Oracle 物理读,逻辑读的理解

Oracle 物理读,逻辑读的理解

1.物理读(physical read)

当第一次读取数据块时,就会缓存到buffer cache 中, 而第二次读取和修改该数据块时就在内存buffer cache了，以下是例子:

1.1  第一次读取:

# su - oracle

$ sqlplus /nolog

SQL*Plus: Release 10.2.0.4.0 - Production on Thu Jul 21 20:12:46 2011

Copyright (c) 1982, 2007, Oracle.  All Rights Reserved.

idle> conn lsf/lsf

Connected.

lsf@ORCL> create table test as select * from all_objects;

Table created.

lsf@ORCL> set autotrace traceonly

lsf@ORCL> select * from test;

50120 rows selected.

Execution Plan

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

Plan hash value: 217508114

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

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

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

|   0 | SELECT STATEMENT  |      | 58253 |  7281K|   156   (1)| 00:00:02 |

|   1 |  TABLE ACCESS FULL| TEST | 58253 |  7281K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

 

Statistics

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

        288  recursive calls

          0  db block gets

       4086  consistent gets

        690  physical reads  --690个物理读

          0  redo size

    2903969  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

1.2  第二次读取

lsf@ORCL> select * from test;

50120 rows selected.

Execution Plan

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

Plan hash value: 217508114

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

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

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

|   0 | SELECT STATEMENT  |      | 58253 |  7281K|   156   (1)| 00:00:02 |

|   1 |  TABLE ACCESS FULL| TEST | 58253 |  7281K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

 

Statistics

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

          0  recursive calls

          0  db block gets

       3982  consistent gets

          0  physical reads –-没有发生物理读，直接从buffer cache中读取

          0  redo size

    2903969  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

1.3  数据块被重新读入buffer cache ,这种发生在

如果有新的数据需要被读入Buffer Cache中，而Buffer Cache又没有足够的空闲空间，Oracle就根据LRU算法将LRU链表中LRU端的数据置换出去。当这些数据被再次访问到时，需要重新从磁盘读入。

 

lsf@ORCL> alter session set events 'immediate trace name flush_cache';

Session altered.  --清空数据缓冲区

lsf@ORCL> select * from test;

50120 rows selected.

Execution Plan

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

Plan hash value: 217508114

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

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

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

|   0 | SELECT STATEMENT  |      | 58253 |  7281K|   156   (1)| 00:00:02 |

|   1 |  TABLE ACCESS FULL| TEST | 58253 |  7281K|   156   (1)| 00:00:02 |

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

 

Note

-----

   - dynamic sampling used for this statement

Statistics

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

          0  recursive calls

          0  db block gets

       3982  consistent gets

        691  physical reads  --又重新发生了物理读

          0  redo size

    2903969  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

2.逻辑读(buffer read)

逻辑读指的就是从（或者视图从）Buffer Cache中读取数据块。按照访问数据块的模式不同，可以分为即时读（Current Read）和一致性读（Consistent Read）。注意：逻辑IO只有逻辑读，没有逻辑写。

• 即时读

即时读即读取数据块当前的最新数据。任何时候在Buffer Cache中都只有一份当前数据块。即时读通常发生在对数据进行修改、删除操作时。这时，进程会给数据加上行级锁，并且标识数据为“脏”数据。

lsf@ORCL> select * from test for update;

50120 rows selected.

Execution Plan

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

Plan hash value: 4255797980

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

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

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

|   0 | SELECT STATEMENT   |      | 58253 |  7281K|   156   (1)| 00:00:02 |

|   1 |  FOR UPDATE        |      |       |       |            |          |

|   2 |   TABLE ACCESS FULL| TEST | 58253 |  7281K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

 

Statistics

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

        139  recursive calls

      51136  db block gets

       4826  consistent gets

          1  physical reads

   10456788  redo size

    5754048  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

• 一致性读

Oracle是一个多用户系统。当一个会话开始读取数据还未结束读取之前，可能会有其他会话修改它将要读取的数据。如果会话读取到修改后的数据，就会造成数据的不一致。一致性读就是为了保证数据的一致性。在Buffer Cache中的数据块上都会有最后一次修改数据块时的SCN。如果一个事务需要修改数据块中数据，会先在回滚段中保存一份修改前数据和SCN的数据块，然后再更新Buffer Cache中的数据块的数据及其SCN，并标识其为“脏”数据。当其他进程读取数据块时，会先比较数据块上的SCN和自己的SCN。如果数据块上的SCN小于等于进程本身的SCN，则直接读取数据块上的数据；如果数据块上的SCN大于进程本身的SCN，则会从回滚段中找出修改前的数据块读取数据。通常，普通查询都是一致性读。

下面这个例子帮助大家理解一下一致性读：

会话1中：

lsf@ORCL> set autotrace traceonly;

lsf@ORCL> select * from test;

50120 rows selected.

 

 

Execution Plan

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

Plan hash value: 217508114

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

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

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

|   0 | SELECT STATEMENT  |      | 49816 |  6227K|   156   (1)| 00:00:02 |

|   1 |  TABLE ACCESS FULL| TEST | 49816 |  6227K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

 

Statistics

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

        288  recursive calls

          0  db block gets

       4076  consistent gets –没有发生update之前

        690  physical reads

          0  redo size

    2903969  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

lsf@ORCL> update test set owner='LSF';

50120 rows updated.

Execution Plan

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

Plan hash value: 3859524075

 

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

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

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

|   0 | UPDATE STATEMENT   |      | 49816 |   827K|   156   (1)| 00:00:02 |

|   1 |  UPDATE            | TEST |       |       |            |          |

|   2 |   TABLE ACCESS FULL| TEST | 49816 |   827K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

Statistics

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

         87  recursive calls

     103217  db block gets

       1634  consistent gets

          0  physical reads

   26599188  redo size

       1125  bytes sent via SQL*Net to client

       1180  bytes received via SQL*Net from client

          4  SQL*Net roundtrips to/from client

          2  sorts (memory)

          0  sorts (disk)

      50120  rows processed

会话2中:

lsf@ORCL> set autotrace traceonly

lsf@ORCL> select * from test;

50120 rows selected.

Execution Plan

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

Plan hash value: 217508114

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

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

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

|   0 | SELECT STATEMENT  |      | 49816 |  6227K|   156   (1)| 00:00:02 |

|   1 |  TABLE ACCESS FULL| TEST | 49816 |  6227K|   156   (1)| 00:00:02 |

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

Note

-----

   - dynamic sampling used for this statement

Statistics

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

          0  recursive calls

          0  db block gets

      97113  consistent gets –没有事务做update时是4076个consistent gets 说明多了93037个consistent gets，多出来的consistent gets是从回滚段中获取的。

          0  physical reads

      44204  redo size

    2903969  bytes sent via SQL*Net to client

      37243  bytes received via SQL*Net from client

       3343  SQL*Net roundtrips to/from client

          0  sorts (memory)

          0  sorts (disk)

      50120  rows processed

--End--