I have been testing with this product for some of our clients. It is easy to use and definately worth looking at. Limitation is bandwidth, but fairly quick and data is encrypted and sent to offsite storage. Restoration is just as easy, and there are several options for fequesncy of backup. After the initial backup, you can set it to take a snapshot several times during the day.

So... how much incremental data is transmitted during a hard drive defrag?

so better than using a gmail account then

I use Jungle Disk, which can backup your Mac, Windows, and Linux computers to Amazon's S3 data centers. Jungle Disk is a one-time $20 cost and you only pay Amazon for the bytes you actually use.

Plus you can mount your Jungle Disk backups from another PC. It's like an infinitely-large USB drive.

http://www.jungledisk.com/

Biggest thing that annoys.... no network drive support for either of these products....

Have a very small client network with a NAS for storing all data (Raid 5 - so at least some protection from loss) - but want off site backup as well.

This would be perfect, but unless I want to go to the effort of double staging - ie, copying everything from the NAS onto a local drive and then off to Mosy / Carbonite (which I don't want to hassle with and frankly don't have a single area large enough to stage it all)... it does not help at all..

Shame really..

How can it be possible to provide client-side encryption AND block level backups? Surely to allow block level backups, the files must be transferred unencrypted? There would be little point of doing a block level "diff" of two encrypted files.

Brian,

Unless your defrag re-organised the data in the files it wouldn't make a scooby's of difference to this file level (albeit operating at a block level within the file) backup tool.

AC -

File level with encryption is easy ... when you do a backup you see if a file has been modified, if it is you encrypt the file as a sequence of encrypted blocks and send a hash for those blocks to the backup server the backup server then tells you by comparing the hashes with those for encrypted blocks it holds for that particular file which blocks to send. This can be optimised by the application holding a database of "last state" encrypted block hashes locally eliminiating the need for the remote check. I've seen both methods work.

The real downside of the encryption (apart from local processing) is the amount of disk space you consume - in an organisation you do this sort of backup and the server will de-duplicate disk blocks to ensure for example the 50 copies of debbie does dallas stored on peoples hard drives actaully only is stored once - but referenced 50 times. Encryption makes this impossible.

Big question does it work with Linux?

The per-block encryption you describe would only work if the blocks stay aligned. If you prepend one byte to a file, no block hashes will match, the more efficient approach is to use rolling checksums. I know all this because I'm one of the independent competitors to these mentioned products (my app is S3 Backup) and currently implementing the block-level differential scheme as well.

I second the guy who voted for Jungledisk (and S3). It might not be as cheap (in theory) as an "all-you-can-eat" plan, but $0.15 per GB stored per month (plus bandwidth) is still cheap, and you get network drive support, block level uploads, user defined encryption keys, yada yada yada. AND I'd rather trust my data to Amazon, frankly (who copy it across multiple geographically remote sites). My experience with some of these "all you can eat" operations was great right up until I wanted to restore some lost data, then the *really* slow response times (still waiting after days) got my goat and I thought - "hey you gets what you pays for here..."

Oh, and Jungledisk does Windows, Max OS and Linux (all included for the one price).

And no, I'm not linked to them, just a very happy customer.