NQS (Network Queuing System) Brief Guide
In the IJIS system, 14 Solaris machines (Sun Fire V440) and 24 Linux machines (Sun Fire V20) are clustered by Sun Microsystems N1 Grid Engine 6.0. The N1 Grid Engine enables users to submit their jobs without searching for resources in each node. The jobs are executed through a Network Queuing System (NQS).
- Job submission through NQS
- "qsub" command
- Shell script creation
- Checking the status of submitted jobs
- Job cancellation
- Executable directory
- The "fast" queue is for jobs with runtimes of less than one hour
- The "large" queue is for a job with runtime of more than ten hours
- Further details about NQS
Job submission through NQS
NQS (Network Queuing System) controls batch processing on UNIX machines.
Master hosts receive program compilation and job submission. In order to submit your job to the NQS, you should log onto the master host and use the "qsub" command. See "Connections from the User Workstation to the user support servers" to log into master hosts.
"qsub" command
Job submission is specified by the qsub command and execution queue names, which are summarized in the following table. The jobs submitted to each queue are processed in order of registration. When the number of requests exceeds the number of simultaneously executable jobs (the number of slots), the new request is suspended until any of the present executions have completed.
The form of the qsub command
Shell script creation
Users should write a shell script to submit a job with the qsub command. Here is an example of execution script. The execute command is described in the shell script, and the script file name is specified as the argument of the qsub command.
Example
- Writing a shell script
$ vi example.sh
- Example of a shell script
# Batch request script example # comment
#!/bin/sh # describing the shell
f77 -o sample.exe sample.f # execution command
sample.exe # execution command
exit
- Job submission
$ qsub exsample.sh
Your job 4130 ("exsample.sh") has been submitted.
- Status of the submitted job
$ qstat
job-Id prior name user state submit/start at queue slots ja-task-ID ---------------------------------------------------------------------------------------------------------------------- 4130 0.00000 exsampl.sh test qw 01/01/2006 12:00:00 1
Checking the status of submitted jobs
This command allows for checking the status of submitted jobs (running, waiting, etc.).
Example
| job-Id | prior | name | user | state | submit/start at | queue | slots | ja-task-ID |
| -------------------------------------------------------------------------------------------------------------------------------------- | ||||||||
| 5437 | 0.00000 | test.sh | test | qw | 06/08/2006 16:51:48 | 1 | ||
| 5438 | 0.56000 | sample.sh | test | t | 06/08/2006 16:50:47 | fast@xxx.ijis.iarc.uaf. | 1 | |
Meaning of items
| job ID | ID number for the submitted job | ||||||||||||||||||||
| prior | Priority of the job | ||||||||||||||||||||
| name | Name of the submitted job (The script file name is given as default.) | ||||||||||||||||||||
| user | The owner of the job | ||||||||||||||||||||
| state | Job status
|
||||||||||||||||||||
| submit/start at | The time and date of submission | ||||||||||||||||||||
| queue | The name of queue containing the job fast queue: fast@ijis-xx-xx.ijis.iarc.uaf.edu |
||||||||||||||||||||
| slots | The number of used slot | ||||||||||||||||||||
| ja-task-ID | Job array task ID (blank for the non-array job) |
Job cancellation
Users who submit the job or the superuser can cancel the job with the "qdel" command irrespective of the execution status.
Executable directory
A user's jobs are performed in the user's home directory. The following option is used to perform a job on a current directory.
The "fast" queue is for jobs with runtimes of less than one hour
Please specify the "fast" queue when you run a job in priory to other jobs. The program execution time of the job must be less than 1 hour. Keep in mind that the job will be canceled by the system if the execution time exceeds the 1-hour limit.
The "large" queue is for a job with runtime of more than ten hours
The execution time of the "normal" queue is restricted to 10 hours. Please specify the "large" queue when executing a job longer than the limit. Note that the priority of the job would be set as the lowest.
Further details about NQS
For more information about NQS, please see the following Web pages.
- http://docs.sun.com/app/docs/doc/817-6117/ (English)
- http://docs.sun.com/app/docs/doc/817-7680/ (Japanese)
