Common data sets are accessible via
It's better to always use the
/bdd absolute path in your code because we could move dataset in underlaying filesystems
Some data sets are in a protected mode on the cluster (ecmwf , cmip ...). you have to ask access to support
Some dataset are too big to be hosted in our dacenters : (cmip ...). They are remotely accessed from national french research computing center. Access could be disabled when national computing center are in maintenance, have infrastructure or network problems. In thoses cases, we can not use our 'magic' on the remote data sets.