After an MPI program is debugged and ready to run in a production environment, it is often useful to submit it to a queue to be scheduled for execution. The Network Queuing Environment (NQE) provides this capability. NQE selects a node appropriate for the resources that an MPI job needs, routes the job to a node, and schedules it to run.
This chapter explains how to use the NQE graphical interface on IRIX systems to submit an MPI program for execution. For information on using NQE to submit UNICOS or UNICOS/mk programs, see the NQE User's Guide.
Before you begin, set your DISPLAY variable so that the NQE screens appear on your workstation. Then enter the nqe command, as shown in the following example:
setenv DISPLAY myworkstation:0 <nqe |
Figure 7-1 shows the NQE button bar, which appears after your entry.
To submit a job, click the Submit button on the NQE Job Submission window. Figure 7-2 shows the NQE Job Submission window with a sample job script ready to be submitted.
Notice in this figure that the difference between an NQE job request and a shell script lies in the use of the #QSUB identifiers. In this example, the directive #QSUB -A nqearray tells NQE to run this job under the nqearray project account. The directive #QSUB -a 8:05pm tells NQE to wait until 8:05 p.m. to start the job.
Also notice in Figure 7-2 that the MPI program is already compiled and distributed to the proper hosts. The file array/hostlist has the list of parameters for this job, as you can see in the output from the following cat command:
% cat array/hostlist homegrown, disarray, dataarray |
To see the status of jobs running under NQE, click the Status button to display the NQE Status window.
Figure 7-3 shows an example of the NQE Status window. Notice in this figure that the MPI job is queued and waiting to run.
To verify the scheduled starting time for the job, position the mouse cursor on the line that shows the job and double-click it.
This displays the NQE Detailed Job Status window, shown in Figure 7-4. Notice that the job was created at 8:26 PDT and is to run at 20:05 PDT.
For more information on using NQE, see the following NQE publications:
Introducing NQE
NQE Release Overview
NQE Installation
NQE Administration
NQE User's Guide
The preceding publications are also available in the Cray Online Software Publications Library at the following URL:
http://www.cray.com/products/software/publications |
PostScript files of NQE publications are available through the Cray Online Publications Software Library. To download a publication, select Summary next to the book title you want on the Titles Web page, which is located at the following URL:
http://www.cray.com/products/software/publications/dynaweb/docs/titles.html#N |
For general information about NQE, see the following URL:
http://www.cray.com(search for NQE) |