I have a small business providing IT support to very small companies (30 people and under). We back up our customer's data to our office using rsync over SSH (scripted, using pre shared keys). The service has proven to be so popular that I need a way to expand my storage periodically without creating new logical volumes on the server each time I do. What would be a good Linux solution for this? I'd like to just have one logical volume that I can expand as needed - either by adding storage to a device or by adding nodes to my LAN. My clients are quite small so the budget is very limited (right now I back up 17 clients to a Mac Mini (of all things) that is attached to 4TB of storage in two logical volumes.

Suggestions involving CENTOS are especially appreciated, as this is by far my favorite distro.

There is a distro called openfiler which is a NAS server, (Network Attached Storage), this would be what you are after. It's free and is easy to install. It allows you to add drives as it is needed, and uses a web based administration tool so you can maintain it from anywhere, not sure what distro it is based on, definitely worth looking at.

Hi,

Your first priority should be redundancy. It is one thing to back stuff up, but being able to restore is what counts. Make sure you have some sort of real hardware raid controller helping you out. You might consider something like a Dell MD3000 or MD1000. You can get them refurbed on outlet.dell.com for the cheap (Relatively speaking). They CAN take SATA drives (Up to 15 of them). Perfect if you get 15 TB Drives and set it up as raid 10 or 5 (For maximum redundancy, Go raid 10).

------------------------------------
Michael Piekarski
Network Engineer
mpiekarski@hostmysite.com
www.hostmysite.com

Thanks for the replies; I think the issue with both suggestions is one of cost. In order for this product to be worthwhile financially, we need to keep the costs quite low. I was hoping for an approach a bit more like OpenAFS. Has anyone implemented open AFS? Could I scale a single logical volume just by adding more nodes to my LAN?

You might want to look at AoE ATA-over-Ethernet

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

http://www.coraid.com/

Ok im a little confused, Openfiler is free to use, so it costs you nothing.

AFS works on a file system level, so I'm not sure this would do what you want. I haven't used AFS before but it's probably worth looking at.

AoE (like another poster mentioned) provides block-level access to storage devices over the network, e.g. a SAN, which may help you. Or iSCSI. Both provide block-level access to storage over a network such that you could add the devices to an existing logical volume on another system.

AoE sends the data in Ethernet frames so it is faster but possibly less reliable (I think in most uses of AoE, it's used on a dedicated storage LAN so this is less of an issue). iSCSI sends the data over TCP so it is slower but more reliable. Since your customers are backing up over the Internet speed is probably not an issue. iSCSI is more widely used / supported.

And Coraid makes good RAID boxes for AoE, but again, you'd have to consider spending some money.

- Arch

You have to spend money to make money as they say.. You are offering a service to your customers to backup their critical data, and you are going to charge them for that service and perceived security those off-site backups provide, yet you are trying to bargain basement your solution to the point where you won't spend any cash..

I currently backup ALL my servers locally to a large RAID array and to a remote Array/SAN at my backup solution providers location across the internet. similar to what you are describing. I had to pay $12,000.00 for the license for the Secondary Server backup License to run on their server (it's portable license I own it and can setup a secondary server of my own at a later date). The on-site software License was 20,000.00 and annual maintenance / support is around $6,000.00. on top of that I pay for the amount of storage space used on the off-site server storage array. I'm telling you this so you can get an idea of some of the actual costs that can involved to the customer for this type of solution if they want to do it themselves.

Do you have any idea what Data is worth ? If our company lost all their data we would be out of business.. Backups are not something you skimp on., especially with data.that belongs to someone else.

If you aren't even going to attempt to do this properly, by putting a decent storage solution in place, and I'm not even talking something expensive like a SAN, or even what I have in place, just something that is fault tolerant and not cobbled together with twine, spit and duct tape.. . Compliance (HIPPA, PCI, GLBA, etc.. dictates off-site backups and DR planning.. people MUST do this and they should be aware it's not free. The suggestion for the used DELL PE array and the AoE solution are reasonably priced large volume options which you can grow as needed and you seem to have discounted just for the fact that you would have to actually buy something..

Data is worth far more than you seem to admit, please don't do this if you aren't going to be the least bit serious about it. you may be able to afford to screw around, but I would bet your customers can't.

[/rant]

I'm not looking for bargain basement! The economics just have to work. I've priced this product pretty competitively and with only 17 small customers signed up there is no way it would be profitable for $20,000 upfront and and $6,000 per year! We're talking under 3TB of data here.

As the epilogue, here is what I decided on: I copies all the users on a dedicated Centos VM and used a script to copy all of my rsync keys there. Then I have NFS shares that this VM accesses on my physical Centos servers (which only fill the role of VMWare hosts other than this NFS role). I have a central folder on the VM with symbolic links to the NFS mounts. Now when a drive gets too full I can just rsync a few customers' data to a new larger disk on another system, then simply change the symbolic link to hit the new location. Each night, the customers rsync their data to these locations Perfect! Now I have nice scalable storage without the fear of RAIDs collapsing and losing a bunch of customers' data in on swoop. I for one frankly just don't trust RAIDs any more. This way if a disk dies I can just just re copy the customers' data and start fresh. Since these backups are purely for disaster recovery (there is another onsite backup for routine nightly backups there) everything should be nice and safe.

Just wanted to share in case my experience can help others!