Life of a typical computing job (1)

• A parallel program is compiled and linked on the master node

• A batch script is written, naming the program file, telling how many processors that shall run it and maxjobtime

• On the master node, the script is submitted to a batch queue. The batch system queues the job, giving it a unique job number

• The job waits, while the scheduler tries to find a time for the job to start, when the required number of compute nodes are free to use

• The required number of compute nodes are reserved for the job, and the job gets a list of their names