domains · fusion
diagnostics and simulations, one archive
A fusion experiment produces hundreds of diagnostic channels per shot, and the simulation campaigns that interpret them are larger still. Earthmover stores both as versioned tensor data, so measurement and model sit side by side.
the data
What we store
Magnetics and equilibria
Flux loops, Mirnov coils, and the reconstructed equilibria derived from them, at the rates control and analysis actually need.
Profile and imaging diagnostics
Thomson scattering, interferometry, spectroscopy, bolometry, and fast-camera data, each keeping its own time base and calibration.
Gyrokinetic and MHD simulation
Turbulence and stability output from codes like GENE, CGYRO, and NIMROD, which routinely dwarfs the experiment that motivated it.
Integrated and transport modeling
Whole-device scenario and transport runs, stored against the shots and equilibria they were initialized from.
on the platform
What Earthmover adds
-
A campaign, not a shot — thousands of shots as one logical dataset, so cross-shot analysis does not mean opening thousands of files.
-
Measurement beside model — diagnostics and simulation output live in one catalog, under one set of permissions.
-
Append as the run proceeds — each shot commits transactionally, so downstream analysis always reads a complete state.
-
Reproducible training sets — pin a disruption-prediction or surrogate-model dataset to an immutable snapshot.