Why a browser download or a sync client is the wrong tool
Pulling 10 TB out of object storage through a browser, a cloud console or a sync client built for documents means a TCP stream per object, throttled by distance to the region, a restart from the top when a laptop sleeps, and no record that what landed matches what was stored. Every team hits it eventually: an ML team pulling a training set to a GPU box, a lab restoring an archive to an instrument PC, a geospatial firm bringing imagery on-prem, an IT team seeding a backup appliance.
Bucket to on-prem in two steps
Step one: import. Connect the bucket and choose what to pull; the import runs on our servers into a workspace folder, not through your laptop. Any S3-compatible bucket works, on every plan, and the workspace shows each file with its size and the checksum recorded on arrival.
Step two: pull it on-prem at line rate. Put a download watch folder on the server, in the desktop app or the headless Linux agent under systemd, pointed at that workspace folder; everything that lands appears on the NAS within a minute, sub-folders included, and nothing is ever deleted locally. Or script it with boita download from the CLI. The transfer runs through the accelerated transfer engine over UDP, in parallel sessions, resuming after any interruption.
Throughput is the lower of the plan ceiling and the line. 10 TB is 80,000,000 megabits. On Studio, fifty sessions at 45 Mbps allow 2.25 Gbps, so over a 1 Gbps line the pull is line-limited: about 22 hours, unattended. On Enterprise with 100 sessions over a 10 Gbps line the 4.5 Gbps plan ceiling takes over: about 5 hours. A 10 TB set just fits Studio’s 10 TB of storage; larger sets run in batches or on Enterprise, where storage is custom.
On-prem to bucket: the reverse
Going the other way, an upload watch folder on the server, or boita upload from a script, sends new files to a workspace folder as they finish; allowed hours and a workspace bandwidth window, enforced by the transfer server, keep the day’s connection free. Enterprise pairs the folder with your own S3-compatible bucket or transfer server and pushes every file that lands within minutes, retried on failure, with a delivery log. On other plans, share the folder as a link or let the far end pull it with a download watch folder.
- Import from any S3-compatible bucket, on every plan, run on our servers
- Download watch folder, Linux agent or CLI pulls to the NAS at line rate with resume
- Upload watch folder or CLI sends back; auto-delivery to your own bucket on Enterprise
- Checksum on every file; allowed hours and bandwidth windows enforced by the server
Other clouds, and what is not offered yet
Google Cloud Storage works today through its S3-compatible endpoint; native GCS and Azure Blob Storage connectors are coming. An SFTP endpoint is coming; Boita does not offer one today. The agent as a Docker image and as packages for the common NAS platforms is coming; today it installs on any Linux server with a shell.
What it costs, and what it does not change
Boita adds no per-GB fee of its own, in either direction, and no charge for downloads. Egress from a bucket is your cloud provider’s charge and is unchanged by Boita; what changes is the hours, the retries and the person watching them. Plans are flat: Pro ₹499 with 1 TB, Studio ₹7,999 with 10 TB, Enterprise custom.