Direct I/O (DIO) and Concurrent I/O (CIO) on AIX 5L
Subject: Direct I/O (DIO) and Concurrent I/O (CIO) on AIX 5L
Doc ID: Note:257338.1 Type: BULLETIN
Last Revision Date: 13-SEP-2005 Status: PUBLISHED
PURPOSE
-------
To discuss the use of Direct I/O (DIO) and Concurrent I/O (CIO) on AIX 5L
SCOPE & APPLICATION
-------------------
System Administrators managing AIX 5L that has Oracle RDBMS on top.
DIRECT I/O (DIO) AND CONCURRENT I/O (CIO) on AIX
------------------------------------------------
If you use a Journaled File System, it is easier to manage and maintain database
files than it is if you use raw devices. In prior versions of AIX, file systems
used to support only buffered read and write and added extra contention due to
undesirable inode locking. These two issues are solved by JFS2's Concurrent I/O
and GPFS' Direct I/O, allowing file systems to be used in place of raw devices
even when optimal performance is required.
AIX 5L includes Direct I/O support on JFS, JFS2 and GPFS. In addition, AIX 5L
version 5.2 and later includes Concurrent I/O support. Direct I/O and Concurrent
I/O support allows database files to exist on file systems while bypassing the
operating system buffer cache and removing inode locking operations that are
redundant with functionality provided by Oracle Database 9i. Where possible,
Oracle recommends isolating Oracle data files on their own dedicated file
system(s) on which the Concurrent I/O or Direct I/O option is enabled.
The following table lists file systems available on AIX and the recommended
setting:
File System Option Descrīption
----------- ------ -----------
JFS dio Concurrent I/O is not available on JFS.
Direct I/O (dio) is available, but performance
is degraded compared to JFS2 with Concurrent I/O.
JFS large file none Oracle does not recommend using JFS big file for
Oracle Database because it's 128 KB alignment
constraint prevents you from using Direct I/O.
JFS2 cio Concurrent I/O (cio) is a better setting than
Direct I/O (dio) on JFS2 because it has support
for multiple concurrent readers/multiple
concurrent writers on the same file.
GPFS - Oracle Database 9i release 2 and later silently
enables Direct I/O on GPFS for optimum
performance. GPFS already supports multiple
reader/multiple writers/multiple nodes.
Therefore, Direct I/O and Concurrent I/O are the
same thing on GPFS.
If GPFS is used, the same file system can be used for all purposes including
$ORACLE_HOME, datafiles, logs. A large GPFS block size (typically at least 512KB)
should be used for optimal performance. GPFS is designed for scalability and there
is no requirement to create multiple GPFS file systems as long as the amount of
data fits in a single GPFS file system.
If placing the Oracle Database's logs on a JFS2 filesystem, the optimal configuration
is to create the file system with option "agblksize=512" and to mount it with option
"cio". This delivers logging performance similar, within a few percents, to the
performance of a raw device.
In Oracle Database 9i, Direct I/O and/or Concurrent I/O can only be enabled at
filesystem level on JFS/JFS2 and therefore $ORACLE_HOME and datafiles have to be
placed in separate filesystems for optimal performance: $ORACLE_HOME would reside on
a file system mounted with default options, whereas datafiles and logs would reside
on file systems mounted with the "dio" or "cio" option.
Having on ORACLE_HOME on a filesystem mounted with "cio" option is not supported.
Such a configuration will cause, installation, relinking and other unexpected
problems.
來自 “ ITPUB部落格 ” ,連結:http://blog.itpub.net/7318139/viewspace-893841/,如需轉載,請註明出處,否則將追究法律責任。
相關文章
- Direct I/O and Redo writes.
- WARNING:Could not increase the asynch I/O limit to 164 for SQL direct I/OMITSQL
- 讀書筆記 Improving Database Performance With AIX Concurrent I/O筆記DatabaseORMAI
- WARNING:Could not lower the asynch I/O limit to 160 for SQL direct I/O.MITSQL
- expdp匯出時卡死 Could not increase the asynch I/O limit to for SQL direct I/OMITSQL
- WARNING:Could not increase the asynch I/O limit to 224 for SQL direct I/O.MITSQL
- 【AIX】AIX 下磁碟 I/O 效能分析AI
- expdp匯出時卡死 Could not increase the asynch I/O limit to XXX for SQL direct I/OMITSQL
- Veritas Quick I/O and Cached Quick I/OUI
- 計算機I/O與I/O模型計算機模型
- I/O埠和I/O記憶體記憶體
- AIX 下磁碟 I/O 效能分析[轉]AI
- Java I/OJava
- AIX系統磁碟I/O效能評估AI
- Python 音訊 I/O 庫 PyAudioPython音訊
- Java(8)I/OJava
- 【java】I/O流Java
- Java I/O流Java
- 檔案I/O
- I/O基礎
- java的I/OJava
- c++ I/OC++
- Python教程:精簡概述I/O模型與I/O操作Python模型
- goldengate extract abended unable to queue I/O, I/O beyond file sizeGo
- 【面試】I/O 複用面試
- Java 非同步 I/OJava非同步
- 網路I/O模型模型
- 系統級 I/O
- JAVA I/O系統Java
- GDAL並行I/O並行
- 優化磁碟I/O優化
- 02. I/O 操作
- 從資料庫層面理解:隨機 I/O & 順序 I/O資料庫隨機
- 在AIX下為Oracle調整磁碟I/O引數AIOracle
- AIX使用非同步I/O(aio)提高系統效能(轉)AI非同步
- Linux下的5種I/O模型與3組I/O複用Linux模型
- not set asynch I/O limit to nnnn for SQL directo I/O.It is set to 200_848859.1MITSQL
- 流?I/O 操作?阻塞?epoll?