We love S3. So much so that we will be adding S3 as a possible destination for CrashPlan customers this year. That said, there are a few points we've discovered when working with their APIs that affect using them as a backup service.

1. Guarenteed restore. The problem with S3 is once you send your data there, you have to TRUST it's still in working order. Want to verify it's good? You have to do a restore, and that costs money. CrashPlan intelligently tests your backup archive at the destination automatically to validate your backup archive is still in good working order. You can't do this with S3.

2. It's cheaper. Amazon charges you every time you move data to and from their data center. There are no monthly fees with CrashPlan if you back up to your own destination (a friends, computer, the office, a neighbor). Our central storage is a flat $0.10 / GB / month, no bandwidth charges, also cheaper than amazon.

3. No limits. Amazon has a 4GB file size limit. CrashPlan has none. We'll leverage our engine to circumvent this limit, but most (all?) products today currently have a 4GB limit I think.

4. Better performance. S3's limited API resulting in poor performance. CrashPlan has better compression, less bandwidth.

5. Better security. With CrashPlan you know where your data is. It's not in the "cloud" you trust to work. Some people are ok with this, for me personally, I like knowing my macbook pro is backed up to my old XP box in the basement of my house & in our rack at the office.

S3 is an awesome service, but it's limited API limits what you can accomplish in terms of backup & restore. I think they're a great addition to a backup strategy, but I personally would not make them my primary off-site strategy.

2nd or 3rd tier? Absolutely.

~Matthew