Backups with Arq and Amazon Glacier
07 Jan 2013It's good practice to have a few different levels of backups: your local backup (for me, that's via Apple Time Machine; for you, that could just be the backup you take every few weeks—or months—to your USB drive); an offsite backup (if you have two USB drives, you might leave one at the office and rotate your backup drives); and an online backup (something like CrashPlan or Carbonite, where some software uploads files in the background to "the cloud").
The idea behind this is that, if something really bad happened and you needed a backup but it was destroyed, you'd have your offsite backup; and if something really really bad happened, you'd have your online backup. You'd always be able to get back to something reasonably recent.
Until recently, I've not quite followed those best practices. I do have my daily backup (Time Machine), and a weekly cloned backup, but I didn't have an online backup, for a couple of reasons. First, they can get pricy (generally between $50 and $100 annually, though you might argue that's a small price to pay for security). I always justified it that a) I had really important stuff on Dropbox, and b) other important stuff (pictures, music, etc.) were synced to my other desktop computer and synced to Apple's servers (iTunes Match).
A few months ago, Amazon announced Amazon Glacier, which is their super high reliability storage, but at a really low price because it's slooooooow (get it, Glacier). You can store about 100GB of data for $1 or so per month.
But there's a catch. The idea is that you should not need to download the data very often, so pulling the data back down is where it gets expensive.
I figured, though, that this would be worth it. The Glacier backup is my super emergency backup. I have multiple other copies of my data, if something really really bad happens, I'd be willing to pay the price to get my data back
Paired with a really nice piece of software called Arq, which manages the uploads to Amazon, the encryption, and the tracking of which files need to be updated/uploaded to Amazon, I set out to get myself a nice online backup.
It took almost a week to get the whole thing uploaded (I was backing up somewhere north of 100GB of data). Once uploaded, it takes Arq about 10 minutes each day to figure out what it needs to update.
Total cost to me? $1.75/month.
Let's assume I end up averaging about $2/month. In year one, that's $30 for Arq, and $24 for Amazon storage, for a total of $54. That's almost exactly what an equivalent CrashPlan or Carbonite account would cost. But in year two, when it only costs me $24, it'll be less than half of what those accounts cost.
Now, really all I'm doing is hedging my bets. I'm assuming that I won't need to restore my data any time soon, and I'm banking that money. If you think you're going to need to restore files all the time, CrashPlan or Carbonite might be better for you. But if you think you won't need to restore very often at all, then this solution might be the right option for you.
If I do have to restore… that's when things get interesting. Restoring 120GB of data over 4 hours from Amazon will cost about $230. Over 12 hours, about $90. Over 24 hours, about $50.
So, if I needed to restore my entire computer from my Glacier backup, I could do it for $50 if I was willing to let it run for a day (and for $30 if I was willing to let it run for two days).
That seems like a reasonable risk to me. In most cases, if I lost some data, I would be able to go to a number of local backups to get it (cost to me: zero). In the event I lost stuff catastrophically, paying between $50 and $200 to get it back seems entirely reasonable. Your particular use case/risk tolerance may change that equation for you.