oracle sql 执行计划分析

转自http://itindex.net/detail/45962-oracle-sql-%E8%AE%A1%E5%88%92

一、首先创建表

SQL> show user
USER is "RHYS"
SQL> create table A(col1 number(4,0),col2 number(4,0), col4 char(30));
create table B(col1 number(4,0),col3 number(4,0), name_b char(30));
create table C(col2 number(4,0),col3 number(4,0), name_c char(30));
Table created.
SQL> 
Table created.
SQL> 
Table created.

第二、查看一下执行计划。
1、

SQL> select a.col4 from c,a,b
  2  where c.col3=5 and a.col1=b.col1 and a.col2=c.col2 and b.col3=10;
Execution Plan
----------------------------------------------------------
Plan hash value: 1485247927
------------------------------------------------------------------------------
| Id  | Operation             | Name | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |      |     1 |   110 |     6   (0)| 00:00:01 |
|*  1 |  HASH JOIN            |      |     1 |   110 |     6   (0)| 00:00:01 |
|   2 |   MERGE JOIN CARTESIAN|      |     1 |    52 |     4   (0)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL  | C    |     1 |    26 |     2   (0)| 00:00:01 |
|   4 |    BUFFER SORT        |      |     1 |    26 |     2   (0)| 00:00:01 |
|*  5 |     TABLE ACCESS FULL | B    |     1 |    26 |     2   (0)| 00:00:01 |
|   6 |   TABLE ACCESS FULL   | A    |     1 |    58 |     2   (0)| 00:00:01 |
------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("A"."COL1"="B"."COL1" AND "A"."COL2"="C"."COL2")
   3 - filter("C"."COL3"=5)
   5 - filter("B"."COL3"=10)
Note
-----
   - dynamic sampling used for this statement (level=2)

执行计划主要查看：访问路径，连接顺序，连接方法
执行计划顺序为上内原则，同层次上边先执行，内层先执行。
plan hash value：当sql第一次在shared pool中进行执行的是硬解析并生产该hash值
id，只是一个标号，并不是实际执行顺序
operation：从字面意思也看出来就是操作的类型
name：对象的名字
rows：oracle估计该操作返回的行数
bytes：产生的数据量
cost：表示该sql执行到此步骤的时候sql执行代价。
该sql的执行步骤如下：
首先执行id 3-》id5-》id4—》id2-》id6-》id1-》id0
首先对id3进行全表扫描过滤条件为filter("C"."COL3"=5)，然后对表b进行全表扫描，条件为 filter("B"."COL3"=10),完了之后再进行buffer sort排序，最后把3和4的row source 进行merge join 笛卡尔积操作，并把所有的结果作为row source1 ，也就是驱动表，然后把表A作为被探测表，两者进行hash join。这就是这一个过程信息。
注意此处在id5和id3没有关联的条件，就采用了笛卡尔积，这是不好的现象。
2、

SQL> select /*+ordered*/  a.col4 from c,a,b
  2  where c.col3=5 and a.col1=b.col1 and a.col2=c.col2 and b.col3=10;
Execution Plan
----------------------------------------------------------
Plan hash value: 531790806
----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |     1 |   110 |     6   (0)| 00:00:01 |
|*  1 |  HASH JOIN          |      |     1 |   110 |     6   (0)| 00:00:01 |
|*  2 |   HASH JOIN         |      |     1 |    84 |     4   (0)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| C    |     1 |    26 |     2   (0)| 00:00:01 |
|   4 |    TABLE ACCESS FULL| A    |     1 |    58 |     2   (0)| 00:00:01 |
|*  5 |   TABLE ACCESS FULL | B    |     1 |    26 |     2   (0)| 00:00:01 |
----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("A"."COL1"="B"."COL1")
   2 - access("A"."COL2"="C"."COL2")
   3 - filter("C"."COL3"=5)
   5 - filter("B"."COL3"=10)
Note
-----
   - dynamic sampling used for this statement (level=2)

使用hints可以调整optimizer的执行连接方法，在此例中我们指定了ordered使得采用hash join选取from 之后从左到有第一个表c作为驱动表。
执行顺序为：id3全表扫描过滤条件为filter("C"."COL3"=5)-》id4 全表扫描，然后表c为驱动表，a为探测表以此来进行hashjoin-》id5 全表扫描过滤条件为filter("B"."COL3"=10)，此后执行id2为外部表，id5为被探测表进行hash join，从access访问路径可以看出首先是id2为("A"."COL2"="C"."COL2")此后为 id1access("A"."COL1"="B"."COL1")。
这是整个sql执行的整个过程。
为了便于理解分析一下数据，
首先我要取到在表c中col3=5的所有数据，然后再内存进行hash，作为hash table，然后我在去使用该hash table去探测A表进行匹配，取出的数据为access("A"."COL2"="C"."COL2")，把最后的匹配结果作为row source，再次建立hash table表，然后再去探测b表，方式为：access("A"."COL1"="B"."COL1")。最终获得了0执行的结果信息。
对于note动态采样信息请参考：
http://www.oracle.com/technetwork/issue-archive/2009/09-jan/o19asktom-086775.html
由于本次没有对表进行analyze所有存有动态取样。

SQL> select /*+ordered use_nl(a c)*/  a.col4 from c,a,b
  2  where c.col3=5 and a.col1=b.col1 and a.col2=c.col2 and b.col3=10;
Execution Plan
----------------------------------------------------------
Plan hash value: 1446226736
----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |     1 |   110 |     6   (0)| 00:00:01 |
|*  1 |  HASH JOIN          |      |     1 |   110 |     6   (0)| 00:00:01 |
|   2 |   NESTED LOOPS      |      |     1 |    84 |     4   (0)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| C    |     1 |    26 |     2   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| A    |     1 |    58 |     2   (0)| 00:00:01 |
|*  5 |   TABLE ACCESS FULL | B    |     1 |    26 |     2   (0)| 00:00:01 |
----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("A"."COL1"="B"."COL1")
   3 - filter("C"."COL3"=5)
   4 - filter("A"."COL2"="C"."COL2")
   5 - filter("B"."COL3"=10)
Note
-----
   - dynamic sampling used for this statement (level=2)

在这个语句中，表c和a进行了nested loops然后把结果惊醒hash table在与表b做jash join。

另外对于表有索引的情况进行如下分析。
首先创建表a的组合索引，索引列为（col1，col2）
eg：

SQL> create index inx_col12A on a(col1,col2);
Index created.
SQL> select A.col4
  2  from C , A , B
  3  where C.col3 = 5 and A.col1 = B.col1 and A.col2 = C.col2
  4  and B.col3 = 10;
Execution Plan
----------------------------------------------------------
Plan hash value: 2122808611
-------------------------------------------------------------------------------------------
| Id  | Operation                    | Name       | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT             |            |     1 |   110 |     4   (0)| 00:00:01 |
|   1 |  NESTED LOOPS                |            |     1 |   110 |     4   (0)| 00:00:01 |
|   2 |   NESTED LOOPS               |            |     1 |   110 |     4   (0)| 00:00:01 |
|   3 |    MERGE JOIN CARTESIAN      |            |     1 |    52 |     4   (0)| 00:00:01 |
|*  4 |     TABLE ACCESS FULL        | C          |     1 |    26 |     2   (0)| 00:00:01 |
|   5 |     BUFFER SORT              |            |     1 |    26 |     2   (0)| 00:00:01 |
|*  6 |      TABLE ACCESS FULL       | B          |     1 |    26 |     2   (0)| 00:00:01 |
|*  7 |    INDEX RANGE SCAN          | INX_COL12A |     1 |       |     0   (0)| 00:00:01 |
|   8 |   TABLE ACCESS BY INDEX ROWID| A          |     1 |    58 |     0   (0)| 00:00:01 |
-------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   4 - filter("C"."COL3"=5)
   6 - filter("B"."COL3"=10)
   7 - access("A"."COL1"="B"."COL1" AND "A"."COL2"="C"."COL2")
Note
-----
   - dynamic sampling used for this statement (level=2)

这个比较有意思了。首先看一下执行顺序，首先对表c进行全表扫描过滤条件为col3=5取出数据作为row source1，然后再对b进行全表扫描过滤条件为col3=10，因为走的是merge join 笛卡尔积的排序连接，然后再buffer 进行sort作为row sources2 ，完了之后row source1和row source2作合并连接，完了之后作为row source1 是驱动表，然后再进行index range scan（索引范围扫描）访问路径为： access("A"."COL1"="B"."COL1" AND "A"."COL2"="C"."COL2")，完了之后把结果作为row source1 然后再去与表A进行嵌套循环操作，不过A也就是id8 走的是index rowid。完了之后再进行0获得数据。太繁琐了。呵呵。

SQL> select /*+ ORDERED USE_NL (A C)*/ A.col4
  2  from C , A , B
  3  where C.col3 = 5 and A.col1 = B.col1 and A.col2 = C.col2
  4  and B.col3 = 10;
Execution Plan
----------------------------------------------------------
Plan hash value: 1446226736
----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |     1 |   110 |     6   (0)| 00:00:01 |
|*  1 |  HASH JOIN          |      |     1 |   110 |     6   (0)| 00:00:01 |
|   2 |   NESTED LOOPS      |      |     1 |    84 |     4   (0)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| C    |     1 |    26 |     2   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| A    |     1 |    58 |     2   (0)| 00:00:01 |
|*  5 |   TABLE ACCESS FULL | B    |     1 |    26 |     2   (0)| 00:00:01 |
----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - access("A"."COL1"="B"."COL1")
   3 - filter("C"."COL3"=5)
   4 - filter("A"."COL2"="C"."COL2")
   5 - filter("B"."COL3"=10)
Note
-----
   - dynamic sampling used for this statement (level=2)

当改变optimizer选择的执行计划时候，添加了hints，然后我们使用嵌套循环，驱动表为c，被驱动表为A，完了之后再作为row source1做为hash table，然后与表B进行hash join。

SQL> select /*+ USE_NL (A C)*/ A.col4
  2  from C , A , B
  3  where C.col3 = 5 and A.col1 = B.col1 and A.col2 = C.col2
  4  and B.col3 = 10;
Execution Plan
----------------------------------------------------------
Plan hash value: 2122808611
-------------------------------------------------------------------------------------------
| Id  | Operation                    | Name       | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT             |            |     1 |   110 |     4   (0)| 00:00:01 |
|   1 |  NESTED LOOPS                |            |     1 |   110 |     4   (0)| 00:00:01 |
|   2 |   NESTED LOOPS               |            |     1 |   110 |     4   (0)| 00:00:01 |
|   3 |    MERGE JOIN CARTESIAN      |            |     1 |    52 |     4   (0)| 00:00:01 |
|*  4 |     TABLE ACCESS FULL        | C          |     1 |    26 |     2   (0)| 00:00:01 |
|   5 |     BUFFER SORT              |            |     1 |    26 |     2   (0)| 00:00:01 |
|*  6 |      TABLE ACCESS FULL       | B          |     1 |    26 |     2   (0)| 00:00:01 |
|*  7 |    INDEX RANGE SCAN          | INX_COL12A |     1 |       |     0   (0)| 00:00:01 |
|   8 |   TABLE ACCESS BY INDEX ROWID| A          |     1 |    58 |     0   (0)| 00:00:01 |
-------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   4 - filter("C"."COL3"=5)
   6 - filter("B"."COL3"=10)
   7 - access("A"."COL1"="B"."COL1" AND "A"."COL2"="C"."COL2")
Note
-----
   - dynamic sampling used for this statement (level=2)
SQL>

注意当我们对表进行了分析之后，那么就不会有动态分析了，动态分析只是为了进行执行计划的选择。
对于分析表知识详解：
http://blog.csdn.net/xiaohai20102010/article/details/8777158

<pre>SQL> set autotrace off
SQL> analyze table a compute statistics;
Table analyzed.
SQL> analyze table b compute statistics;
Table analyzed.
SQL> analyze table c compute statistics;
Table analyzed.
SQL> analyze index  inx_col12A compute statistics;
Index analyzed.
SQL> select A.col4
  2  from C , A , B
  3  where C.col3 = 5 and A.col1 = B.col1 and A.col2 = C.col2
  4  and B.col3 = 10;
no rows selected
SQL> set auotrace trace explain
SP2-0158: unknown SET option "auotrace"
SQL> set autotrace trace explain
SQL> r
  1  select A.col4
  2  from C , A , B
  3  where C.col3 = 5 and A.col1 = B.col1 and A.col2 = C.col2
  4* and B.col3 = 10
Execution Plan
----------------------------------------------------------
Plan hash value: 2122808611
-------------------------------------------------------------------------------------------
| Id  | Operation                    | Name       | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT             |            |     1 |   110 |     4   (0)| 00:00:01 |
|   1 |  NESTED LOOPS                |            |     1 |   110 |     4   (0)| 00:00:01 |
|   2 |   NESTED LOOPS               |            |     1 |   110 |     4   (0)| 00:00:01 |
|   3 |    MERGE JOIN CARTESIAN      |            |     1 |    52 |     4   (0)| 00:00:01 |
|*  4 |     TABLE ACCESS FULL        | C          |     1 |    26 |     2   (0)| 00:00:01 |
|   5 |     BUFFER SORT              |            |     1 |    26 |     2   (0)| 00:00:01 |
|*  6 |      TABLE ACCESS FULL       | B          |     1 |    26 |     2   (0)| 00:00:01 |
|*  7 |    INDEX RANGE SCAN          | INX_COL12A |     1 |       |     0   (0)| 00:00:01 |
|   8 |   TABLE ACCESS BY INDEX ROWID| A          |     1 |    58 |     0   (0)| 00:00:01 |
-------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   4 - filter("C"."COL3"=5)
   6 - filter("B"."COL3"=10)
   7 - access("A"."COL1"="B"."COL1" AND "A"."COL2"="C"."COL2")
</pre>

来自为知笔记(Wiz)

posted @ 2015-06-18 19:14 hao_xiaoyu 阅读(297) 评论(0) 编辑收藏举报

会员力量，点亮园子希望

刷新页面返回顶部

CHINA ORACLE-BASE

Sharing and recording of Oracle and Linux knowledge

oracle sql 执行计划分析

公告