I'm wrapping up production on a movie and need to archive roughly 80 TB of footage. Most files are large video files, with hundreds of files ranging from 10–40 GB each. I'm trying to follow the 3-2-1 backup rule and am considering AWS S3 Glacier Deep Archive as the off-site copy.
Upload bandwidth isn't a concern because I have effectively unlimited high-speed upstream capacity. The plan is to store the footage for at least a year. In a disaster, I might need to restore around 5 TB, although ideally I would never need to retrieve anything.
My rough estimate is $1,000–$1,500 per project per year, including storage and a possible 5 TB restore. Does that seem realistic? Are there hidden costs beyond the 180-day minimum storage period, such as retrieval, egress, API requests, or lifecycle fees? I'd also like to understand the real-world advantages and drawbacks, especially recovery times and whether another archival solution would make more sense for professional video production.
4 Answers
This is almost exactly what Deep Archive is designed for: large, rarely accessed files that serve as disaster insurance. Large video objects avoid the small-file overhead, and the storage cost is far lower than maintaining another large local system. The main question is whether you can accept waiting many hours, potentially up to a couple of days, before the files become downloadable.
Your use case sounds appropriate, provided this is genuinely a cold off-site copy and not the only place you can access the footage. Keep at least one usable local copy and periodically verify that the archive inventory and metadata are intact. LTO or another storage provider could also work, but they bring their own costs, maintenance, hardware-obsolescence, and data-integrity concerns. For a one-year, rarely restored disaster copy, Deep Archive is a reasonable choice if the recovery delay and egress bill are acceptable.
Be much more careful with the restore budget than the storage estimate. Deep Archive has no instant or expedited retrieval option. Bulk retrieval is cheaper but slower, while Standard retrieval is faster and more expensive; either way, you still pay internet egress after the objects are restored. A 5 TB recovery could cost several hundred dollars once retrieval and egress are included, and restoring the entire 80 TB could cost several thousand dollars. Also remember that deleting or replacing objects before the 180-day minimum still incurs the remaining minimum-storage charges.
Upload directly using the Deep Archive storage class if your software supports it; you don’t have to upload into ordinary S3 storage first and then transition the files. For hundreds of large objects, use multipart uploads, verify checksums, and keep your own manifest or catalog of every object. Batch Operations can help submit restore requests later. Test the upload and restore process with a small sample before committing the entire archive.
That recovery window is fine for my use case. This would be a last-resort archive rather than something I need to edit from regularly.