Skip to Main Content

Breadcrumb

Question and Answer

Tom Kyte

Thanks for the question, Robert.

Asked: March 28, 2009 - 7:59 pm UTC

Last updated: June 22, 2009 - 12:20 pm UTC

Version: 10.2.0

Viewed 1000+ times

You Asked

Hi Tom,

If we are using ASM (and the data is externally protected (mirrored disks)) ---
Should our redo multiplexing strategy change?

Currently without ASM our strategy is to multiplex online and archived redo to the flash recovery area (one copy to the 'regular' file system and one copy to the FRA).

However as we migrate toward ASM it has been suggested that since the ASM disks are mirrored, that the added multiplexing redundency is no longer required.
The idea is that nowadays, with huge mirrored SAN systems, the main reason for redo multiplexing is to guard against human/operator error (e.g. DBA or SA accidentally deletes a redo log)..... But that with ASM, since DBA's and SA's will not be navigating around the ASM file system, there is little likelihood of ever accidentally deleting a redo file.... hence no need to multiplex the redo.

What are your thoughts on this?

Thank you,

Robert.




and Tom said...

hardware based mirroring can go bad as well. Mirroring is not a guarantee. If it was, you could just skip backing up. Mirroring is a higher level of confidence.

Since redo is so important (lose it, lose committed work), I don't see the point in skimping on it. Do you?

Multiplexing decreases the change of losing redo.
Mirroring the disk further decreases it.

I would not want to increase it, only decrease it. Hence, multiplexing is still in vogue.

You would want to make sure these huge SAN systems had a little bit of thought put into them as well - that you have access to physically segregated components of disk/hardware - so that a single fault won't hit everything.

Rating

  (3 ratings)

Is this answer out of date? If it is, please let us know via a Comment

Comments

Thank you Tom.

Robert, March 30, 2009 - 11:17 am UTC


Raj, June 22, 2009 - 3:33 am UTC

Hi Tom,
I multiplex but I have a problem. I have two ASM diskgroups, +DATA(tier1 SAN superfast disk) and +FRA(tier2 SAN not so fast to save money). +FRA is only used to keep backups.
I have two members in each online log group but BOTH are on +DATA and so are all controlfile copies. I dont think its wise to put one member on slower disk(+FRA) because it will affect lgwr performance.
So my questions are:
1)Am I being foolish to put both members of log groups on the same diskgroup?
2)Do I need to create another diskgroup(tier1 san) just for the sake of multiplexing my redo logs?
3)Will putting a controlfile on slower disk affect database performace significantly?

Thank you,
Raj.

Tom Kyte
June 22, 2009 - 12:20 pm UTC

depends on how protected that diskgroup is, what is the redundancy on it

could putting control files on slower devices affect performance? depends if your foregrounds are waiting on something control file related. The only answer is "it depends, it might - it might not"

Recovery Area on slower disk

Raj, July 28, 2009 - 6:50 pm UTC

Hi Tom,
Thank you for your reply. My asm diskgroups are "external" redundancy as the underlying asm disks are LUNS (on RAID). A LUN can be lost for eg. misallocation due to human error, in which case I lose the whole diskgroup and all my online redo logs along with it. I have created a separate diskgroup to remedy this and mirrored the redo logs on it.

I felt the need to comment here because DBA's might get the impression that they can use slower disk for flash recovery area, but they shouldn't. Slower disk can be used for backups but not flash recovery area (area to keep ALL files needed for recovery with space managment by oracle). Oracle/EMC docs say that we can use the "inner" cylinders for FRA and "outer" cylinders for DATA, but that is not the same thing as using disk of a completely different bandwidth.

Thank you.