The writing of a service corresponding to a parallel or batch job is very simple. The SeD programmer builds a shell script that he would have normally used to execute the job, i.e., a script that must take care of data replication and executable invocation depending on the site.
In order for the service to be system independent, the SeD API provides some meta-variables which can be used in the script.
$DIET_NAME_FRONTALE: frontale name
$DIET_USER_NBPROCS: number of processors
$DIET_BATCH_NODESLIST: list of reserved nodes
$DIET_BATCH_NBNODES: number of reserved nodes
$DIET_BATCH_NODESFILE: name of the file containing the
identity of the reserved nodes
$DIET_BATCH_JOBID: batch job ID
$DIET_BATCHNAME: name of the batch system
Once the script written in a string, it is given as an argument to the
following function: