Dear Lazyweb,
I want to buy some online data storage.
I don’t want to have to learn any new APIs for accessing my storage. I already know how to interact with files in a computer filesystem, so I’d like to just access my cloud storage that way. In other words, I want to buy a networked mount point with a designated amount of storage behind it, where I’m charged based either on the amount of storage reserved or on the amount used, I don’t care which. It’s okay if it’s slow: we can copy data to faster local disk when we’re working with it. I just want a place to put large amounts of data, a place that’s backed up by someone who’s paid to back it up, such that it’ll be easily accessible to programs running on a server.
(Actually, in an ideal world I want to find two such offerings, so I can use them both and have one be a backup of the other, for organizational redundancy in our backups.)
I would have thought this service would be completely commoditized by now, but apparently not — or, possibly, I’m just not searching for it the right way.
I’ve been looking at Wikipedia’s Comparison of File Hosting Services, and maybe one of those will turn out to be it. I also had a conversation about using Ceph to do this, in the #ceph chatroom on the OFTC IRC network (many thanks to the people there who responded to my questions). Word there is no one’s offering this yet with Ceph, though it might be an offering in the future.
Any ideas?
It’s not a perfect fit for your requirements, but the Tarsnap service provides a very cheap cloud version of a tar file ( http://www.tarsnap.com/about.html ). I’ve only discovered Tarsnap in the last week or so, so I can’t vouch for it yet; but the post mortem of the most recent AWS outtage looks promising ( http://www.daemonology.net/blog/2012-07-04-tarsnap-outage.html ).
You might find Gluster useful here. It was built to do exactly this, with the bonus that you can keep your local data local.
Ec2 plus SMB or NFS won’t do it?
Max, Gunnar, Eric, thanks for your comments! Taking them individually:
@Max: Very interesting. Tarsnap doesn’t seem to be a simple mount point — you have to download client code, which isn’t even open source (it’s under a no-derivs license). The apparently one-person-show nature of the company itself is also not quite the bus factor > N that I was looking for :-).
@Gunnar: Gluster is software — I’m looking to buy a service. It looks like we could deploy Gluster across a bunch of AMIs, and scale it out as needed… as long as we remember to monitor it correctly, and we get the configuration right, etc. But the whole point is to pay a service that does all that for us and just bills us. I want to buy a mount point, not a chunk of EBS that I then have to learn how to stitch together. My brain is already full, and I’m at the point where trading brain for money is a good deal. Hey — does Red Hat sell this?
@Eric: Well, I guess my answer to Gunnar above says it. Yes, we could stitch this together ourselves. But the whole point is not to have to do that. We don’t want to have to think about it, we just want to get a monthly bill and have access to our storage through filesystem calls that we already learned.
Ah, and on Identi.ca, James Vasile recommends searching for “vps storage”. Results promising so far…
What about rsync.net?
@Josh: rsync.net might be just the thing — reading more now at rsync.net/resources/faq.html. Thanks for pointing it out.
On Linux I use Box.com as a WebDAV filesystem (davfs2). In the past I’ve used GMX.com the same way.
My prejudices would lead me to trust rsync.net more for production use, however.