raid 0+1 or 1 for redologs and archives why ?

[Emilio Brusa]

Hi.

I read some documents for raids in Oracle

I read "RAid 0+1 should not be used with redo logfiles "
OR
"Avoid RAid 0+1 whit Redologs and archive logs"

But idont understand why.
is for the performance or the cost ?

If raid 0+1 is faster 1, why no put the redo and archive in raid 0+1 ?

Thanks a lot

E.

[Florian Heigl (new acc)]

I'm a bit confused here too, as the redologs and archive logs only get large, sequential updates to my understanding.

but, I'd guess, the risk of losing this data is just too high, and a 0+1 scenario might have bad consequences in case of controler failures.

but this is only my guesswork, I don't know if there is a more specific reason.

0+1 needs as many disks as a regular raid1 of the same size, and the performance should be better, so I think there must be another reason.

but I think raid1 will do well enough, if the actual spindles in Your storage aren't shared to other lun's any current system should easily be able to cope with redolog writing, so 0+1 might just be an unnecessary risk.

[Nguyen Anh Tien]

Hi Emillo
Cost: raid 0+1 is more expensive than raid5
Performance: raid 0+1 is better than raid 5.
Raid 0+1 is faster than raid5 (no need to caculate Parities)
that all.

[Steven E. Protter]

Reasoning:

Redo logs are write intenseive. Raid 5 writes take a log longer than Raid 1 writes.

Oracle recommends Raid 1 or Raid 10 for redo logs and data for this reason.

I know shops that get away with raid 5 for archive logs, because the writes are simple and sequential.

The reason is perfomrance and reliability and because Oracle says so, and they should know a bit about their product.

SEP

[A. Clay Stephenson]

I've learned over the years to take Oracle's disk recommendations with a grain of salt. I really can't think of any valid reason to avoid RAID 1/0 for any Oracle files. 1/0 is my RAID level of choice. I do avoid RAID5/RAID5DP/RAID6 for any Oracle because of the added overhead. I always prefer performance over disk capacity utilization.

Now if they had said avoid RAID 0 -- that would have made sense because there is no redundancy.

My criteria in order of importance are reliability, performance, and disk capacity. RAID 1/0 at least gets the first two nicely.

[Florian Heigl (new acc)]

You people are are that 1/0 and 0/1 are two different things? ;)

I have a 1+0 setup at my home mailserver and I learned to be sorry about this decision - the OS in question (windows 2000) can't really handle stale PEs and on unexpected reboots the system wastes a lot of time and performance re-syncing the mirrors...

[Emilio Brusa]

Thanks everyone!!!

Florian:

I`m confused.
Why 0+1 might just be an unnecessary risk ?

E.

[TwoProc]

Emilio,

I've only seen those documents saying to avoid RAID 5 for redo and archive logs, not the other way around.

I agree with Clay that it is certainly best when it is all RAID 0/1. But, if you can't do that - then redo, archive logs, temp files, and undo file should be RAID 0/1. Also, if you can, separate hot write active tables/tablespaces onto RAID 0/1.

Florian, please enlighten me how Raid 1/0 is different (if that can be short). I just ignorantly assumed they were the same (tia).

[Jean-Luc Oudart]

Hi

like metioned in other posts avoid RAID5 alltogether. Saying that if your drives are on a SAN you may not see the difference.
Can afford mirroring (RAID1) use it !

see attachement SAME method (Oracle)

Regards
Jean-Luc

[Yogeeraj_1]

hi,

You may be having limited resources. Below the guidelines that we follow:

Normally, you should try to dedicate specific devices to the following:
a. online redo
b. archive
c. temp

For online redo logs AND control files, use: no raid, raid 0 or raid 0+1
=======================================
Oracle will multiplex them even if you mirror them. There are opportunities. You should still let Oracle multiplex them by its own even if you mirror them. Oracle has more opportunities for failure if the raid subsystem reports a "warning" back to Oracle -- if Oracle has multiplexed them -- Oracle is OK with that.

For temporary datafiles (used with temporary tablespaces): no raid or raid 0
=======================================
No raid/raid 0 is sufficient. If you lose these, who cares? You want speed on these, not reliability. If a disk fails, drop and recreate temp elsewhere.

For archives: no raid, raid 0 or raid 0+1
=======================================
Again, let Oracle multiplex if you use no raid or raid 0, let the OS do it (different from online redo log here) if you use 0+1.

For Rollback: raid 0+1
========================
High write activity. It is important to have protected. Oracle cannot multiplex them so let the OS do it (unless you are using Automatic Storage Management - ASM). Use this for datafiles you believe will be HEAVILY written. Bear in mind, Oracle buffer writes to datafiles, they happen in the background so the poor write performance of raid 5 is usually OK except for the heavily written files (such as rollback).

For datafiles that experience Medium or Moderate write activity: raid 5
=======================================
Because this activity takes place in the background typcially (not with direct path loads and such). Raid 5 can typically be safely used with these. As these files represent the BULK of your database and the above represent the smaller part -- you achieve most of the cost saving without impacting performance too much.

hope this helps too!

regards
yogeeraj