1

We are planing cloud solution for our customers and we are really concerned with designing safest storage servers for their files.

This storage won't be IO demanding, as it will contain only files which are uploaded by users. Databases will be stored on completely other servers, and are not topic of this discussion.

We are ordering 2 storage servers where one of them will replicate other one.

I'm torn between these two approaches:

  1. 12 HDD split in 6 x 2 HDD in Raid 1
  2. put all 12 HDD in Raid 6

The other machine would replicate first one, and would have the same choice of disk organising.

I'm not troubled by sharding in case 1., i can easily solve that in my software and i'm not worried by loosing storage space, or storage speed.

Because, we plan to store large amount of data, we have chosen NL-SAS (basically SATA) drives.

If we take safety as main deciding factor, which approach would you choose and why.

Thanks

Improve this question
0

1 Answer 1

Reset to default
1

That's simple: If you are after safety from disk failures - use Raid 6. It handles 2 drive failures (complete or bit errors) at the same time. Your solution number 1 is basically a raid 10 which handles most but not all 2-disk-errors and is therefor less safe than Raid 6.

Improve this answer
5
  • But, am i wrong to think that if i have second computer doing replication of same data on another two disks which are also in raid 1, basically i would have 3 tolerable drive failures?
    – Goran Radulovic
    Jun 17, 2013 at 11:47
  • 1
    In reality, no. While there's a high chance of one drive in both to fail (so you have two), there's always the chance of a double-drive failure on the same server resulting in loss of the array.
    – Nathan C
    Jun 17, 2013 at 11:52
  • Yep, but with Raid 6 you have still one tolerable failure more (4) and more space net. Besides that consider the amount of time it needs to rebuild a complete array during which you run on reduced redundancy while your remaining disks are under heavy io - mostly restoring data which is not really lost on the failed array...
    – tim
    Jun 17, 2013 at 12:19
  • @tim when you say that rebuilding array takes time, you are talking about RAID6, right?
    – Goran Radulovic
    Jun 17, 2013 at 15:16
  • @GoranRadulovic rebuilding an array after a drive failure takes time on any raid level (except RAID 0, which has no provision for such things - you spend your time restoring from backups instead).
    – voretaq7
    Jun 17, 2013 at 16:04

Not the answer you're looking for? Browse other questions tagged .