HN Explorer

Ask HN: Is AWS bandwidth pricing making backup storage infeasible?

3 points by t0mas88 ↗ HN
Maybe this is glaringly obvious to a lot of you, but AWS bandwidth pricing seems to be the biggest factor in S3 / Glacier pricing, much higher than the "retrieval fee" that everybody worries about.

Some background: I was considering using S3 Glacier Deep Archive as a solution to backing up large volumes of media files (mostly raw video/audio). For production/editing these files are kept on fast SSD based storage, but after that they're archived together with all the project files, notes etc. It's done on normal 4TB sized spinning disks and they're read maybe once a year when archive footage is needed for a new edit.

When expanding storage there was a discussion on whether backing up to AWS would be better than buying twice the amount of disks to do RAID 1. The initial idea was that non-RAID disks + AWS is better, because it's off-site for when the building burns down, and you can make make the permissions "append only" so it's much more resistant to ransomware or other issues.

The price of S3 Glacier Deep Archive is great at $ 0.00099 per GB per month, that's about two years of storing 4 TB at the price of a 4 TB drive. The much feared retrieval fees in case one of our local drives is lost (crashes/corruption happen) is $ 0.003 per GB so $ 12 to get everything back. So the fear of retrieval fees actually isn't bad at all with the S3 Glacier Deep Archive pricing if you accept having to wait 48 hours.

But then comes the shocker: You're going to pay 30 times as much as the retrieval fee to get that data out of AWS, because bandwidth is $ 0.09 per GB, so retrieving that "crashed disk" 4TB will cost $ 372 in total instead of just $ 12. Did I miss something? Or does AWS bandwidth pricing make using them as a backup a terrible idea if you ever want to be able to retrieve that backup?

6 comments

[ 2.7 ms ] story [ 28.5 ms ] thread
throwaway888abc
Get same conclusion for photos. Ended up with Backblaze and Restic.

https://www.backblaze.com/

https://restic.readthedocs.io/en/latest/010_introduction.htm...

Set and forget.

See how AWS S3, Azure, and Google Cloud compare to B2 Cloud Storage.

https://www.backblaze.com/b2/cloud-storage-pricing.html

t0mas88
But Backblaze is cheating in how they compare pricing, because they compare to a much more expensive S3 storage class. In the "4TB disk crash" case it's really not as their website describes:

AWS: $ 48/year to store, $ 312 to retrieve Backblaze: $ 240/year to store, $ 40 to retrieve

If I store it for more than 17 months before the crash, Backblaze is actually more expensive than AWS.

atYevP
Yev here -> the reason we have S3 pricing on our pricing page and not Deep Glacier is because our performance is closer-tied to the S3 service, meaning both Backblaze B2 and S3 are hot storage. Deep Glacier is...well..cool or cold storage in that it takes more time to access your data.

You're not wrong, just wanted to give a bit more context on the why!

t0mas88
Ok, makes sense. Time to launch a lower cost slow retrieval product? Because for most backups I expect there are a lot of customers that would trade retrieval time for a lower cost per GB.
QuinnyPig
The retrieval fee used to work differently, leading to multi-thousand dollar bill surprises. A lot of the negative reaction stems from old data.

AWS’s hilariously bad egress pricing is of course still nuts.