利用集群因子优化

create index END_DT_IDX1 on F_AGT_BUSINESS_CONTRACT_H(end_dt);

SQL> explain plan for  select * from  F_AGT_BUSINESS_CONTRACT_H t where t.end_dt = date '2999-12-31';

Explained.

SQL> select * from table(dbms_xplan.display());

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 3544262987

-----------------------------------------------------------------------------------------------
| Id  | Operation	  | Name		      | Rows  | Bytes | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |			      |   238K|   140M| 45269	(1)| 00:09:04 |
|*  1 |  TABLE ACCESS FULL| F_AGT_BUSINESS_CONTRACT_H |   238K|   140M| 45269	(1)| 00:09:04 |
-----------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("T"."END_DT"=TO_DATE(' 2999-12-31 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))

13 rows selected.

为什么没走索引呢？

1.查看数据分布

SQL> select end_dt,count(*) from F_AGT_BUSINESS_CONTRACT_H 
group by end_dt
having end_dt = date '2999-12-31';  2    3  

END_DT	     COUNT(*)
---------- ----------
2999-12-31     246369

总条数:
SQL> select count(*) from F_AGT_BUSINESS_CONTRACT_H;

  COUNT(*)
----------
   1614953


2.查看集群因子:
SQL> select index_name,clustering_factor from user_indexes where table_name='F_AGT_BUSINESS_CONTRACT_H'
  2  and index_name='END_DT_IDX1';

INDEX_NAME		       CLUSTERING_FACTOR
------------------------------ -----------------
END_DT_IDX1				  557965


3.查看表的块数:

SQL> select count(distinct dbms_rowid.rowid_block_number(rowid)) from F_AGT_BUSINESS_CONTRACT_H a
  2  where a.end_dt = date '2999-12-31';

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
					      33459

SQL> select count(distinct dbms_rowid.rowid_block_number(rowid)) from F_AGT_BUSINESS_CONTRACT_H;

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
					      96435
如果clustering factor 接近block 数，说明表的存储和索引存储排序接近，也就是说表中的记录很有序，这样在做index range scan 的时候能，读取少量的data block 就能得到我们想要的数据，代价比较小。如果clustering factor 接近表记录数，说明表的存储和索引排序差异很大，在做index range scan 的时候，会额外读取多个block，因为表记录分散，代价较高。

4. 重建表：
CREATE TABLE F_AGT_BUSINESS_CONTRACT_H_1 AS SELECT * FROM F_AGT_BUSINESS_CONTRACT_H ORDER BY end_dt;

SQL> create index END_DT_IDX2 on F_AGT_BUSINESS_CONTRACT_H_1(end_dt);

BEGIN
  DBMS_STATS.GATHER_TABLE_STATS(ownname          => 'DWF',
                                tabname          => 'F_AGT_BUSINESS_CONTRACT_H_1',
                                estimate_percent => 30,
                                method_opt       => 'for all columns size repeat',
                                no_invalidate    => FALSE,
                                degree           => 8,
                                cascade          => TRUE);
END;

查看此时的集群因子:
SQL> select index_name,clustering_factor from user_indexes where table_name='F_AGT_BUSINESS_CONTRACT_H_1';

INDEX_NAME		       CLUSTERING_FACTOR
------------------------------ -----------------
END_DT_IDX2				  171023

SQL> select count(distinct dbms_rowid.rowid_block_number(rowid)) from F_AGT_BUSINESS_CONTRACT_H_1;

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
					     161791

此时的集群因子和块数接近；

查看此时需要访问的块数:
SQL> select count(distinct dbms_rowid.rowid_block_number(rowid)) from F_AGT_BUSINESS_CONTRACT_H_1;

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
					     161791

SQL> select count(distinct dbms_rowid.rowid_block_number(rowid)) from F_AGT_BUSINESS_CONTRACT_H_1 a
where a.end_dt>date'2014-03-01'  2  ;

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
					      24971

此时集群因子接近了表的块数

SQL> explain plan for select * from  F_AGT_BUSINESS_CONTRACT_H_1 t where t.end_dt = date '2999-12-31';

Explained.

SQL> select * from table(dbms_xplan.display());

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 1303973883

-----------------------------------------------------------------------------------------------------------
| Id  | Operation		    | Name			  | Rows  | Bytes | Cost (%CPU)| Time	  |
-----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT	    |				  |  1478 |   894K|   163   (0)| 00:00:02 |
|   1 |  TABLE ACCESS BY INDEX ROWID| F_AGT_BUSINESS_CONTRACT_H_1 |  1478 |   894K|   163   (0)| 00:00:02 |
|*  2 |   INDEX RANGE SCAN	    | END_DT_IDX2		  |  1502 |	  |	6   (0)| 00:00:01 |
-----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("T"."END_DT"=TO_DATE(' 2999-12-31 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))

14 rows selected.