This post actually started as a comment on this post. It got a tad long for a comment, and had some good opinions in it, so I thought I’d just trackback to it instead.
The idea is that with more and more services like Dreamhost and Media Temple’s (gs) service available these days, that one could conceivably create an online backup system with distributed storage, utilizing cheap, readily available shared web-hosting accounts.
The issues arrise when you take into consideration the AUPs of these companies. Generally, they’re going to include some kind of clause to ensure that they can terminate accounts which start to cost them money1.
In theory, the idea sounds great. If companies like Dreamhost would let you store 100 GB of data with 1 TB of bandwidth a month for $20, you could easily cover that with a few customers and offer your backup service as a type of “reseller” of that space. In practice, however, I don’t think you’ll find too many companies offering that kind of space that would take kindly to you actually using it all. This all goes back to the entire base concept of overselling: promise the client more than you think they’ll ever actually use2.
One possible scenario I see would be to start off that way: either using straight small hosting accounts (such as through Dreamhost), or work out some kind of deal with such companies to become a sort of “reseller” using their services3 at a slightly higher cost to you than would be otherwise. Then, quickly move to some kind of self-hosted service. The idea would be to use the basic “cheater” approach until you actually got some paying clients, so that you don’t have a huge amount of overhead cost before you even get any income coming in.
Another variation on that theme might be to use dedicated hosting. You’d be paying a good bit more for your space then, but that’s what it’s for: higher-demand services. Consider that you could then get a couple hundred GB of storage and a few TB of bandwidth for under $100 a month, and you’re not yet rolling in the dough, but you also are able to offer a service to your users without risking getting shut down for a TOS violation either.
Personally, I think a perhaps more viable option may be to actually utilize S3 yourself. The key would be adding features that aren’t natively available and which the majority of people wouldn’t have the technical expertise to implement on their own. Picture something like Box.net, backed by S3. You provide the killer interfaces4, all while saying that you’re backed by the amazing S3 infastructure that the market leader Amazon uses for their own services. There’s no risk of users losing their data, because it’s all stored on this highly redundant and distributed system from Amazon. A smaller company like Box.net isn’t going to be able to make that same claim. Sure, that’s what they’re aiming for, but how confident are you in that claim?
Again, there’s nothing to say that you couldn’t move beyond S3 in the future, should you think you could provide the same reliability on your own at a lower cost, but it would be a very convenient way to at least get such a service started. Since I see one of the largest barriers to S3 adoption being that there isn’t any actual client interface software available at present (that’s widely marketed, at any rate), I think offering something like S3 Backup (re-branded for your own uses, of course) and a web-based interface to the service would be a fairly practical business model to start out with. Yes, you’d cost a little more than other companies, but I think the promise of that Amazon reliability would more than outweigh the slight price difference.
Of course, now you wonder why I haven’t actually done anything like this, since I think it’s such a great idea and have obviously put so much time into thinking about it… Well, I don’t (alone) have the technical expertise to implement any type of desktop system that would hardness S3. I also don’t have any available time to work on such a system (at least until January, when I’m free of school). Woe is me…