FDB documentation#
The FDB is part of ECMWF’s high‑performance data infrastructure and stores different file formats, such as GRIB, ODB, BUFR. In case of GRIB, it stores each message as a field, indexes it by meteorological metadata (e.g., parameter, level, date/time), and serves recent outputs to post‑processing tasks and users. In operational use, FDB acts as a hot cache in front of the long‑term MARS archive, enabling fast access to newly generated data.
PyFDB is the Python interface to the FDB, a domain‑specific object store designed to efficiently archive, index, list, and retrieve GRIB fields produced by numerical weather prediction workflows. It provides a thin, idiomatic Python layer over the FDB client library installed on your system, so you can drive FDB operations directly from Python scripts and notebooks.
Z3FDB is the Python-Zarr interface to the FDB. It provides a thin, idiomatic Python layer over the FDB client library installed on your system, so you can extract Zarr data from underlying GRIB data by creating a virtual Zarr store. The view is described via a MARS request and can be used to create a virtual Zarr store containing the data of the MARS request. For further information, see Z3FDB or visit the Zarr project.