1
I'm looking to use AWS Glacier for personal backup, or the similar 'Nearline storage' from Google. I would prefer manually uploading stuff (instead of an automated backup solution). I'm aiming to store pictures and video, about 110G currently. This is backup, so I hope to not need to retrieve it, therefor (reasonable) costs and time for retrieval are not an issue.
Because Glacier seems to cost slightly less in storage (0.007$ vs 0.01$), I looked at that mostly so far. Google Nearline seems to be only marginally more expensive though, and maybe somewhat simpler in use? I'd appreciate any experience on the difference between the 2 in practice.
For Glacier, I'm now wondering: should I look at using it directly or through S3? What are the consequences for both?
- Some sites seem to suggest getting a list of files in folders from glacier-without-s3 is difficult (or costly?)
- Docs on s3-with-glacier seem to suggest there will be some storage on s3 as well (to store the data about the file being in Glacier, if I understand correctly), it's not clear to me how much that will add to the cost.
- Some backup solutions like cloudberry seem to support only s3-with-glacier, making me wonder what is the reason behind that.
Is there anything else to consider, anything I've missed? Hidden costs, things that would make using this difficult, ...?
Update (because I don't think this is a full answer) after trying nearline and comparing to a friend trying glacier:
- glacier has reasonable tools on the aws site to calculate cost etc, however, even getting an index of your files takes 4 hours, which can be annoying. It has some (at this time) unique features, like a write-once setup so files won't be overwritten by accident
- nearline works like normal online storage, without the glacier 4 hour wait etc, so the only thing to be aware of seems to be the cost of early deletion (also when overwriting!) and the cost of downloading
- backblaze now has online storage without any of the nearline/glacier limits, below the cost of glacier. Very 1.0 though, so support in clients like cloudberry isn't all there yet.
Legolas
Posted 2016-06-19T20:39:17.027
Reputation: 389
If you need to do a recovery, how fast will you need the data back? Is it okay if you need to wait several hours before you can begin retrieving it? How long will you be storing any one backup? – Brandon Yarbrough – 2016-06-20T06:21:16.027
Yes, several hours/days even weeks to restore everything are fine. The idea is to have 1 backup that is never updated/deleted but only grows with new things. – Legolas – 2016-06-21T10:38:37.673