欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识！
您现在的位置是： 首页  >  数据库

optimizer_index_caching和optimizer_index_cost_adj两个参数说

程序员文章站 2024-01-03 18:41:58
...

一、optimizer_index_cost_adj参数 优化器计算通过索引扫描访问表数据的cost开销，可以通过这个参数进行调整。参数可用值的范围为1到10000。默认值为100， 超过100后越大则越会使索引扫描的COST开销越高 （计算的），从而导致查询优化器更加倾向于使用全表扫

一、optimizer_index_cost_adj参数 优化器计算通过索引扫描访问表数据的cost开销，可以通过这个参数进行调整。参数可用值的范围为1到10000。默认值为100，超过100后越大则越会使索引扫描的COST开销越高（计算的），从而导致查询优化器更加倾向于使用全表扫描。相反，值越小于100，计算出来的索引扫描的开销就越低。 注意： 1、这里描述的开销，仅仅为优化器评估出来的而已，而非实际执行的开销；例如同样的SQL语句，同样的执行路径，修改这个参数以后，计算出来的cost不同，但是，SQL语句执行的实际物理路径、时间、逻辑读都是一样的。 2、这个参数影响优化器评估索引访问的IO开销 3、可以参见后面的示例进一步理解这个参数
二、optimizer_index_caching参数 用于在执行in-list遍历和嵌套循环连接时，优化器评估已经存在于buffer cache中的索引块的数量（以百分比的方式）。参数的取值范围是0到100，默认值为0,取值越大就越减少优化器在评估In-list和嵌套循环连接的索引扫描的开销COST。 1、这里的描述仅仅为优化器的评估结果。换句话说，它不是用来指定数据库实际缓存的每个索引块的数量 2、可以参见后面的示例进一步理解这个参数
三、索引范围扫描的成本计算公式（与两个参数相结合） cost = { (blevel+leaf_blocks * effective index selectivity)*(1-optimizer_index_caching/100)+ cluster_factor * effective table selectivity)*(optimizer_index_cost_adj/100) } 说明 1、blevel=索引的层数，即dba_indexes视图中的blevel列值 2、leaf_blocks为索引的叶子块数量,即dba_indexes视图中的leaf_blocks列值 3、effective index selectivity指的是SQL语句中用于索引扫描的谓词驱动条件的列字段的选择率（where条件中） 4、cluster_factor为索引的聚簇因子（表示索引数据与表数据排序后的相近程度） 5、effective table selectivity指的是SQL where条件中能够在索引上进行过滤的所有谓词列字段的选择率（可以理解为通过索引扫描后（包含驱动和过滤），最后通过rowid定位表的数量） 6、一般是4、5两个部分往往是整个计算公式中权重最大的因素。
注意 1、从上面的成本计算公式和说明，我们知道，optimizer_index_cost_adj参数对优化器的影响要远远大于optimizer_index_caching参数 2、随着优化器越来越智能这两个参数的作用已经慢慢被 系统统计信息 给替换了。 3、10gR2开始，这两个参数的值尽量保持为默认值。如果实在需要调整，请严格测试！

四、optimizer_index_cost_adj参数示例 SQL> create table t as select rownum rn ,object_name name from dba_objects a where rownum Table created.
SQL> create index t_idx on t(rn);
Index created.
SQL> BEGIN
2
3 DBMS_STATS.GATHER_TABLE_STATS(OWNNAME=>'SYS',
4 TABNAME=>'T',
5 ESTIMATE_PERCENT=>30,
6 METHOD_OPT=>'FOR ALL COLUMNS SIZE 1',
7 NO_INVALIDATE=>FALSE,
8 CASCADE=>TRUE,
9 DEGREE => 4);
10 END;
11 /
PL/SQL procedure successfully completed.
SQL> explain plan for select * from t where rn Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 470836197

-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 199 | 4179 | 3 (0)| 00:00:35 |
| 1 | TABLE ACCESS BY INDEX ROWID| T | 199 | 4179 | 3 (0)| 00:00:35 |
|* 2 | INDEX RANGE SCAN | T_IDX | 199 | | 2 (0)| 00:00:24 |
-------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("RN"
14 rows selected.

SQL> alter session set optimizer_index_cost_adj=50;
Session altered.
SQL> explain plan for select * from t where rn Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 470836197
-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 199 | 4179 | 2 (0)| 00:00:18 |
| 1 | TABLE ACCESS BY INDEX ROWID| T | 199 | 4179 | 2 (0)| 00:00:18 |
|* 2 | INDEX RANGE SCAN | T_IDX | 199 | | 1 (0)| 00:00:12 |
-------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("RN" 14 rows selected.

SQL> alter session set optimizer_index_cost_adj=500;
Session altered.
SQL> explain plan for select * from t where rn Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 1601196873
-------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 199 | 4179 | 6 (0)| 00:01:10 |
|* 1 | TABLE ACCESS FULL| T | 199 | 4179 | 6 (0)| 00:01:10 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("RN" 13 rows selected.

五、optimizer_index_caching参数示例 IN-LIST示例 SQL> set linesize 200 pagesize 9999
SQL> explain plan for select * from t where rn in (1,2,3,4,5,6,7,100,130,200,240,2000);
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 3506647781
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 12 | 252 | 4 (0)| 00:00:43 |
| 1 | INLIST ITERATOR | | | | | |
| 2 | TABLE ACCESS BY INDEX ROWID| T | 12 | 252 | 4 (0)| 00:00:43 |
|* 3 | INDEX RANGE SCAN | T_IDX | 12 | | 3 (0)| 00:00:31 |
--------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - access("RN"=1 OR "RN"=2 OR "RN"=3 OR "RN"=4 OR "RN"=5 OR "RN"=6 OR
"RN"=7 OR "RN"=100 OR "RN"=130 OR "RN"=200 OR "RN"=240 OR "RN"=2000)
16 rows selected.
SQL> alter session set optimizer_index_caching=10;
Session altered.
SQL> explain plan for select * from t where rn in (1,2,3,4,5,6,7,100,130,200,240,2000);

Explained.

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

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 3506647781

--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 12 | 252 | 3 (0)| 00:00:31 |
| 1 | INLIST ITERATOR | | | | | |
| 2 | TABLE ACCESS BY INDEX ROWID| T | 12 | 252 | 3 (0)| 00:00:31 |
|* 3 | INDEX RANGE SCAN | T_IDX | 12 | | 2 (0)| 00:00:19 | --可以看到cost确实降低了
--------------------------------------------------------------------------------------

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

3 - access("RN"=1 OR "RN"=2 OR "RN"=3 OR "RN"=4 OR "RN"=5 OR "RN"=6 OR
"RN"=7 OR "RN"=100 OR "RN"=130 OR "RN"=200 OR "RN"=240 OR "RN"=2000)

16 rows selected 嵌套循环连接示例： SQL> explain plan for select /*+ use_nl(a b) */ * from t a,t b where a.rn=b.rn and b.name='sss';
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 752965310
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 42 | 8 (0)| 00:01:34 |
| 1 | NESTED LOOPS | | | | | |
| 2 | NESTED LOOPS | | 1 | 42 | 8 (0)| 00:01:34 |
|* 3 | TABLE ACCESS FULL | T | 1 | 21 | 6 (0)| 00:01:10 |
|* 4 | INDEX RANGE SCAN | T_IDX | 1 | | 1 (0)| 00:00:12 |
| 5 | TABLE ACCESS BY INDEX ROWID| T | 1 | 21 | 2 (0)| 00:00:24 |
--------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - filter("B"."NAME"='sss')
4 - access("A"."RN"="B"."RN")
18 rows selected.
SQL> alter session set optimizer_index_caching=10;
Session altered.
SQL> explain plan for select /*+ use_nl(a b) */ * from t a,t b where a.rn=b.rn and b.name='sss';
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 752965310
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 42 | 7 (0)| 00:01:22 |
| 1 | NESTED LOOPS | | | | | |
| 2 | NESTED LOOPS | | 1 | 42 | 7 (0)| 00:01:22 | --可以看到cost确实降低了
|* 3 | TABLE ACCESS FULL | T | 1 | 21 | 6 (0)| 00:01:10 |
|* 4 | INDEX RANGE SCAN | T_IDX | 1 | | 0 (0)| 00:00:01 |
| 5 | TABLE ACCESS BY INDEX ROWID| T | 1 | 21 | 1 (0)| 00:00:12 |
--------------------------------------------------------------------------------------

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

3 - filter("B"."NAME"='sss')
4 - access("A"."RN"="B"."RN")

18 rows selected.

非in-list和嵌套循环操作，调整这个参数时，不会影响oracle优化器成本的运算，如下
SQL> set linesize 200 pagesize 999
SQL> explain plan for select * from t where rn Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 470836197
-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 199 | 4179 | 3 (0)| 00:00:35 |
| 1 | TABLE ACCESS BY INDEX ROWID| T | 199 | 4179 | 3 (0)| 00:00:35 |
|* 2 | INDEX RANGE SCAN | T_IDX | 199 | | 2 (0)| 00:00:24 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("RN" 14 rows selected.
SQL> alter session set optimizer_index_caching=1;
Session altered.
SQL> explain plan for select * from t where rn Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 470836197
-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 199 | 4179 | 3 (0)| 00:00:35 |
| 1 | TABLE ACCESS BY INDEX ROWID| T | 199 | 4179 | 3 (0)| 00:00:35 |
|* 2 | INDEX RANGE SCAN | T_IDX | 199 | | 2 (0)| 00:00:24 | --可以看到cost确实没有改变
-------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("RN" 14 rows selected.
相关标签： optimizer index caching cost a

上一篇：

下一篇：

推荐阅读