Server configuration file

The programmer of a batch service available in a SeD do not have to worry about which batch system to submit to, except for its name, because DIET provides all the mechanisms to transparently submit the job to them.

DIET is currently able to submit batch scripts to OAR (version 1.6 and 2.0), PBS/Torque, LoadLeveler, and SGE. The name of the batch scheduler managing the parallel resource where the SeD is running has to be incorporated with the keyword batchName in the server configuration file. This is how the SeD knows the correct way of submitting a job.

Furthermore, if there is no default queue, the DIET deployer must also provide the queue on which jobs have to be submitted, with the keyword batchQueue.

You also have to provide a directory where the SeD can read and write data on the parallel resource. Please note that this directory is used by DIET to store the new built script that is submitted to the batch scheduler. In consequence, because certain batch schedulers (like OAR) need the script to be available on all resources, this directory might be on NFS (remember that DIET cannot replicate the script on all resources before submission because of access rights). Note that concerning OAR (v1.6), in order to use the CoRI_batch features for 0AR 1.6 (see Section 10.2.3), the Batch SeD deployer must also provide the keyword internQueue with the corresponding name. For example, the server configuration file can contain the following lines: