2.5.1. Data Structure

The Python dispatch.snapshot(n,run) command caches the metadata needed to access the binary data, using the file and directory structure described below:

Data from execution of experiments ends up under the subdirectory data/. It is often convenient to let data/ be a softlink to a directory on a dedicated disk system (e.g. a Luster file system on a cluster, or an external disk drive on a workstation or laptop). When using the default input file (input.nml) data is stored directly in data/, while if using e.g. run.nml, the data is stored in data/run/.

Metadata describing the snapshots is stored as namelists in files such as

data/run/params.nml
data/run/nnnnn/snapshot.nml
data/run/nnnnn/snapshot_rrrrr.nml

where nnnnn is the snapshot number and rrrrr is rank (both with 5 digits). The binary data are written to

data/run/nnnnn/snapshot_rrrrr.dat