3 Local resource management systems
The switch -g runs jobs as a so-called Fork job, which means the job is to be run immediately, on the machine which runs the WS-GRAM service without going to the compute nodes via the batch queueing system.
Using Fork is recommended to run only those jobs which are needed to run on the front-end/login node. Since fork jobs are executed immediately, this method is well suited to debugging job scripts quickly, thus avoiding long batch queue waiting times. No production jobs may be submitted to the Fork scheduler! NB it is the default method, so -Ft Fork can also be omitted.
Production jobs have to be submitted to the batch scheduling system of the target site. That is done using the deisa_service script with the -b switch. Recall that this script is available at DEISA machines and it will automatically set up the correct batch system identifier. The following example performs a job submission on a DEISA machine to Huygens at SARA:
globusrun-ws -submit -F `deisa_service -i -b sara` -c /bin/date