Help understanding what RAID I should use for my storage server setup

Question

I have Googled as much as I can but really haven't found an answer that I can understand.

I'm creating a website where I want to give each user 5gb of disk space where their personal files, theme pages, and script pages will be served from.

I have looked at every host provider and decided to go with SoftLayer because they have 24HDD servers.

My idea on how I would like this to work out is

1 Web server to hold web pages for the landing portion of the site

1 Storage server where all files that belong to the user are stored

I want to mount the storage server onto the web server so that i can easily redirect users to the mounting point via php.

Now, my question is that because I am giving users 5gb of space I would need a lot of space to get a good handful of registered users on board per server. SoftLayer offers 2TB HDDs, so I was thinking of going with 12x2TB drives and having the other 12 as a raid or backup

What would be a good production environment here, they only offer RAID 0,1,10,5 at the moment. Because the I plan on seeing a high IO on the storage server I'm afraid that a disk would fail and there being days of down time in rebuilding a raid.

What are some good suggestion for my case?

Or, what would be a good solution for my storage needs

Answer 1

You seem to be confused. RAID is not BACKUP, but you use these terms interchangeably.

All your eggs are in 1 basket (all disks are on the same server) so whatever you do: I will not call it BACKUP. It saves you from a disk-failure, but is no protection whatsoever if the data-center catches fire or someone hacks into it and does a "rm -fr *" A proper BACKUP is off-site (or at least in a different fire-zone) and not accessible at the file-system level from the system that is backed up.

Raid 10 seems most appropriate. Else Raid 6: RAID 5 or 6 with that many disks means days of rebuilding if a disk fails. Most raid5 implementations by the way degrade in operational performance beyond 9 disks. At least Raid 6 will give you protection against a 2-disk failure, which is safer when you are having long rebuild times. So I would rather use 6 (if possible) than 5.

Regardless off all the above: I'm really wondering if this is all such a smart idea. Given the price of cloud storage these days it would appear it's a lot cheaper and certainly less hassle if you just would hire cloud-storage from the likes of Amazon or Google. Let them worry about the backup, redundancy, etc.

Besides: Standard solutions are easier to maintain and manage than home-brew. Don't forget: If you roll your own, you are the only one who can support it. If you sell a 24/7 SLA to a customer you are the one who needs to be on call 24/7 to handle support.

Answer 2

I'd avoid RAID 0 like the plague with so many disks, just imagine - lose one of those 24 disks and you lose everything!

RAID 1 is really RAID 10 when you have more than 2 disks.

RAID 5 would worry me on an array with so many disks in, the rebuild time and exposure to array failure during rebuild gives me shivers.

RAID 6 (if you can configure this) would give you a lot of space, be more resilient that RAID 5 and perform fine for this kind of setup, especially if you're not doing >80% writes.

Certainly a RAID 10 (i.e. 12+12) setup will be secure and actually probably faster than you can serve out.

So I'd suggest either R6 or R10.