Dropdock takes multi-terabyte deliveries from cameras, render farms and partners, verifies them on arrival, and hands them to whatever runs next. Built for the transfers where a browser upload gives up at 4 %.
Dropdock is deliberately not a workspace, a player or a review tool. It is the part of the pipeline where bytes land intact and something else picks them up.
Objects are split into parts and pushed on up to 32 streams. A dropped link resumes from the last acknowledged part, not from zero — a 40-hour delivery survives a router reboot.
Each part carries a BLAKE3 digest computed by the sender and re-computed on write. A mismatch fails the part immediately instead of surfacing three weeks later in a grade.
A partner or a DIT on location gets a signed upload URL with a prefix, a size cap and an expiry. They drag a folder into a browser; you see it arrive under the right project.
A webhook fires per object or per completed manifest. Kick off transcoding, sync to your NAS, or file a ticket — Dropdock does not care what runs next.
Move a prefix to cold storage after 30 days, delete proxies after 90, keep masters forever. Rules are declarative and dry-runnable before they touch anything.
Four European docks with 40 Gbit/s uplinks. Data can stay in the dock it landed in, or replicate to a second one on a schedule you set.
Point an existing tool at the endpoint and it works. When you need more than S3 gives you, the native API adds manifests, signed upload links and lifecycle rules.
# one-off endpoint, no config file rclone copy /Volumes/A001 \ :s3:ingest/prod-zk/day12 \ --s3-provider Other \ --s3-endpoint https://fra.dropdock.pro \ --transfers 16 --checksum
POST /v1/upload-links Authorization: Bearer $DD_TOKEN { "prefix": "prod-zk/vendor-vfx/", "max_bytes": 2199023255552, "expires_in": "72h", "webhook": "https://pipeline/dd" }
Uploading is always free — you should never hesitate before sending us data. You pay for what sits on disk and what leaves.
Not a practical one. Multipart uploads go up to 100,000 parts and a part can be 5 GB, so a single object can exceed 400 TB. The largest thing anyone has stored with us so far is a 61 TB EXR sequence delivered as one manifest.
With 16 parallel streams from a well-connected European host we routinely see 6–8 Gbit/s to the nearest dock. A single stream tops out around 1.2 Gbit/s because of TCP round-trip behaviour, which is why the client parallelises by default.
No. Ingest is free on every plan, including parts that are retried. You are billed for stored bytes measured hourly and for egress, nothing else.
Objects stay listed and their metadata stays hot. The first byte of an archived object arrives after two to four minutes while it is staged back; after that it reads at full speed for 24 hours. There is no per-request retrieval fee.
Yes. Replication is opt-in per bucket, so a bucket created in Helsinki stays in Helsinki until you say otherwise. The console shows the physical location of every prefix.
Accounts start empty and free. Ingest costs nothing, so the first real test can be an actual delivery rather than a demo file.