Cold Storage & Archive

Most files in a long-lived system are read once and then sit untouched for months: last quarter's database dumps, rotated logs, finished project deliverables, compliance records you must retain for seven years but hope never to open. Paying hot-storage rates to keep them online is waste; paying with multi-hour restore latency is pain. files.link's cold storage tier is built for exactly this gap.

It uses an instant-retrieval cold class, so archived files come back immediately through the same API call as any other file — there is no 'initiate restore, wait 3–12 hours, then download' dance that classic Amazon S3 Glacier (Flexible/Deep Archive) forces on you, and no juggling of Expedited/Standard/Bulk retrieval tiers where the fast option costs many times the slow one.

The trade you accept for the lower storage price is a 90-day minimum retention window: a file deleted before 90 days is still billed for the remainder of that window, so the tier rewards genuinely cold data, not files you churn weekly. The decision rule is simple — if a file is read less than roughly once a month, archive it; if it's actively served and benefits from edge caching, leave it in standard storage.

Migrating an existing archive is straightforward because the upload contract is identical to standard storage: POST file metadata to /v1/files/{folderId}, PUT the bytes to the returned presigned URL, then POST /v1/files/confirm-upload with the file id — you simply flag the storage tier. A nightly cron that sweeps your standard-tier files older than N days into cold storage is a few lines around those three calls.

Because delivery is folded into prepaid credits, there are no separate per-GB egress fees layered on top when you do retrieve — the bill-shock pattern of S3 (storage + requests + restore + egress as four meters) collapses into one balance you watch. That makes cold storage here a clean fit for database backup storage, Kubernetes volume snapshots, log archives, Terraform state history, and any compliance dataset where you need durability and a fast path back, but rarely walk it.

It is honestly not a fit for hot, frequently-read content — for that the standard tier with CDN caching is cheaper in practice once retrieval volume is counted.