0

What are the advantages/disadvantages of raid 1+0 vs raid 1 regarding the following:

  1. Recoverability. How easy is it to recover data from a raid 1+0 array if the raid array controller fails and you have to use a different manufactures controller (or maybe a different model from the same manufacturer)? Has anyone done this? I know with raid 1 this is possible (I have done this in the past).
  2. Reliability of the array. Does this differ between implementations (ie: Do some vendors support 2 failed disks in a 4 disk raid 1+0 array as long as these are not the other disk's pair or is this "standard")
  3. I/O. Is a 4 disk raid 1+0 always twice as fast as a raid 1 array or are there certain conditions (read/write/random/sequential) where this is not true.
  4. Array size. Again obviously raid 1+0 wins, however if there are recoverability advantages to raid 1 (above) is there a sensible solution to combine arrays without reducing reliability/recoverabilty.
  5. Anything else?

There seems to a bit of information about this on the internet, however no one really addresses the first point (which is my main concern running a home lab where reliability and recoverability are of uttermost importance) and it would be nice to cover all of the advantages/disadvantages...

maloo
  • 162
  • 1
  • 8
  • BTW - aware that RAID is not a backup, however there may be times where your backups are old and recovering that missing data is important... – maloo Jan 12 '20 at 07:05

3 Answers3

2

  1. I'll let someone with more experience than what I have answer, but this exact scenario is why I prefer software RAID where possible (at least when running Linux or BSD servers). In Windows with hardware RAID I am ready to use my backups to recover in case of a controller failure, even though it will most likely be possible to recover using an identical controller.

  2. RAID 1+0 is a striped array of mirrors. Any mirror should be its own failure domain, but you really shouldn't expect failures to adhere to the happy case: plan for the worst and enjoy the fact that this rarely happens.

  3. The I/O capacity of a RAID10 is probably up to its implementation, but I would expect the striping to help with performance in any read or write that's larger than the stripe size.

  4. The KISS principle usually applies: For robustness, stick to the most basic solution that solves your problem. In this case: Recovering from a controller failure is possible - it is a solved problem. Just make sure you don't purchase one that doesn't promise its users this ability if you're the one deciding what to buy.

Mikael H
  • 5,031
  • 2
  • 9
  • 18
2

Regarding question #1:

Don't depend on your disks being usable in another manufacturers hardware controller without data loss. Most controllers will write metadata on each disk identifying details about the RAID group and type it belongs to. AFAIK, there is no standard for this metadata. So, should you have to plug your disks into another controller, they may or may not be usable with some sort of hardware specific "initialization" (i.e. writing the metadata).

If replacing the controller is a concern, you should consider purchasing and testing a spare to keep on the shelf.

Brandon Xavier
  • 2,022
  • 13
  • 15
0

The biggest single difference between R1 and R10 is that with R1 you're limited to two disks, so your maximum capacity isn't huge - with R10 it can be several times larger.

Chopper3
  • 101,299
  • 9
  • 108
  • 239