Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

5446

September 1st, 2010 12:00

Information Storage and Management Book ISM Page 60

If someone has this book, please tell me that page 60 figure 3-7 is confused. RAID 1+0 and RAID 0+1 are in the wrong places right? I have looked at all other sources for nested RAID/ RAID 10/RAID 01 and the ISM book doesn't follow any other source I have looked at. I have attached the picture

Thanks,

Mark

1 Attachment

September 29th, 2010 03:00

Hello Mark,

I have re-posted my hand-drawn diagram again - you should not need to rotate it this time (well, that's the plan).

There are 4 pages to the diagram.

  • Page 1 - a healthy RAID 0+1 layout. Notice we do the striping first (RAID 0), and then the mirroring, to produce the final RAID 0+1.

  • Page 2 - one of the drives in the RAID 0+1 has failed (bolt of lightning). Notice that, since RAID 0 is not redundant, the effect is that we lose the entire RAID 0 stripe. This means that when we replace the failed drive, we actually have two drives to rebuild - we have to rebuild the entire RAID 0 stripe from its mirrored copy. This RAID 0+1 diagram has a total of 4 drives. Imagine instead that it was 10 drives, then, when one drive fails, you would need to rebuild 5 drives.

  • Page 3 - a healthy RAID 1+0 layout. Notice we do the mirroring first (RAID 1), and then the striping, to produce the final RAID 1+0.

  • Page 4 - we lose a single drive. As you can see, since RAID 1 is redundant, we are in much better shape. So, when we replace the failed drive, we only have that one drive to rebuild. Also, we may be able to survive (still access our data), certain combinations of multiple drive failures (eg. losing the 0,2,4,6 drive at the top, and the 1,3,5,7 drive at the bottom).

So the main distinction is this:

  • RAID 0+1     Lose a drive - rebuild the entire stripe (typically n/2 drives, where n is the total number of drives in the 0+1 layout).
  • RAID 1+0     Lose a drive - rebuild only the failed drive.

Does that make sense?

Cheers.

1 Attachment

September 2nd, 2010 21:00

Mark..

The images are correct.

If you take a look at RAID 1+0 : Imcomming data is Mirrored and then Striped which the way it is

RAID 0+1 : imcomming data is first Striped and then Mirrored.

Guys please correct me if I am wrong.

Cheers!!

September 3rd, 2010 07:00

Hi Mark,

I've attached a diagram which is another way of looking at this. Please bear with me, it is 4 pages and is just hand-drawn.

To quickly explain the diagram:

  • 0,1,2,3,4,5,6,7 are 'strips'  (this is the ISM terminology. You may be familiar with this as the 'stripe-unit' - Veritas, 'interlace' - Disk Suite / Solaris Volume Manager, 'element' - Clariion etc etc. I drew the diagram a long time ago, before ISM, and I used the term 'stripe-unit').
  • The first column is physical - hard drives.
  • The second column is the first layer of logical abstraction.
  • The third column is the second layer of logical abstraction.
  • As a small exercise for the reader, there is a tiny mistake - a digit 2 that should be a 3.

Please let me know if you have any questions. I hope the diagram helps.

Thanks, Richard.

1 Attachment

29 Posts

September 27th, 2010 15:00

Well here is one source that is confusing me. Look at the Six disk 0+1 example. I must be looking at something wrong. It is the same picture as (a) RAID 1+0 figure on page 60.

http://en.wikipedia.org/wiki/Nested_RAID_levels

September 28th, 2010 04:00

Hello Mark,

The wiki diagram is really the same as my hand-drawn diagram (mine is constructed left to right, whilst the wiki diagram is constructed bottom to top; also, the wiki diagram uses 6 disks for 0+1, whilst my diagram uses just 4. Apart from these minor differences, the diagrams are logically identical). If you find the page 60 diagram confusing, then I would just forget about it and refer to one of the other two diagrams for your understanding. If you understand either my diagram or the wiki diagram, then you understand the distinction between 1+0 and 0+1.

Best regards.

September 28th, 2010 07:00

By the way, I reckon RAID 1+0 vs RAID 0+1 has probably been one of the hottest/most popular discussion topics in storage management over the last 15 years, or maybe longer. There has also been confusion in the field about which one was striped mirrors, and which one was mirrored stripes - (mirroring + striping) was often called striped mirrors, and (striping + mirroring) was often called mirrored stripes.

However, RAID 0+1 is now often considered to be legacy/historical. In the last 15 years the only product that I have personally encountered (and there will be others that I am unaware of) where you could only implement 0+1, and had no choice to implement 1+0, was with Veritas Volume Manager version 2.x. When Veritas Volume Manager 3.x shipped (many years ago), then Veritas introduced a new layered volume architecture. This new architecture allowed the implementation of 1+0, along with other nice features such as dynamically migrating from one RAID level to another (eg. 5 -> 1+0).

There can be 1+0 subtle implementation differences between different vendors and different products. As one example, some vendors allow 1+0 using an odd number of disks, whilst many vendors mandate an even number of disks.

Cheers.

29 Posts

September 28th, 2010 09:00

Thank you very much for your responses. They are helpful. I think the main thing that will help me understand which is which is this: which Nested RAID 1+0 or 0+1 can sustain only a two disk failure. So if I have 10 disks and I take 5 disks and stripe them and then take the other 5 disks and stripe them and the data is mirrored between the two stripes I can only have 2 disks fail from the whole group and not from the same striped set. So which Nested RAID is this?

thanks,

Mark

29 Posts

September 30th, 2010 13:00

DING DING DING! That is what I expected, but just needed it clear. Thanks for all your responses. I completely understand now.

Mark

No Events found!

Top