Recently, I’ve been looking at using Amazon Glacier for some of my off-site backup needs. I already have a small amount of my data being stored in s3, and the lower cost of Glacier ($0.004/GB-month) make it a possibility for backing up even more data.
Since I intend for these backups to be fairly long time, I wanted to do a little analysis, comparing it with something like LTO-6. A LTO-6 tape drive looks to cost about $3000. There’s probably a little bit more cost involved for the needed SAS controller to be able to use the drive, but I’ll just use the $3k for my analysis here. Tapes (which hold 2.5TB, I’ll assume no compression, since most of my backup solutions will use already compressed data), run about $25, which amounts to $0.01/GB.
Doing a bit of math, I can conclude that the break-even point for storing some amount of data in glacier (continuously) vs tape is:
|5 year||13 TB|
|10 year||6 TB|
|15 year||4 TB|
Tape technology is likely to change within 5 years, and definitely within 15 years. It is also likely that in 5-10 years there will be other solutions than Amazon/Glacier.
I currently have multiple TB of data, but nowhere close to 13TB. Even if I considered backing up things like videos I record, it seems better for me to use Glacier.
Fortunately, Amazon seems to have fixed the absurd restore pricing that Glacier used to have (which was based on the rate you restored at, and could be absurdley expensive, costing hundreds or even thousands of times their normal data transfer rates). Now, a glacier restore, including the download, runs about $0.10/GB. A full restore of 13TB this way (which would also exceed Comcast’s data cap) would cost $1300, which is a lot, but still reasonable, and not the cost-of-a-house rate it would be with the old pricing.