How Abed works

Running the tasks on the cluster is done as follows:

1. In the PBS the following preprocessing steps are taken:

   1. A results directory is created on the scratch filesystem (see above)

   2. An email is sent to the current user with a brief summary of the number of tasks to be performed. This is sent to the users email, an may require a forwarding file to be delivered correctly.

   3. Optional modules are loaded

   4. Optional environment variables are exported

   5. mpicopy is used to copy files from the home directory to the scratch filesystem

2. abed run is executed through mpiexec. This requires Abed to be in the PATH variable. This command is run in a timeout command, with the execution time reduced by a configurable number of seconds. This should allow enough time for the post-processing steps

3. The postprocessing steps are:

   1. Create a bzips directory in the current release directory

   2. Create a compressed archive for each dataset in the result directory, this is done in parallel through pbzip2

   3. Copy the compressed files to the current release directory

   4. Sent an email to the user that the task is completed.