topic Re: RAID 1+0 vs RAID 0+1 in ProLiant Servers (ML,DL,SL)

RAID 1+0 vs RAID 0+1

Andrew_235 — Tue, 22 Jun 2004 07:41:40 GMT

Hi,

Could someone explain the differences between RAID 1+0 and RAID 0+1?

Also, which does the Integrated Smart Array 5i that is ony our DL 380 G3 support? I have seen some documentation show it as RAID 0+1 and some as RAID 1+0.

As far as performance is concerned, are they as quick as eacho other?

Thanks,
Andrew

Re: RAID 1+0 vs RAID 0+1

SAKET_5 — Tue, 22 Jun 2004 08:23:10 GMT

Hi Andrew,

So here you go (everything you wish to know about the difference between 2*5 Vs. 5*2...hehe i mean RAID 1+0 vs. RAID 0+1)
http://www.pcguide.com/ref/hdd/perf/raid/levels/multXY-c.html

saves me typing....DL380 G3 will support it but in your case, if i presume it right (from your other query on the forum) you will actually need 4 drives in your RAID 1+0 set or vice versa, an even number of drives > 2.

3 drives in your existing raidset wont cut it.

performance and all your other queries will be dealt with in the above link...

hope it was of help and dont forget to assign points:)

regards,

Re: RAID 1+0 vs RAID 0+1

SAKET_5 — Tue, 22 Jun 2004 08:25:15 GMT

Hi Andrew,

another handy hint: (hope you like this)

RAID 0+1 implements striping without any drive redundancy which is mirrored, a back up of the first.

Its only advantage is speed.

RAID 1+0 I would understand this to mean RAID 10 which is fault-tolerant, where mirror volumes can be striped across multiple disks.

Takes up more disk space

P.S. dont forget to assign points:)

regards

Re: RAID 1+0 vs RAID 0+1

Uwe Zessin — Tue, 22 Jun 2004 09:00:27 GMT

I don't see why one implementation should be faster or take up more space than the other. We always have striping implemented and we always write the data to two different disks.

All storage controllers I am aware implement striped mirrorsets. The advantage is that in case of a disk failure the controller only needs to recover that disk. If a disk in a stripeset became inoperative it means the whole stripeset is affected.

Re: RAID 1+0 vs RAID 0+1

Doug de Werd — Tue, 22 Jun 2004 09:57:11 GMT

Smart controllers implement RAID 1+0, or 10. What this means is that each individual drive is mirrored, and then striped. This provides much higher availability than 0+1.

I have tested this and it really works.

This is the difference. Let's assume 8 drives. for simplicity, look at it like this:

1 3 5 7
2 4 6 8

In a RAID 0+1, the odd number drives would be striped first (essentially a RAID 0 with drives 1,3,5,7), and then the stripe set mirrored to the even number drives (another RAID 0 with 2,4,6,8). If you lost drive 3, you would still be OK, but if you lost drive 3 and drive 6, you would be hosed because both RAID 0 mirror sets would be broken.

In RAID 1+0, drives 1-2 are mirrored, 3-4, 5-6, and 7-8, then striped. You can lose any combination - up to 4 drives total, as long as they are both not a mirrored pair. So you could lose 1, 4, 5, and 8 and still be running. But if you lost any of the mirrored pairs (like 1-2, 3-4, 5-6, or 7-8) you would be hosed.

And all of this doesn't matter if you only have 2 drives because there's no striping :-)

Thanks,
Doug

In a RAID 0+1

Re: RAID 1+0 vs RAID 0+1

Tim D Fulford — Tue, 22 Jun 2004 16:31:36 GMT

Hi

You learn something new every day... Before reading this thread I was sure this was all guff & nonsense!! As far as I can see there is no difference in physical data layout between RAID1+0 & RAID0+1.. The only difference is subtle, but real, the logic and how it is implemented by the controller. This result is such that RAID0+1 has a very high probability of data loss compared to RAID1+0, especially for longer stripes.

If there are 10 disks and the probability of a single disk loss is, say, 1%, then the RELATIVE probabilies of data loss is
RAID1+0 ~ 1/9 as only one disk can fail causeing data loss out of remaining 9 disks
RAID0+1 ~ 5/9 as ANY 5 disks of the remaining 9 can fail causing data loss!
So RAID1+0 is 5 (well N/2) times more robust/fault tolerant than RAID0+1!

Regards

Tim

Re: RAID 1+0 vs RAID 0+1

Ernest Ford — Tue, 22 Jun 2004 18:20:55 GMT

This is all theoretical - for convenience I will define RAID 0+1 as a mirrored stripe set and RAID 1+0 as a striped mirror set - and I agree with the math that says a mirrored stripe set is the less reliable of the two configurations.

It's theoretical simply because I've worked with different RAID controllers that define it differently, but when you go through the explanation in the manual it comes back to the same thing - mirror the disks for redundancy and then stripe them for speed

It's also theoretical, because it's dictated by the controller logic and the user really doesn't have a choice.

I note that one gentleman claims he tested it - I'd love to know how that was done, because I tried to test it and my 5i wouldn't let me.

Re: RAID 1+0 vs RAID 0+1

Doug de Werd — Wed, 23 Jun 2004 09:12:32 GMT

I was the one who tested it.

Very simple... set it it up just like I described, and then start pulling drives out while you're doing disk I/O

If you start with 8 drives, you can pull up to 4 and it will still be running - just don't pull any 2 that are a mirrored pair!

Also, if you are going to try different combinations of drive pulls, remember that if you put a drive back, you need to wait until it has been fully rebuilt before pulling it's mirror pair.

Thanks,
Doug

Re: RAID 1+0 vs RAID 0+1

Ernest Ford — Wed, 23 Jun 2004 10:58:41 GMT

Ah - so what you in fact tested was RAID 1+0 - what I defined as a striped mirror set - that is definitely "do-able".

The difficulty that I have experienced lies in finding a controller that will allow me to configure RAID 0+1 or a mirrored stripe set.

Re: RAID 1+0 vs RAID 0+1

Doug de Werd — Wed, 23 Jun 2004 11:06:49 GMT

yes you are correct. HP Smart Controllers have always done 1+0, so that's basically all I could test.

In some of our documentation quite a few years back, we incorrectly talked about 0+1 as a feature - back when it first came out, nobody in Marketing really thought it through that there might be a difference. This lead to an internal discussion about what the difference between 0+1 and 1+0 was, and when we got the Smart Controller engineering team involved, they verified that from the very beginning, Smart Controllers have done 1+0. So we tried to change all of the references to 0+1 in our documents, but it's possible that some may have been overlooked.

And there is no option within the ACU for Smart Controllers to select 0+1 - it is always 1+0.

Thanks,
Doug

Re: RAID 1+0 vs RAID 0+1

Uwe Zessin — Wed, 23 Jun 2004 12:53:01 GMT

Ah, that is exactly why I try to avoid saying RAID-1+0 / 0+1 or at least reach an understanding with my dialog partner what the difference between 1+0 vs 0+1 is, because one can view this bottop-up or top-down.

Re: RAID 1+0 vs RAID 0+1

Tim D Fulford — Sat, 26 Jun 2004 12:30:46 GMT

Hi

You can test RAID0+1.. it is fairly expensice though!!! (and probably not of any practical value)

1 get hold of 2 complete EVA3000 with appropriate number of disks for each.
2 - Create on EVA-1 a single RAID0 LUN and on EVA-2 do the same.
3 - use LVM to miror between the two LUNs
4 - pull one disk from each of the EVA's and it will fail
5 - comment on futility of test, and how no one in their right mind would be stupid enough to do this in a live environment. (hence RAID0+1 is a dumb idea)
6 - decide never to mention RAID0+1 again....

Regards

Tim