How much can the network move while a shipment is in transit?
The fair comparison is not “network vs. courier” — it is how much data each moves in the same elapsed time. This table assumes a well-engineered path running at about 90% of the link.
| Link | 1 day | 3 days | 1 week | 2 weeks |
|---|---|---|---|---|
| 1 Gbps | 10 TB | 29 TB | 68 TB | 136 TB |
| 10 Gbps | 97 TB | 292 TB | 680 TB | 1.4 PB |
| 25 Gbps | 243 TB | 729 TB | 1.7 PB | 3.4 PB |
| 100 Gbps | 972 TB | 2.9 PB | 6.8 PB | 13.6 PB |
| 400 Gbps | 3.9 PB | 11.7 PB | 27.2 PB | 54.4 PB |
Formula: TB per day = link Gbps × 0.9 × 0.45 TB per hour per Gbps × 24. These figures need a path that actually reaches line rate — if your current tools realize a fraction of the link, scale them down accordingly, and see why transfers run slower than the link.
The hidden steps of shipping
Transit is only part of a shipment’s timeline. Door to door, you also wait for:
- Getting the devices. Ordering, provisioning, and delivery of drives or transfer devices before any data moves.
- Copying data on. Loading the devices is itself a data-movement job, limited by the same storage and host throughput that slows network transfers.
- Packing, transit, and receiving. Courier time, customs for cross-border moves, and handling at the destination.
- Copying data off and verifying it. Ingest into the target storage or cloud, then checksumming that every file arrived intact.
- Catching up on changes. Data written at the source after the copy started still has to be synced — usually over the network anyway.
Add these up honestly before comparing. A move that looks like “two days by courier” is often one to several weeks door to door.
When does shipping still make sense?
- There is no usable network at the source — a vessel, a remote field site, or a disconnected location.
- The available link is very slow relative to the data, and upgrading it is not an option.
- It is a one-time move of static data, with no deadline pressure and no ongoing flow afterward.
When does the network win?
- You have a fast link. At 10 Gbps and above running at line rate, the network usually beats door-to-door shipping for anything that is not enormous.
- Data keeps arriving. Instruments, sequencers, training pipelines, and sensors produce data continuously; shipping cannot keep up with a flow.
- Data changes during the move. The network can sync changes up to cutover; a shipment is out of date the day it leaves.
- Custody matters. Over the network, data never leaves your control, and every transfer can be encrypted and checksummed end to end.
- It is recurring or multi-site. Replication, DR, and hybrid-cloud flows happen every day, not once.
A hybrid that often works best
For a very large one-time seed over a slow link, ship the bulk and then use the network to sync everything that changed since the copy — and to carry the ongoing flow from then on. The faster the network path, the smaller the share worth shipping.
A five-minute decision checklist
- How much data, and is it static or still growing?
- What link speed do you have — and what share of it do your tools actually achieve today?
- What is the realistic door-to-door shipping time, including copy-on, copy-off, and verification?
- Is this a one-time move or the first of many?
- What are your custody, encryption, and audit requirements?
Then compare the shipping timeline with the table above. Try your own figures in the transfer-time calculator.
What line rate looks like in practice
With SLAC National Accelerator Laboratory and ESnet (U.S. DOE), Zettar moved 1 PB in 29 hours over a 5,000-mile path at 96% link utilization, encrypted and checksummed — while capped at 80 Gbps. On a full 100 Gbps link that is a petabyte a day: faster than most shipments, with the data never leaving your custody. The Zettar zx Appliance delivers that engineering turnkey, so the network option is available the day it is racked.
