Preservation
Why S3 is the reference
Amazon S3 defined object storage, and its interface became the de facto standard: most competing providers offer an “S3-compatible” API. Choosing a tool that speaks S3 therefore keeps the option of changing provider without changing your workflow.
For a photographic collection, S3 provides what a disk cannot: a copy sharing no physical risk with your studio, replicated by the provider, available without you maintaining anything.
Storage classes, and the trap
S3 offers several classes, from standard storage to deep archive classes such as Glacier. The logic is simple: the less you expect to read, the less you pay to store — and the more you pay to read.
The trap is retrieval time. A deep archive class can take hours before a file becomes retrievable. For a frozen archive you’ll only open after a disaster, that’s an excellent trade. For a collection you consult, it’s a bad choice you discover at the worst moment.
Practical rule: deep archive for frozen originals, standard class for anything you might want to read within the year.
The cost people forget to calculate
Storage is cheap. What costs is egress: retrieving your data bills a per-gigabyte transfer. A full restore of several terabytes can be a serious sum, and that’s precisely the moment you have no choice.
Before committing, estimate the cost of a total restore. If the figure surprises you, better to learn it now.
The settings that actually matter
- Versioning, to enable. Without it, a deletion propagated by a sync tool is final.
- Encryption at rest, available server-side — and, if your collection is sensitive, client-side encryption before upload.
- A dedicated least-privilege user: write to one bucket, no delete permission. If the machine is compromised, the archive isn’t.
- Lifecycle rules, to move objects automatically to a cheaper class after a delay.
- The region, chosen far from you for independence, and compatible with your personal-data obligations.
In practice
- Create a dedicated user with no delete permission.
- Enable versioning before the first upload.
- Choose the class by acceptable retrieval time, not by headline price.
- Estimate the cost of a full restore.
- Verify integrity from the bucket after upload, not only locally.
Obscura Vault sends its archives to Amazon S3 and compatible storage, resumes an interrupted upload where it stopped, and checks the copy at the destination once the upload completes: manifest, signature, size and, on Amazon S3, the content’s SHA-256 checksum. It does not offer the Glacier Flexible Retrieval and Deep Archive classes, whose objects cannot be read back until thawed.