SQL优化—nested loop优化

跑批时间段22:00-23:00，生成AWR报告

 

分析sql:SQL_ID='5hfw4smzs2pqw'

 执行计划：

SQL> select *  FROM TABLE(DBMS_XPLAN.DISPLAY_CURSOR('5hfw4smzs2pqw',NULL,'ALL'));

SQL_ID  5hfw4smzs2pqw, child number 0

-------------------------------------

SELECT a.SEQ_NO,a.ACCTNO,a.POST_TXN_CD,b.POST_METHOD_CD,b.AMT_TYPE,a.BIL

L_AMT,a.POST_DATE,a.EXPLAIN FROM T_EMPLOYY_A a INNER JOIN

T_EMPLOYY_B b ON a.POST_TXN_CD=b.POST_TXN_CD AND a.POST_DATE=:1  AND

a.ACCTNO=:2  AND POST_STATUS_CD='0'

Plan hash value: 3635671702                  

--------------------------------------------------------------------------------------------------

| Id  | Operation                    | Name              | Rows  | Bytes | Cost (%CPU)| Time     |

--------------------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT             |                   |       |       |  5248 (100)|          |

|   1 |  NESTED LOOPS                |                   |     1 |    71 |  5248   (1)| 00:00:01 |

|   2 |   NESTED LOOPS               |                   |     1 |    71 |  5248   (1)| 00:00:01 |

|*  3 |    TABLE ACCESS FULL         | T_EMPLOYY_A    |     1 |    61 |  5247   (1)| 00:00:01 |

|*  4 |    INDEX UNIQUE SCAN         | PK_T_EMPLOYY_B |     1 |       |     0   (0)|          |

|   5 |   TABLE ACCESS BY INDEX ROWID| T_EMPLOYY_B    |     1 |    10 |     1   (0)| 00:00:01 |

--------------------------------------------------------------------------------------------------

Query Block Name / Object Alias (identified by operation id):

-------------------------------------------------------------

   1 - SEL$58A6D7F6

   3 - SEL$58A6D7F6 / A@SEL$1

   4 - SEL$58A6D7F6 / B@SEL$1

   5 - SEL$58A6D7F6 / B@SEL$1

Predicate Information (identified by operation id):

---------------------------------------------------

   3 - filter(("A"."ACCTNO"=:2 AND "A"."POST_DATE"=:1 AND "POST_STATUS_CD"='0'))

   4 - access("A"."POST_TXN_CD"="B"."POST_TXN_CD")

Column Projection Information (identified by operation id):

-----------------------------------------------------------

   1 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200],

       "B"."POST_METHOD_CD"[NUMBER,22], "B"."AMT_TYPE"[CHARACTER,2]

   2 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200],

       "B".ROWID[ROWID,10]

   3 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200]

   4 - "B".ROWID[ROWID,10]

   5 - "B"."POST_METHOD_CD"[NUMBER,22], "B"."AMT_TYPE"[CHARACTER,2]

Note

-----

   - this is an adaptive plan

分析：

       执行计划：3-->2-->4-->5-->1-->0

       第一层循环：根据POST_DATE=:1,ACCTNO=:2,POST_STATUS_CD='0'嵌套循环，全表扫描驱动表T_EMPLOYY_A；

       第二层循环：根据a.POST_TXN_CD=b.POST_TXN_CD，嵌套循环，根据索引PK_T_EMPLOYY_B，得到rowid，访问表T_EMPLOYY_B 

索引信息：

SQL> select TABLE_NAME,INDEX_NAME,COLUMN_NAME from dba_ind_columns where TABLE_NAME='T_EMPLOYY_A';

TABLE_NAME           INDEX_NAME           COLUMN_NAME

-------------------- -------------------- ------------------------------

T_EMPLOYY_A       PK_T_EMPLOYY_A    SEQ_NO

SQL> select TABLE_NAME,INDEX_NAME,COLUMN_NAME from dba_ind_columns where TABLE_NAME='T_EMPLOYY_B';

TABLE_NAME           INDEX_NAME           COLUMN_NAME

-------------------- -------------------- ------------------------------

T_EMPLOYY_B       PK_T_EMPLOYY_B    POST_TXN_CD

表T_EMPLOYY_A条件列、连接条件字段没有索引 

表的总行数

SQL> select count(*) from SCOTT.T_EMPLOYY_A;

  COUNT(*)

----------

   2029447

SQL> select count(*) from SCOTT.T_EMPLOYY_B;

  COUNT(*)

----------

        52   

连接列的匹配情况：

SQL> select count(distinct(POST_TXN_CD)) from SCOTT.T_EMPLOYY_A;

COUNT(DISTINCT(POST_TXN_CD))

----------------------------

                          26

SQL> select count(distinct(POST_TXN_CD)) from SCOTT.T_EMPLOYY_B;

COUNT(DISTINCT(POST_TXN_CD))

----------------------------

                          52             

ACCTNO列在T_EMPLOYY_A表的筛选性

SQL>  select count(distinct(ACCTNO)) from SCOTT.T_EMPLOYY_A;

COUNT(DISTINCT(ACCTNO))

-----------------------

                 225427

POST_DATE列在T_EMPLOYY_A表的筛选性

SQL> select count(distinct(POST_DATE)) from SCOTT.T_EMPLOYY_A;

COUNT(DISTINCT(POST_DATE))

--------------------------

                       736

SQL> select count(*) from ( select distinct ACCTNO,POST_DATE from  SCOTT.T_EMPLOYY_A);

  COUNT(*)

----------

   1690282

优化建议：1：在表T_EMPLOYY_A的ACCTNO, POST_DATE列创建联合索引     

2：利用hind,修改驱动表为T_EMPLOYY_B.但是需要修改SQL语句,需上线调整,所以不使用。

ACCTNO筛选性强，放在联合索引列的前面                                 

创建联合索引

SQL> create index IDX_ACCTNO_POST_DATE on SCOTT.T_EMPLOYY_A (ACCTNO, POST_DATE)   tablespace SCOTT_IDX01

SQL> select *  FROM TABLE(DBMS_XPLAN.DISPLAY_CURSOR('5hfw4smzs2pqw',NULL,'ALL')); 

SQL_ID  5hfw4smzs2pqw, child number 0

-------------------------------------

SELECT a.SEQ_NO,a.ACCTNO,a.POST_TXN_CD,b.POST_METHOD_CD,b.AMT_TYPE,a.BIL

L_AMT,a.POST_DATE,a.EXPLAIN FROM T_EMPLOYY_A a INNER JOIN

T_EMPLOYY_B b ON a.POST_TXN_CD=b.POST_TXN_CD AND a.POST_DATE=:1  AND

a.ACCTNO=:2  AND POST_STATUS_CD='0'

Plan hash value: 1730680787

--------------------------------------------------------------------------------------------------------------

| Id  | Operation                             | Name                 | Rows  | Bytes | Cost (%CPU)| Time     |

--------------------------------------------------------------------------------------------------------------

|   0 | SELECT STATEMENT                      |                      |       |       |     5 (100)|          |

|   1 |  NESTED LOOPS                         |                      |     1 |    71 |     5   (0)| 00:00:01 |

|   2 |   NESTED LOOPS                        |                      |     1 |    71 |     5   (0)| 00:00:01 |

|*  3 |    TABLE ACCESS BY INDEX ROWID BATCHED| T_EMPLOYY_A       |     1 |    61 |     4   (0)| 00:00:01 |

|*  4 |     INDEX RANGE SCAN                  | IDX_ACCTNO_POST_DATE |     1 |       |     3   (0)| 00:00:01 |

|*  5 |    INDEX UNIQUE SCAN                  | PK_T_EMPLOYY_B    |     1 |       |     0   (0)|          |

|   6 |   TABLE ACCESS BY INDEX ROWID         | T_EMPLOYY_B       |     1 |    10 |     1   (0)| 00:00:01 |

--------------------------------------------------------------------------------------------------------------

 Query Block Name / Object Alias (identified by operation id):

-------------------------------------------------------------

    1 - SEL$58A6D7F6

   3 - SEL$58A6D7F6 / A@SEL$1

   4 - SEL$58A6D7F6 / A@SEL$1

   5 - SEL$58A6D7F6 / B@SEL$1

   6 - SEL$58A6D7F6 / B@SEL$1

Predicate Information (identified by operation id):

---------------------------------------------------

   3 - filter("POST_STATUS_CD"='0')

   4 - access("A"."ACCTNO"=:2 AND "A"."POST_DATE"=:1)

   5 - access("A"."POST_TXN_CD"="B"."POST_TXN_CD")

Column Projection Information (identified by operation id):

-----------------------------------------------------------

   1 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200],

       "B"."POST_METHOD_CD"[NUMBER,22], "B"."AMT_TYPE"[CHARACTER,2]

   2 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200],

       "B".ROWID[ROWID,10]

   3 - "A"."SEQ_NO"[NUMBER,22], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8],

       "A"."POST_TXN_CD"[NUMBER,22], "A"."BILL_AMT"[NUMBER,22], "A"."EXPLAIN"[VARCHAR2,200]

   4 - "A".ROWID[ROWID,10], "A"."ACCTNO"[VARCHAR2,40], "A"."POST_DATE"[CHARACTER,8]

   5 - "B".ROWID[ROWID,10]

   6 - "B"."POST_METHOD_CD"[NUMBER,22], "B"."AMT_TYPE"[CHARACTER,2]

Note

-----

   - this is an adaptive plan

两天跑批时间段22:00-23:00的性能对比

经优化后，数据库逻辑读下降，CPU负载下降 

20180808逻辑读：

 

20180809逻辑读：