MGRID Portal How-To

Submit a Simple Interactive Job

Job Submission

To run a job, follow the steps below to make the screen look like the one above.

Go to the MGRID Portal
Click MGRID Job Submission
Select Cluster name "MacOSX_cluster"
Enter "/bin/date"
Check "Run Interactive?"
Click "Run Job"

The following job output should appear, giving the current date and time:

Job Submission

More Information on fields

[ Job name ]

This allows users to give a unique name to the job. This can be useful for organization when submitting multiple jobs. This field is only required when the "Run Interactive?" field is not checked.

[ Cluster name ]

Specifies on which host the job is to run. The "MacOSX_cluster" cluster was chosen for this demonstration because a wide variety of users are permitted to submit jobs.

[ Queue ]

When submitting to a cluster controlled by a job manager such as PBS (Portable Batch System), Condor, or SGE (Sun Grid Engine) it may be necessary to specify the queue to which the job will be submitted. Administrators use the different queues to more finely control access to clusters by limiting specific users to given queues. They may also control load by giving higher priority to some queues over others.

[ Executable ]

The full path of the binary or script to be executed. In the example, "/bin/date" prints out the current date and time.

[ Arguments ]

Command line arguments to the executable. As an alternate example to the one above, enter "/bin/uname" for the executable, and "-a" for "Arguments". The output should show name of the head node.

[ Run Interactive ]

If checked, the browser will wait for the job to finish and then send the output to the screen. If not checked, then the job will be submitted and the browser will return immediately with an indication as to whether the submission was successful or not. If it was successful, then a job number something like:

will be shown. This can be used later to retrieve the output.

[ Email ]

Some job schedulers will send an email notification when your job finishes, so you may enter your email address here. Note also that some job managers (currently those running on morpheus and hypnos) require an email address.

[ Walltime ]

This is the maximum amount of time that your job will be given to run. If not specified, most systems will assume a default such as 15 minutes. In general, jobs with smaller times will run sooner.

[ Host Count ]

When running on a cluster, this specifies the number of nodes on which to run. Note that many nodes have two processors. In general, requesting more nodes will run later.

[ Standard Output File ]

File into which standard output from the Executable will be placed. This must be specified if "Run Interactive?" is not checked.

[ Standard Error File ]

File into which standard error from the Executable will be placed. This must be specified if "Run Interactive?" is not checked.

[ Run Job ]

Tells the system to run the job with the given parameters.

[ Expert Mode ]

Allows a sophisticated user to modify the Globus command before it is submitted. The Globus commands have a number of options that are not frequently used, however are occasionally useful. For full details, refer to the Globus documentation.

[ Monitor Submitted Jobs ]

This allows you to review the jobs you have submitted and view their status. Refer to Monitoring Jobs for more information.

MGRID How-to

Grid-Enable Your Browser
Submit a Simple Interactive Job
Schedule a Non-Interactive Job
Monitor Your Non-Interactive Jobs
Change Authorization
When Your Certificate Will Expire
How Much Resources People Use
Upload and Download Your Files Into and Out of MGRID
What Computing Clusters are Available from MGRID
What Technology MGRID Uses
Typical MGRID User Steps