM_RULE_REF_TBL, ELIMREF_IDX_02) */ DISTINCT ELIM_NBR
  FROM ELIM_RULE_REF_TBL
WHERE SUBSTR(ELIM_NBR, 2, 1) != '9';Regards,
Randolf
Oracle related stuff blog:
http://oracle-randolf.blogspot.com/
SQLTools++ for Oracle:
http://www.sqltools-plusplus.org:7676/
http://sourceforge.net/projects/sqlt-pp/