Home » Server Options » RAC & Failsafe » multiplexing to reduce I/O contention (10.2)
multiplexing to reduce I/O contention [message #288892] Wed, 19 December 2007 00:36 Go to next message
asangapradeep
Messages: 128
Registered: October 2005
Location: UK
Senior Member
Hi,

What’s the best way to organize various files in a RAC environment with two ASM disk groups?

There are two disk groups created one disk group would be used for data files (named DATA) and other would be used as the flash recovery area (named FLASH). From what I read this is what oracle recommends (two disk groups) Question comes when it comes to multiplexing various files such as control files, online redo log files and even archive logs.

The SAN has two controllers managing all its disks and all disks from one controller are assigned to one data group.

By default online redo log files are multiplexed on the two disk groups. Wouldn’t this create contention between database writer process (dbwn) and the log writer process (lgwr) ?

If all online redo log files are multiplexed on one single disk group wouldn’t it defeat the whole purpose of multiplexing?

Another thing to be considered is where to put archive logs. Putting them in the FLASH disk group would create additional contention between lgwr and arc processes.
Archiving the logs locally on each RAC node would make it risky in case a node is completely lost and need to do recovery.

So I would like to know what would be the ideal configuration for a production system.

thank you.





Re: multiplexing to reduce I/O contention [message #289020 is a reply to message #288892] Wed, 19 December 2007 05:08 Go to previous messageGo to next message
ebrian
Messages: 2794
Registered: April 2006
Senior Member
It really depends on how the storage subsystem is configured. You may want to simply consider the S.A.M.E. methodology.
Re: multiplexing to reduce I/O contention [message #289028 is a reply to message #289020] Wed, 19 December 2007 05:25 Go to previous messageGo to next message
asangapradeep
Messages: 128
Registered: October 2005
Location: UK
Senior Member
Hi,


Since disk groups are ASM does SAME has to be implemented on the underlying disk subsystem? Won't ASM take care of it since The SAME approach is the foundation of the Oracle10g Automatic Storage Management (ASM) approach
Re: multiplexing to reduce I/O contention [message #289947 is a reply to message #289028] Wed, 26 December 2007 17:11 Go to previous message
ebrian
Messages: 2794
Registered: April 2006
Senior Member
ASM can be configured to be a form of SAME, however Oracle actually recommends that you use external RAID for data protection. The SAME methodology can be utilized to provide the redundancy for the disk and then ASM disk group would be configured with EXTERNAL REDUNDANCY.
Previous Topic: RAC Windows 2003 server hangs
Next Topic: what tool can be possible to simulate the EMC storage?
Goto Forum:
  


Current Time: Sat Nov 23 22:51:08 CST 2024