The CCP1 GUI Project - Job Submission
Introduction
The Job Submission tools appear in the Job Tab of the
Calculation Editors, which generally appears as the right-most of the
Tabs. Clicking on the button labelled Job Submission, will
bring up a list of the supported job submission types, allowing one to
be selected. This is shown below for the GAMESS-UK job tab.
The different items in the list are the different job submission
types that are supported. Pushing the Configure... button when a
particular job submission type is selected will then bring up a job
editor that can be used to configure the job in question.
The different types currently supported are listed below, together
with links that point at more detailed documentaion for that job type.
- csevig6.dl.ac.uk: this the name
of the machine that the CCP1GUI is running on and will be different
depending on the name of the machine being used. This job type is
default setting and causes the job to be run on the local
machine.
- SSH: Run the job on a remote machine via ssh.
- RMCS: Run the job remotely using the My Condor Submit
technology developed by the eMinerals project
- Nordugrid: Run the job on remote machine that is running
the Nordugrid Arc software stack.
- Globus: Run the job on remote machine that is running
the Globus software stack.
Running Locally
When running locally, the job editor displayed below is used.
The Executable tools can be used to select the location of the
binary that is to be run, and the Local Directory tool can be
used to select the local directory that the job should run
in. In both cases the path can be typed in directly or the
Browse: tool can be used to browse to the file/directory.
If the Accept button is clicked, then the job editor will close and
the selected values will be used for all jobs in this session.
If the Save As Defaults button is clicked, then the selected values
will be saved by the CCP1GUI and will be used as the default settings
for any new jobs when the CCP1GUI is next restarted.
If the Cancel button is clicked, then the job editor will
close but none of the changes made will be saved.
Nordugrid
Globus
Globus can be used to run jobs on remote machines that have the
Globus middleware stack running on them. The CCP1GUI relies on the
Globus client libraries to manage the submission and movement of data,
and although the CCP1GUI only uses a small section of the Globus
client library, it can still be rather awkward to get this installed
on the Client machine. For when the CCP1GUI is runing on Linux/Mac
OSX systems, the GROWL toolkit
is the recommended way to install the relevant Globus libraries on the
system.
When running via Globus, the job editor displayed below is used.
Selecting the machine to run on
The machines section of the editor is used to select the machine to
run the job on. To add a machine to the list of available ones, type
the name of the machine in the box to the right of the list and click
Add. To delete a machine from the list of available ones,
select the machine from the list with the mouse and then click the
Del button.
The machines that are highlighted are the ones that have been
selected from the list. For Globus jobs, one, and only one machine
should be selected as the target machine.
If a machine is selected and the Save As Defaults button at
the bottom of the editor is clicked, all the values (number of
processors, directory path etc.) are saved for that machine. Clicking
the Update parameters from host when a machine is selected will
fill out the different widgets with any values that were saved for
that particular machine.
Setting up the job on the remote machine
The remaining widgets set up various attributes of the job on the
remote machine as described below.
| Number of processors |
The number of processors to execute the job on. This sets the
count RSL value. If count > 1, then the jobtype RSL
is set to "mpi", otherwise it is set to "single".
|
|
| Executable Name |
This is the path to executable on the remote machine. If the path
begins with a slash (/), then the path is expected to be a full
path. If the path starts without a slash then the path is expected to
be relative to the user's home directory on the machine
|
|
| Local Directory |
The directory on the local machine (i.e. the one where the CCP1GUI is
being run from) where the CCP1GUI should write out the input files and
copy back the data from the remote machine.
|
|
| Remote Directory |
This is the path to directory on the remote machine where the job will
be run. If the path begins with a slash (/), then the path is expected
to be a full path. If the path starts without a slash then the path is
expected to be relative to the user's home directory on the machine
|
|
| Globus Jobmanager |
Unfortunately, when running with Globus, one needs to specify the
jobmanager that the job is to be submitted to. By default Globus
submits to to jobmanager-fork, which just forks a new process
on the remote machine. For most parallel clusters , you will probably want to submit
your job to the queuing system on the machine (e.g. LSF, Loadleveller, Sun
GridEngine, etc). In these cases you should selected
jobmanger-sge (for GridEngine) or jobmanager-lsf (for
LSF), etc.
|
|
Setting additional job parameters with RSL
In addition to the tools described above, it is possible to further
fine-tune the job on the remote machine by setting various RSL
Parameters. RSL is the Globus Resource Specification Languge,
which specifies how the job should be executed on the remote
resource. You can read all about it here.
The tools here allow a user to add supported RSL parameters to the job
description. This can be used to set environment variables on the
remote machine, specify the maximum runtime etc. To add a new RSL
parameter, click the Add button to bring up a list of supported
RSL parameters. To delete an RSL parameter, select the one in question
from the menu on the far left and then click the Del button on
the far right.
Return to the CCP1GUI homepage
