The record-holder for moving research data.
Zettar zx set a U.S. national lab's petabyte record — a full petabyte across 5,000 miles in 29 hours, at 96% utilization, encrypted and checksummed. Get instrument and simulation data where it belongs for HPC centers, universities, and national labs — about 10× typical commercial and freeware movers — turnkey, pretuned, and racked beside your storage.
Instruments produce data faster than you can move it.
- Light sources, telescopes, microscopes, and simulations flood storage faster than you can drain it.
- rsync, scp, and Globus weren't built to fill the high-speed R&E links you pay for.
- Your DTNs sit idle — the hardware can do line rate, but the software can't saturate it.
- Multi-institution science means moving the same petabytes between facilities, again and again.
Built in demanding DOE science environments.
SLAC / ESnet / DOE
Engaged since 2015 on LCLS-II, a U.S. DOE exascale-class project. Winner of the Supercomputing Asia 2019 Data Mover Challenge. The proof is on real networks, not a slide.
~10× faster
About 10× typical commercial and freeware movers — line rate, scaling out with no software ceiling. Finally use the network you already pay for.
Drops into your Science DMZ
A pretuned appliance on standard x86 and Linux, with CUBIC as the default congestion control. Rack it beside your DTNs and storage — nothing to tune.
Facility to facility
Move petabytes between institutions and to the cloud. Latency-insensitive — validated over real 5,000- and 12,375-mile transfers, including Poland to Singapore.
ESnet's in-depth evaluation of zx on its own data transfer nodes confirmed 2×–7× the performance of Globus GridFTP.
Backed by the record: 1 PB in 29 hours at 96% link utilization with SLAC & ESnet (U.S. DOE). Read the record →


"The flexible traffic engineering design of ESnet's network ensures that TCP-based transfer protocols can be used effectively for high-throughput data movement."
The hardware behind the record.
The SLAC / ESnet / Zettar system that moved a petabyte in 29 hours over a 5,000-mile loop — sending and receiving zx clusters, NVMe burst buffers, and ESnet's on-demand circuit.
Put a number on it — and make your case.
Beam time and grant clocks don’t wait. When data can’t keep up with the instrument, the science stalls — and allocated time can’t be bought back.
Common questions
How fast can Zettar move research and instrument data?
At line rate — about 90%+ of the bandwidth your path can sustain, scaling out with no software ceiling. With SLAC and ESnet it moved 1 PB in 29 hours at 96% link utilization over a 5,000-mile path.
How does it compare with Globus GridFTP on our DTNs?
ESnet's in-depth evaluation of zx on its own data transfer nodes found 2×–7× the performance of Globus GridFTP. Zettar ships that capability as a turnkey, pretuned appliance that racks beside your DTNs and storage.
Does it fit a Science DMZ?
Yes — it is a pretuned appliance on Linux that sits at the network edge beside your DTNs and storage, uses CUBIC by default, and holds throughput on long, high-latency R&E paths, validated over 5,000- and 12,375-mile transfers.
Use the network you already pay for.
See the zx Appliance move research data at line rate — ESnet measured 2×–7× Globus GridFTP.
