Registration Guide Register for HPC Access to HPCNew User Guide How to Run Batch Jobs
Portal User Guide
About the Service Subscribe to Pay-Per-Use Service Subscribe to Condominium Service Terms and ConditionsUser Guide
High Performance Workspace Service Utility Storage ServiceResearch Data Management ServiceCentre for Computational Science & Engineering
| Need help? | ||
![]() |
||
Requirements
In order to successfully run a job in an LSF queue, please observe the following:
- Your program has to be an executable and is readable on the remote machine where it is intended to be run.
- Your program is compiled to run on the operating system of the remote machine you intend to send it to.
- Your program is in the executable path, otherwise, you need to prepend './' to the command, or use absolute path.
- All input files has to be accessible when the program is running.
- All output files can be written to the destination directory.
- If it is your own compiled program, you may optionally link the LSF checkpoint libraries to your program.
To help you, we have provided temporary workspace on all HPC clusters Atlas2/3, Atlas4 and Atlas5 with /atlas3tmp, /atlas4tmp and /atlas5tmp, respectively. The high performance workspace for the clusters is to meet the demands of the I/O intensive analysis programs using the parallel file system technology. All users are recommended to use this workspace if your program is heavily I/O dependent to get better performance and efficiency. More information for the High Performance Workspace for the Computational clusters is available.
Commands
The basic command for submitting a job bsub. The following demonstrates a simple job submission:
> bsub -q linux64 ./a.out
This will send the job to the "linux64" queue. At this point you may want to have a quick peek at the monitoring commands to check the status of your job.
By default, the system will store all screen outputs from your program run in its buffer and send the results to you via email. If you want to save it to a file instead, use the -o and -e options:
> bsub -q linux64 -o outfile -e errorfile ./a.out
As the names in the example imply, the -o option specifies the file that stores the normal screen output of your program while the -e option specifies the file that stores the error messages that you normally get on screen.
Multiple Line Commands
The bsub command supports multiple line inputs, terminated by a CTRL-D. This is seldom used in SVU since options are not saved for subsequent jobs.
> bsub -q linux
bsub> -o outfile
bsub> -e errorfile
bsub> ./a.out
bsub> press <CTRL>-D
A better way preferred by many users is to use a job submission script.
Using Job Submisstion Scripts
Job submission scripts are text files that contain most of the bsub options that are normally typed in at the command line. An exception if the -q option which cannot be set in the script, but has to be set on the command line. A script for the above example would contain the following lines:
#BSUB -o outfile
#BSUB -e errorfile
./a.out/
After the script has been created (in this example, named myscript), the bsub command can be issued thus:
> bsub -q linux64 < myscript
Job Submission for Software
Job submission for software such as Abaqus, Amber, AutoDock, DiscoveryStudio, Fluent, Matlab and so on can be checked out from the software information page.
