oracle 9i pga使用控制

oracle 9i pga使用控制

我们知道从9i开始oracle提供了自动管理pga的功能，从此我们可以不管sort_area_size，hash_area_size等等烦人的设置，也不用再去计算繁琐的pga大小。但是怎么控制每个session使用pga的大小呢，会不会出现一个session用完所有pga，而其他session得不到空间的状况吗？
带着这个疑问我们可以来做一个实验。

实验的步骤很简单，初始pga_aggregate_target设定为100m，我们做一个排序的大查询，然后查看pga的使用空间。然后修改pga_aggregate_target为1g，在做一个大查询，查看pga使用空间。

SQL> startup pfile=/test/inittest.ora
ORACLE instance started.

Total System Global Area 2710294504 bytes
Fixed Size                   454632 bytes
Variable Size             805306368 bytes
Database Buffers         1903992832 bytes
Redo Buffers                 540672 bytes
Database mounted.
Database opened.

SQL> show parameter pga
NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
pga_aggregate_target                 big integer 100000000
 

SQL> conn test/test
Connected.
SQL> select name,value from v$statname a,v$mystat b where a.statistic#=b.statistic# and a.name like '%pga%';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
session pga memory                                                   243208
session pga memory max                                            243208
这里可以看到初始连接时pga使用了240k左右

SQL> set autotrace trace;

SQL> select title from test order by starts,ends,title,zoo,city;
2997531 rows selected.

Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=129273 Card=2932520 Bytes=173018680)
   1    0   SORT (ORDER BY) (Cost=129273 Card=2932520 Bytes=173018680)
   2    1     PARTITION RANGE (ALL)
   3    2       TABLE ACCESS (FULL) OF 'TEST' (Cost=84738 Card=2932520 Bytes=173018680)

Statistics
----------------------------------------------------------
        475  recursive calls
         75  db block gets
     930726  consistent gets
     962505  physical reads
         60  redo size
  113681365  bytes sent via SQL*Net to client
    2198688  bytes received via SQL*Net from client
     199837  SQL*Net roundtrips to/from client
         37  sorts (memory)
          1  sorts (disk)
    2997531  rows processed

SQL> set autotrace off;
SQL> select name,value from v$statname a,v$mystat b where a.statistic#=b.statistic# and a.name like '%pga%';
NAME                                                                  VALUE
---------------------------------------------------------------- ----------
session pga memory                                                   457496
session pga memory max                                            6355736

做完一个大查询，pga的使用峰值是达到了6m左右，这其中可能包含了一部分pga自己的开销
关闭数据库，修改pga_aggregate_target为1g，再重启数据库
SQL> conn / as sysdba
Connected.
SQL> shutdown immediate;
Database closed.
Database dismounted.
ORACLE instance shut down.
SQL> exit

$ sqlplus "/as sysdba"

SQL*Plus: Release 9.2.0.4.0 - Production on Sun Jul 18 11:46:53 2004

Copyright (c) 1982, 2002, Oracle Corporation.  All rights reserved.

Connected to an idle instance.

SQL> startup pfile=/test/inittest.ora
ORACLE instance started.

Total System Global Area 2710294504 bytes
Fixed Size                   454632 bytes
Variable Size             805306368 bytes
Database Buffers         1903992832 bytes
Redo Buffers                 540672 bytes
Database mounted.
Database opened.
SQL> conn test/test
Connected.
SQL> select name,value from v$statname a,v$mystat b where a.statistic#=b.statistic# and a.name like '%pga%';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
session pga memory                                                   250792
session pga memory max                                            381864

SQL> show parameter pga
NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
pga_aggregate_target                 big integer 1000000000
SQL> set autotrace trace;
SQL> select title from test order by starts,ends,title,zoo,city;
2997531 rows selected.

Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=112995 Card=2932520 Bytes=173018680)
   1    0   SORT (ORDER BY) (Cost=112995 Card=2932520 Bytes=173018680)
   2    1     PARTITION RANGE (ALL)
   3    2       TABLE ACCESS (FULL) OF 'TEST' (Cost=84738Card=2932520 Bytes=173018680)

Statistics
----------------------------------------------------------
       3740  recursive calls
         28  db block gets
     931577  consistent gets
     955856  physical reads
         60  redo size
  113681365  bytes sent via SQL*Net to client
    2198688  bytes received via SQL*Net from client
     199837  SQL*Net roundtrips to/from client
         77  sorts (memory)
          1  sorts (disk)
    2997531  rows processed

SQL> set autotrace off;
SQL> select name,value from v$statname a,v$mystat b where a.statistic#=b.statistic# and a.name like '%pga%';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
session pga memory                                                   456568
session pga memory max                                            50263928

可以看到这次大查询使用了近50m的pga空间，为什么第一次只使用了6m呢，这是因为oracle会限制每个session使用pga的最大空间，
每个session可以使用到最大至5%pga的总空间，所以第一次可以使用到5m左右的空间，加上pga自己的开销，使用了一共6m。第二次
由于总pga大小扩大了，所以session使用的pga大小也差不多是5%的总pga大小.oracle很好的控制了pga的使用，避免了一个session
占用大部分pga的情况。
不过，还存在一种特例，在并行操作的情况下session pga可以使用到30%总pga空间。