Friday, October 19, 2007

Using S3 for Customer Backups

For some time we've been using S3 to back up our customers systems.

Originally we set them up to FTP the data back to our office. This was handy for us because if we wanted to look at their data we had it in-house. But as the number of clients and the size of their data grew we ran out of bandwidth. (Thankfully we did these transfers at night so it didn't slow down our daytime internet access.)

The other problem was the sheer size of the data and the issues of how to guard against disk failure. Although we don't promote this to our clients as a "backup" service many of them still end up falling back on us when they discover their own backups weren't done or were no good. And people often don't discover problems till days or weeks later so they need older copies, not just the most recent.

We decided to give S3 a try and so far it has worked out well. We're up to about 8 gb of data transfer per night and we currently have about 240 gb of storage. This is costing us about $50 per month - a bargain as far as I'm concerned.

We are currently keeping the last 8 days, last 5 weeks, last 13 months, and every year, or something like 25 - 30 copies per customer. On top of this redundancy, Amazon says they store multiple copies of each file.

A minor downside is that when we want to look at someone's data we have to download it, but that's not a big deal. And it's still a lot easier to download from S3 than to download directly from the customer.

There is a potential concern with storing data with a third party, but we encrypt the files and Amazon has decent security on top of that, so it seems ok. It's doesn't seem any worse than other hosting situations.

Overall, we're pretty happy with this setup.

No comments: