Oracle 分区裁剪

explain plan for select count(*)  from  esb2_trans_log t
 where t.trans_date >= sysdate - 3 / 1440;
   
   
   select * from table(dbms_xplan.display());
   
          PLAN_TABLE_OUTPUT
1    Plan hash value: 287218081
2     
3    ------------------------------------------------------------------------------------------------------------
4    | Id  | Operation                 | Name           | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
5    ------------------------------------------------------------------------------------------------------------
6    |   0 | SELECT STATEMENT          |                |     1 |     8 |   878K  (1)| 02:55:42 |       |       |
7    |   1 |  SORT AGGREGATE           |                |     1 |     8 |            |          |       |       |
8    |   2 |   PARTITION RANGE ITERATOR|                |   177 |  1416 |   878K  (1)| 02:55:42 |   KEY |     9 |
9    |*  3 |    TABLE ACCESS FULL      | ESB2_TRANS_LOG |   177 |  1416 |   878K  (1)| 02:55:42 |   KEY |     9 |
10    ------------------------------------------------------------------------------------------------------------
11     
12    Predicate Information (identified by operation id):
13    ---------------------------------------------------
14     
15       3 - filter("T"."TRANS_DATE">=SYSDATE@!-.002083333333333333333333333333333333333333)



  select * from table(dbms_xplan.display_cursor('btfkf4uww4zv4'));

PLAN_TABLE_OUTPUT
SQL_ID  btfkf4uww4zv4, child number 0
-------------------------------------
select count(*)  from  esb2_trans_log t  where t.trans_date >= sysdate 
- 3 / 1440
 
Plan hash value: 287218081
 
------------------------------------------------------------------------------------------------------------
| Id  | Operation                 | Name           | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT          |                |       |       |   878K(100)|          |       |       |
|   1 |  SORT AGGREGATE           |                |     1 |     8 |            |          |       |       |
|   2 |   PARTITION RANGE ITERATOR|                |   177 |  1416 |   878K  (1)| 02:55:42 |   KEY |     9 |
|*  3 |    TABLE ACCESS FULL      | ESB2_TRANS_LOG |   177 |  1416 |   878K  (1)| 02:55:42 |   KEY |     9 |
------------------------------------------------------------------------------------------------------------
 
Predicate Information (identified by operation id):
---------------------------------------------------
 
   3 - filter("T"."TRANS_DATE">=SYSDATE@!-.002083333333333333333333333333333333333333)
 

 
 explain plan for select *
  from esb2_trans_log t
 where t.trans_date >=
       to_date('2018-05-11 08:00:00', 'yyyy-mm-dd hh24:mi:ss');
       
       select * from table(dbms_xplan.display());

       PLAN_TABLE_OUTPUT
1    Plan hash value: 2926439834
2     
3    -----------------------------------------------------------------------------------------------------------
4    | Id  | Operation                | Name           | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
5    -----------------------------------------------------------------------------------------------------------
6    |   0 | SELECT STATEMENT         |                |   155 | 32395 |     2   (0)| 00:00:01 |       |       |
7    |   1 |  PARTITION RANGE ITERATOR|                |   155 | 32395 |     2   (0)| 00:00:01 |     8 |     9 |
8    |*  2 |   TABLE ACCESS FULL      | ESB2_TRANS_LOG |   155 | 32395 |     2   (0)| 00:00:01 |     8 |     9 |
9    -----------------------------------------------------------------------------------------------------------
10     
11    Predicate Information (identified by operation id):
12    ---------------------------------------------------
13     
14       2 - filter("T"."TRANS_DATE">=TO_DATE(' 2018-05-11 08:00:00', 'syyyy-mm-dd hh24:mi:ss'))



分区表有很多好处，以大化小，一小化了，加上并行的使用，在loap中能往往能提高几十倍甚至几百倍的效果。当然表设计得不好也会适得其反，效果比普通表跟糟糕。
为了更好的使用分区表，这里看一下分区表的执行计划。
PARTITION RANGE ALL：扫描所有分区
PARTITION RANGE ITERATOR：扫描多个分区，小于所有个分区数量
PARTITION RANGE SINGLE：扫描单一的分区
KEY，表示执行时才知道哪个分区