DL380G4 RAID 1+0: Really?

[Rob Leadbeater]

Hi,

I'd been doing some searching of my own, and came across the same document as Marek.

http://docs.hp.com/en/J6369-90036/ch01s03.html

Based on HP's own definition, I'd go back on my previous post, and say that the use of the term RAID 1+0 when referring to a mirror of two disks, is incorrect.

Cheers,

Rob

[Adrian Clint]

My interpretation of this has always been:

If you have 2 drives then configuring RAID1 or RAID1+0 always results in a RAID1 config.

The change happens when you expand the array.

You can expand a RAID1 array a drive at a time and you just proceed to get one more mirrored copy of the orginal each drive you add.

You can expand the RAID1+0 array 2 drives at a time and when you do it starts to stripe on the new disks.

Never tried it but thats they way I assumed it works. The RAID1+0 option for 2 drives is there for those that only have 2 drives initially but want to go to 4+ drives later.

Does that make sense?

[PDMeat]

I think these replies are evidence of why HP's use of the term RAID1+0 is confusing re two drives. :) Thanks all!

What I find interesting too, is the fact that if RAID1 using 2 disks can read from either disk (performing like a 2 disk RAID0 array) and if you have significant cache memory (128mb+ BBWB) in the controller to buffer all writes, is there really any difference in performance between a RAID0 stripe set vs RAID1 set *with only 2 disks*?

Even more food for thought is that other OEMs to include Intel now offer multiple RAID types on the same set of physical disks. (not quite the same thing we're talking here...)
http://www.intel.com/design/chipsets/matrixstorage_sb.htm
http://www.ocforums.com/showthread.php?t=467848

It would nice to talk to an HP storage engineer and get their 2c on why the RAID1+0 designation is used for 2 disks.

[HendrikFrans]

I think..


RAID1+0 mirrors the data onto another disk first, and then stripes the data across more disks (obviously, only when more disks become available)

RAID0+1 stripes the data across multiple disks first, and then mirrors the stripe set.
B cannot happen before A has completed.
They are not the same.

[shermaine_1]

Dear Guys,
Thanks for so much information given.I went throught some of the document on RAID level configuration on HP, and finaly confirm that for a heavy load MS SQL database system, the best is implement RAID 1+0.

Currently my database is running on one logical drive with 2 HDD @ HP RAID 1+0 configuration, my server is Hp proliant G4 380 with Smart array 6404/128 controller card, deos any one have any idea whether the I/O performance will get improved if I migrate the RAID 1+0 from 2 HDD to 4 hdd?
If yes, how much difference?

Many thanks for any idea!
Shermaine