首页 » ORACLE 9i-23ai » oracle Memory Management 改进

oracle Memory Management 改进

Automatic Memory Management (AMM) 在ORACLE 11g中被引入,这是从OS级别内存管理的重大改变,回顾ORACLE的版本更新历程,内存管理也在逐步的更新

从ORACLE9I时的pga_aggregate_target 单一的参数统一控制了SORT_AREA,HASH_AREA等小POOL的分配

再到ORACLE10G时的ASMM,也是把shared_pool,buffer_cache等内存分配上升到了一个独立的POOL,有ORACLE自己调配各sub-pool的大小,只要指定sga_target和sga_max_target.
手动指定的方式还是保留,但只是作为自动调整时的下限

到了ORACLE11G,有了更大的改进,那就是把整个实例的内存都可以交给ORACLE现管理就是上面提到的AMM,只需要指定memory_target and memory_max_target. 系统参数,PGA and SGA就有一个大的pool来自由支配。

在oracle10G及以前的版本oracle用的是一种叫做 system v-style shared memory,可以用unix工具查看已存在的共享内存区大小比如ipcs

sys@ANBOB>select * from v$version where rownum=1;

BANNER
----------------------------------------------------------------------------------------
Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Prod

sys@ANBOB>exit
Disconnected from Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Production
With the Partitioning, OLAP and Data Mining options
[oracle@dbserver1 ~]$ ipcs -m

------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch     status      
0x4f2dd44c 2752517    oracle    640        1052770304 18      

可以看到分配了1052770304bytes的共享内存段

NATTCH 列The number of processes attached to the associated shared memory segment
[oracle@dbserver1 ~]$ ps -ef|grep 'ora_'|grep -v 'grep ora_'
oracle   29391     1  0 11:07 ?        00:00:00 ora_pmon_anbob
oracle   29393     1  0 11:07 ?        00:00:00 ora_psp0_anbob
oracle   29395     1  0 11:07 ?        00:00:00 ora_mman_anbob
oracle   29397     1  0 11:07 ?        00:00:00 ora_dbw0_anbob
oracle   29399     1  0 11:07 ?        00:00:00 ora_lgwr_anbob
oracle   29401     1  0 11:07 ?        00:00:02 ora_ckpt_anbob
oracle   29403     1  0 11:07 ?        00:00:01 ora_smon_anbob
oracle   29405     1  0 11:07 ?        00:00:00 ora_reco_anbob
oracle   29407     1  0 11:07 ?        00:00:01 ora_cjq0_anbob
oracle   29409     1  0 11:07 ?        00:00:02 ora_mmon_anbob
oracle   29411     1  0 11:07 ?        00:00:00 ora_mmnl_anbob
oracle   29413     1  0 11:07 ?        00:00:00 ora_d000_anbob
oracle   29415     1  0 11:07 ?        00:00:00 ora_s000_anbob
oracle   29419     1  0 11:07 ?        00:00:00 ora_qmnc_anbob
oracle   29425     1  0 11:07 ?        00:00:00 ora_q000_anbob
oracle   29429     1  0 11:07 ?        00:00:00 ora_q002_anbob
oracle   29431     1  0 11:07 ?        00:00:31 ora_j000_anbob
oracle   30793     1  0 16:07 ?        00:00:00 ora_j001_anbob

因为没有客户端连接目前这有这些后台进程(目前18个

[oracle@dbserver1 ~]$ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sda5 64G 14G 47G 24% /
/dev/sda3 965M 61M 855M 7% /tmp
/dev/sda1 99M 12M 83M 12% /boot
tmpfs 1010M 0 1010M 0% /dev/shm
/dev/hdb 2.9G 2.9G 0 100% /mnt/cdrom

我们再看下11G的

[oracle@rac1 ~]$ ora

SQL*Plus: Release 11.2.0.1.0 Production on Mon Jun 4 01:51:45 2012

Copyright (c) 1982, 2009, Oracle. All rights reserved.

Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production
With the Partitioning, Automatic Storage Management, OLAP, Data Mining
and Real Application Testing options

sys@ANBOB>show parameter memory_target

NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
memory_target big integer 500M

[oracle@rac1 ~]$ ipcs -m

------ Shared Memory Segments --------
key shmid owner perms bytes nattch status
0x714f0ec4 32768 grid 660 4096 0
0x49b6ab3c 98305 oracle 660 4096 0

[oracle@rac1 ~]$ ps -ef|grep ckpt
grid 5172 1 0 01:23 ? 00:00:00 asm_ckpt_+ASM
oracle 5570 1 0 01:47 ? 00:00:00 ora_ckpt_anbob
oracle 5804 5353 0 01:57 pts/0 00:00:00 grep -i ckpt

因为我是在GRID用户下运行的ASM,ORACLE用户下是ORACLE RDBMS,所以目前分配两个内存段。但是得注意,每个内存段仅4K,且nattach为0

当给实例指定了memory_target参数11G使用了操作系统的另一种叫做Posix-Style shared memory management,这种管理方式把共享内存分成了更小的颗粒,也实现了从PGA分配的内存转给SGA或反分配,如果用posix-style那ipcs就不可用了,取而代之的是一种叫shmfs 或才tmpfs虚拟文件系统管理,它们必须mount到/dev/shm下

[oracle@rac1 ~]$ df -m /dev/shm
Filesystem 1M-blocks Used Available Use% Mounted on
tmpfs 1500 482 1019 33% /dev/shm

[oracle@rac1 ~]$ ls -l /dev/shm
total 493332
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_0
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_1
-rw-r----- 1 oracle asmadmin 0 Jun 4 01:48 ora_anbob_98305_10
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_100
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_101
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_102
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_103
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_104
-rw-r----- 1 oracle asmadmin 4194304 Jun 4 01:46 ora_anbob_98305_105
.......
-rw-r----- 1 grid oinstall 4194304 Jun 4 01:23 ora_+ASM_32768_0
-rw-r----- 1 grid oinstall 0 Jun 4 01:23 ora_+ASM_32768_1
-rw-r----- 1 grid oinstall 0 Jun 4 01:23 ora_+ASM_32768_10
-rw-r----- 1 grid oinstall 0 Jun 4 01:23 ora_+ASM_32768_11
-rw-r----- 1 grid oinstall 0 Jun 4 01:23 ora_+ASM_32768_12

每个文件的粒度分0,4MB,16MB三种,当memory_max_target>1024时,颗粒为16MB,否则为4MB,当是0时,就好比虽然现在连接的是PGA,但随时有可以转为SGA

系统环境的工具可以用fuser或者pmap,

[oracle@rac1 ~]$ su - root
Password:
[root@rac1 ~]# /sbin/fuser -v /dev/shm/ora_anbob_98305_99

USER PID ACCESS COMMAND
/dev/shm/ora_anbob_98305_99:
oracle 5548 ....m oracle
oracle 5550 ....m oracle
oracle 5554 ....m oracle
oracle 5556 ....m oracle
oracle 5558 ....m oracle
oracle 5560 ....m oracle
oracle 5562 ....m oracle
oracle 5564 ....m oracle
oracle 5566 ....m oracle
oracle 5568 ....m oracle
oracle 5570 ....m oracle
oracle 5572 ....m oracle
oracle 5575 ....m oracle
oracle 5577 ....m oracle
oracle 5581 ....m oracle
oracle 5583 ....m oracle
oracle 5585 ....m oracle
oracle 5589 ....m oracle
oracle 5591 ....m oracle
oracle 5601 ....m oracle
oracle 5665 ....m oracle
oracle 5668 ....m oracle
oracle 5670 ....m oracle
oracle 5685 ....m oracle
oracle 5774 ....m oracle
oracle 6243 ....m oracle
oracle 6836 ....m oracle
oracle 6838 ....m oracle

[root@rac1 ~]# pmap 5548|grep 98305_99
38c00000 4096K rwxs- /dev/shm/ora_anbob_98305_99

note:
可以11G的这种内存管理方式更加灵活,11G的是新的里程碑。

SUMMARY
ASMM is supported with LINUX x86-64 Hugepages but AMM is not supported with Hugepages.

打赏

, ,

对不起,这篇文章暂时关闭评论。