The new forums will be named Coin Return (based on the most recent vote)! You can check on the status and timeline of the transition to the new forums here.
The Guiding Principles and New Rules document is now in effect.
Multi-disk File Server
electricitylikesme
Registered User regular
Registered User regular
So, my family home has a file server with about 10 drives of various sizes in it for serving movies, videos, the VoIP phones etc. We don't want to stripe any of these disks into a RAID array since that makes it likely that we just lose everything. At the same time though, we would like all the disks to be treated as a single logical drive for the purposes of network sharing.
The problem is, spanned drives are just as bad as far as I can tell - if one goes down you potentially lose everything.
What I would like, is a way to keep all the disks independent, with an identical folder structure on each so when any folder is opened from the network all the files across all the disks are listed as though they were one. Even better would be if files could be copied and be put as a single file onto just one of the disks.
Can this be done? Has anyone ever heard of a system like this? Would RAID-5 just be a better approach?
The problem is, spanned drives are just as bad as far as I can tell - if one goes down you potentially lose everything.
What I would like, is a way to keep all the disks independent, with an identical folder structure on each so when any folder is opened from the network all the files across all the disks are listed as though they were one. Even better would be if files could be copied and be put as a single file onto just one of the disks.
Can this be done? Has anyone ever heard of a system like this? Would RAID-5 just be a better approach?
electricitylikesme on
0
Posts
The only reasonable approach for your situation would be a RAID5 array, but remember that any RAID5 array is going to base it's max size on the minimum capacity of the member disks, so if you have a 40GB and two 60GB, you'll only get 80GB of usuable space (80GB wasted).
Formula for RAID5 capacity (for those who don't already know):
S x N-1
Where S=Smallest physical disk size
N= total number of disks
Add an "A" instead of the "only" and this is right. RAID was designed for this problem. R.edundant A.rray of I.nexpensive D.isks.
I think he meant "place them in a RAID-0 (striping)", which would kill everything upon a single disk failure.
No, not really... in a striped RAID array (RAID 0), if you lose one disk it is possible that you will lose at least one piece of every file stored in the array, as in this type of RAID each file is written in chunks across multiple disks in order to speed access time, although you end up with only one copy of each chunk of the file.
RAID 5 would be a good solution if you have at least 3 disks of the same size, otherwise you will only be able to use as much space on each disk as the size of the smallest drive. If you set up the array using mdadm in Linux you can still use the left over space, though, just not as part of the RAID 5.
I believe that you could set this sort of thing up in two different ways (under Linux... you don't mention what you're running). You could use Samba for file sharing and set up DFS, which basically allows you to group a number of separate file shares together so that they appear as a single file share. You can read about it here:
http://www.faqs.org/docs/samba/ch08.html
Your other option would be to disable all of your existing disk shares, set up a single directory share that contained symlinks to each of the mounted disks root directories. This way you wouldn't have to have 10 mapped drives in Windows, you'd have one that appeared to contain all the rest.
All disk raid but one does striping, and i didn't consider 0 at all as it's not a redundancy solution. Mostly we use 6 now as we have TB's of arrays.
But yeah, good point. Don't raid 0.
First off, the noise and power consumption has got to be a little high. Second, if you're concerned about reliability, having 10 drives really increases the likelihood that one of them is going to fail. There's a reason the 777 has two big engines instead of four little ones. All things being equal, you've got 5x the chance of having a disk failure in a 10-disk array instead of a 2-disk array. If the drives are relatively small (say 10-50GB) then it may be worth it in terms of reliability to just replace a bunch of them with a single 500GB disk. RAID with some kind of mirroring or parity is a nice solution but replacing and rebuilding drives when one fails is a pain in the ass.
I'm also not sure how well modern (software or hardware) RAID controllers handle disks that are of different sizes. The smallest drive in your array may determine the size of your array.
OP - Read a bit more about raid. http://en.wikipedia.org/wiki/RAID. If I were you, I'd raid what I had. If that's not feasible with your current hardware a few 300GB drives with some raid 5 would give you a single unified partition and redundancy if one drive died.
No, I didn't say to replace 10 drives with 1, but replacing 10 drives with 2 or 3 big ones is probably more reasonable.
This is a purely economic analysis. Let's say you want to minimize hassle. If a drive fails, you have a big hassle because you've got to go out and get a new one and rebuild the array. If you lose data, you have a big hassle because you lost your data. My hypothesis is that the expected hassle doesn't go down as the number of disks goes up. For one drive, you have too much risk of data loss. For 10 drives, you have too much risk of disk failure, but you haven't reduced your risk of data loss much. I think there's a sweet spot somewhere in the middle there.
I can't really talk though - I had all my stuff on a single 250MB drive until it decided to grind some holes in the platters. And No, I didn't have a backup...