Automate the Sizing of your SGA in Oracle 10g
How much memory does each of the individual components of the SGA need? Oracle now has methods to determine most of the major parameters all on its own.
What is the SGA
Simply stated, the system global area (SGA) is just shared memory structures that are created at instance startup, hold information about the instance and control its behavior. The following table gives a brief synopsis of the particular components of the SGA, the variables that control the size of memory allocated, some of the areas of the Oracle server the particular component has an influence on, and then a very brief description. What can be seen from this simple list is that there are plenty of options available for us to tweak the SGA and without a complete understanding of what our applications are doing in the background, our ability to guess the appropriate amount of memory to give each of these individual components is not always optimal. What we do not want to have happen in this process of allocation of memory is to waste it.
Components of the SGA
SGACOMPONENT | SIZE CONTROLED BY | AREAS OF INFLUENCE |
SIMPLE DESCRIPTONS |
Shared Pool | SHARED_POOL_SIZE | Library Cache
|
Oracle needs to allocate & deallocate memory as SQL or procedural code is executed based on the individual needs of users' sessions and in accordance to the LRU algorithm. |
Dictionary Cache
|
Highly accessed memory structures that provide information on object structures to SQL statements being parsed. |
||
Java Pool | JAVA_POOL_SIZE |
|
Memory available for the Java memory manager to use for all things Java. |
Streams Pool | STREAMS_POOL_SIZE |
|
New to Oracle 10g, memory available for stream processing. |
Redo Log Buffer | LOG_BUFFER |
|
Holds changes made to data and allows for the reconstruction of data in the case of failure. |
Database Buffer Cache |
DB_2K_CACHE_SIZE DB_4K_CACHE_SIZE DB_8K_CACHE_SIZE DB_16K_CACHE_SIZE DB_32K_CACHE_SIZE DB_KEEP_CACHE_SIZE DB_RECYCLE_CACHE_SIZE |
|
Holds copies of data requested by SQL and reduces requests to disk by having data in memory. You may have many different buffer caches that help segregate on usage patterns. |
Large Pool | LARGE_POOL_SIZE |
|
For large memory allocations. |
You can look at the size of your SGA by looking at the initialization parameters that control its size. Here is a simple query and its output.
select name, value
from v$parameter
where name in ('shared_pool_size', 'java_pool_size', 'streams_pool_size',
'log_buffer', 'db_cache_size', 'db_2k_cache_size', 'db_4k_cache_size',
'db_8k_cache_size', 'db_16k_cache_size', 'db_32k_cache_size',
'db_keep_cache_size', 'db_recycle_cache_size', 'large_pool_size');
NAME VALUE
------------------------- ---------
shared_pool_size 83886080
large_pool_size 8388608
java_pool_size 50331648
streams_pool_size 54525952
db_cache_size 25165824
db_2k_cache_size 0
db_4k_cache_size 0
db_8k_cache_size 0
db_16k_cache_size 0
db_32k_cache_size 0
db_keep_cache_size 0
db_recycle_cache_size 0
log_buffer 262144 13 rows selected.
Letting Oracle Take Control
Using Oracle's Automatic Shared Memory Tuning, you can instruct Oracle to manage a subset of the components that make up the SGA by merely telling the instance the target size of the SGA through the new SGA_TARGET parameter. Oracle will then pool from this value and dynamically distribute memory across selected components of the SGA. You now need not set values for SHARED_POOL_SIZE, JAVA_POOL_SIZE, LARGE_POOL_SIZE, or DB_CACHE_SIZE as Oracle will automatically size them for you. Once you set the SGA_TARGET parameter to a desirable size for your SGA these parameters will take on a value of zero and new parameters will be created designated by __SHARED_POOL_SIZE, __JAVA_POOL_SIZE, LARGE_POOL_SIZE, and __DB_CACHE_SIZE. As workloads go through the system and memory is needed in these areas, Oracle will allocate more memory based on internal statistics trends. Oracle will not manage the DB_KEEP_CACHE_SIZE, DB_RECYCLE_CACHE_SIZE, DBnK_CACHE_SIZE, or the STREAMS_POOL_SIZE and you must still determine the value for these parameters. In order for this all to take place, you must be using an SPFILE as this is the only way Oracle can dynamically make all these changes happen. Also, note that SGA_TARGET is the sum of all parameters that make up the SGA, not just the parameters it controls, so you must take those components it does not control into consideration when you give a value for SGA_TARGET.
Stepping through Letting Oracle Take Control
There is really nothing to switching into automatic shared memory tuning. You only need to set the SGA_TARGET parameter.
1. Take a look and see if you are already in automated sizing of SGA
SQL> show parameter sga_target
NAME TYPE VALUE
------------------------------------ ----------- --------
sga_target big integer 0
2. Alter system to begin automated sizing of SGA
SQL> alter system set sga_target=216m;
System altered.
3. Done
What happens when you switch to Automatic Shared Memory Tuning is a bit interesting. After you alter SGA_TARGET parameter, your SPFILE will undergo a change and now have the following parameters defined. Note that k101 is my instance name and will take on whatever the instance name is.
k101.__db_cache_size=25165824
k101.__java_pool_size=50331648
k101.__large_pool_size=8388608
k101.__shared_pool_size=83886080
In addition, when you issue the previously given SQL to show the parameter settings for the individual components of the SGA you will notice that they now have a value of zero.
select name, value
from v$parameter
where name in ('shared_pool_size', 'java_pool_size', 'streams_pool_size',
'log_buffer', 'db_cache_size', 'db_2k_cache_size', 'db_4k_cache_size',
'db_8k_cache_size', 'db_16k_cache_size', 'db_32k_cache_size',
'db_keep_cache_size', 'db_recycle_cache_size', 'large_pool_size');
NAME VALUE
------------------------- ---------
shared_pool_size 0
large_pool_size 0
java_pool_size 0
streams_pool_size 0
db_cache_size 0
db_2k_cache_size 0
db_4k_cache_size 0
db_8k_cache_size 0
db_16k_cache_size 0
db_32k_cache_size 0
db_keep_cache_size 0
db_recycle_cache_size 0
log_buffer 262144
13 rows selected.
If you truly want to see the parameters after setting the SGA_TARGET you will need to modify the query to include the newly created underscore variables. This can be of some concern if you are relying on the "normal" parameters for any database monitoring scripts.
select name, value
from v$parameter
where name in ('__shared_pool_size', '__java_pool_size',
'streams_pool_size', 'log_buffer', '__db_cache_size', 'db_2k_cache_size',
'db_4k_cache_size', 'db_8k_cache_size', 'db_16k_cache_size',
'db_32k_cache_size', 'db_keep_cache_size', 'db_recycle_cache_size',
'__large_pool_size'); NAME VALUE
------------------------- ---------
__shared_pool_size 67108864
__large_pool_size 4194304
__java_pool_size 8388608
streams_pool_size 0
__db_cache_size 142606336
db_2k_cache_size 0
db_4k_cache_size 0
db_8k_cache_size 0
db_16k_cache_size 0
db_32k_cache_size 0
db_keep_cache_size 0
db_recycle_cache_size 0
log_buffer 262144 13 rows selected.
Switching over to the Automatic Shared Memory Tuning is as easy as setting an initialization parameter. How this will behave under load is yet to be determined but since these numbers are driven by the various advisories and I am mostly happy with them as individual components, I see no reason not to venture down the path of having Oracle automatically size my SGA. Of course, as always, in a test environment first. I would suggest you take a snapshot of your initialization parameters before letting Oracle take control and then compare the end settings that Oracle has implemented. It is always easy to switch back, just reset the SGA_TARGET parameter and set the individual components back to their original values.
Automate the Sizing of your SGA in Oracle 10g的更多相关文章
- oracle 10g升级到11g
Linux 上Oracle RAC 10g 升级到 Oracle RAC 11g 了解如何在 Oracle Enterprise Linux 5 上逐步将 Oracle RAC 10g 第 2 版升级 ...
- [转]oracle性能调优之--Oracle 10g AWR 配置
一.ASH和AWR的故事 1.1 关于ASH 我们都知道,用户在ORACLE数据库中执行操作时,必然要创建相应的连接和会话,其中,所有当前的会话信息都保存在动态性能视图V$SESSION中,通过该视图 ...
- Install Oracle 10g on Red Hat Linux 5.3 Step by Step
一.虚拟机配置 1. 虚拟机(VBox 4.3.12) 2. 配置虚拟机网卡网络.选择host-only.VirtualBox Host-Only Network网卡IP为设置为192.168.1.1 ...
- Oracle 10g R2 Transparent Data Encryption 透明数据加密
Oracle 10g R2 Transparent Data Encryption 透明数据加密 本章介绍如何使用透明数据加密来保护Oracle数据库中的敏感数据,该功能使您可以加密数据库列并管理加密 ...
- [转帖]达梦数据库(DM6)和ORACLE 10g的异同点
达梦数据库(DM6)和ORACLE 10g的异同点 https://bbs.aliyun.com/detail/351337.html 花花浪子 级别: 小白 发帖 0 云币 -41 加关注 ...
- 安装Oracle 10g
本文仅用于学习交流,商业用途请支持正版!转载请注明: http://www.cnblogs.com/mxbs/p/6217052.html 准备: Oracle 10g for Win(32-Bit) ...
- Oracle 10g安全加固(审计、监听密码)
环境: Linux 6.4 + Oracle 10.2.0.4 1. Oracle 10g 审计功能 2. 对数据库监听器的关闭和启动设置密码 1. Oracle 10g 审计功能 Oracle 10 ...
- ORACLE 10g 数据库体系结构图
ORACLE 10g 的数据库体系结构图(ORACLE 10g(Release 2)ARCHITECTURE),非常的全面.系统.高屋建瓴的整体介绍了ORACLE 10g 的数据库体系结构.如果能全面 ...
- RHEL6 64位系统安装ORACLE 10g 64bit 数据库
记得去年4月份的时候,为公司部署测试环境和UAT环境时,在红帽RHEL6 64位系统安装ORACLE 10g 64位数据库时遇到了许多小问题,当时匆匆忙忙也没记录一下这些问题,前几天在虚拟机安装ORA ...
随机推荐
- Hive执行过程
http://blog.csdn.net/wf1982/article/details/9122543
- (转) rabbitmq应用场景
原文:http://blog.csdn.net/wangpengblog/article/details/76405598
- java 算法 - 冒泡排序
冒泡排序: 冒泡排序是专门针对已有的一部分已经排序的数据进行排序的一种排序算法.假如你的数据中只有两个数据输乱序的,那么冒泡排序就是最快的.这种算法的核心思想就是扫描数据清单,找到乱序中相邻的两个数据 ...
- Linux 搭建Hadoop集群错误锦集
一.Hadoop集群配置好后,执行start-dfs.sh后报错,一堆permission denied zf sbin $ ./start-dfs.sh Starting namenodes on ...
- ibatis中的cdata和xml中cdata的含义
ibatis的cdata用于sqlmap文件中,二sqlmap本身就是xml文件,即解析cdata的方法与xml文件的cdata相同. 简单来说:cdata就是用来表明纯文本的,如果没有这个的话 &l ...
- MVC中使用Castle.Windsor
我在MVC中使用Castle.Windsor是这样用的. 首先在UI层安装Install Castle.Windsor 在App_Start中增加一个类WindsorActivator,用于注册和销毁 ...
- 记录一次读取memcache缓存的优化
我们是用mvc做web,大部分数据都用memcache做了缓存 有2台memcache缓存服务器 数据并不大. 某页面响应较慢,大概在4s左右. 页面本身很简单只是显示一个表单. 但是layout相对 ...
- 常常忘记但是很重要的sql语句
一.基础1.说明:创建数据库CREATE DATABASE database-name 2.说明:删除数据库drop database dbname3.说明:备份sql server--- 创建 备份 ...
- C# 小软件部分(一)
自己在空闲时分整合.编写了一款小软件程序,命名为魔法兔子,希望大家可以提出意见和指导,此篇文章主要为软件的部分截图和介绍. 软件详情: 1.首先是登录,注册界面. 可以注册自己的账号,后台是腾讯云服务 ...
- js中的深浅拷贝
js中的深浅拷贝 js中有深拷贝.浅拷贝一说,所谓的深浅拷贝是针对value类型为引用类型(函数.对象.数组)而言的,大概理解的就是: 浅拷贝: 拷贝出的对象c和原始对象o,c和o在key对应的val ...