National University of Singapore

The most efficient and convenient way to run Fluent solver for your CFD simulations, which need hours, days or weeks to finish, is to run the solver in LSF batch/parallel mode. First, you shall setup the CFD problem, including mesh, models, boundary conditions etc., on an interactive Fluent interface. Next save the initial problem to Fluent case and data file pair. Then you can choose a suitable batch queue to submit your case for computing.

Batch Serial Fluent Job

serial | parallel | monitor

  bsub -q linux64 -o myjob.o "fluent 3d -g -i myjob_script"

where:

Here is an example of the script inside the file myjob_script (for steady-state case):

There have many ways to create/edit the text-based script file. To learn those ways, please check at the article on editing files on Linux from the HPC@NUS Issue No. 7 in 2005 March.

When the job is completed, you should find the final data file "test_1000.dat" at the working directory where you submitted the job. To do the post-processing and visualise the results, launch a Fluent interface interactively from the command line and read the case and data files into it to do the visualisation.
Here is another example of the script for unsteady flow simulation, or time-dependent flow:

When the job is completed, you should find 10 (=1000/100) sets of intermediate data file and the final data file "mydata_new.dat". You can use those data files to visualise the results

Using "fluent-job" Command

You can also use command "fluent-job" for batch serial Fluent job submission. Here is an example of the command:

fluent-job -s 3d -n 1 -i myjob_script

Take note that "fluent-job" will try to search and submit the job to a less busy LSF queue. The default job output file is "std-out.txt", and you can use a different file with option of "-o my_job.o". The script file should be created similarly as mentioned above.

Batch Parallel Fluent Job

serial | parallel | monitor

For long hour Fluent simulations, parallel computing with Fluent solver is encouraged. By running simulations on parallel model, or using multiple CPUs, the computing time will be shortened and hence improve the production. Fluent solver is well and ready parallelised using MPI and is able to run in parallel manner using multiple CPUs on all HPC clusters. The parallel queues supported Fluent solver include atlas3_parallel, atlas4_parallel, 4core_parallel and 8core_parallel.
To use parallel computing for a Fluent simulation, similarly as the serial-batch job process you need to setup your Fluent model first and save the model in Fluent case and data files. Initial debug and/or trial is necessary to ensure the model is properly defined.
At the same time, prepare the Fluent script file that is used for the job submission (same as above). To submit a parallel Fluent job, run command "fluent-job" and follow the 5 steps:

fluent-job 

1. Solver type [2d, 2ddp, 3d, 3ddp]: 3d  <Enter>
2. Fluent Inputput/Script file Name: myjob_script <Enter>
3. Number of CPU [1,2,3, ..., 16] <Suggested num is 4>: <Enter> 
4. Output file Name [e.g. std-out.txt]: <Enter>

1. SOLVER..................> 3d
2. Input File..............> run_0.5m
3. Num of CPU..............> 4
4. Stderr Output File......> std-output.txt
5. Run Queue...............> atlas4_parallel 

Enter <1,2,3,4,5> to change the above inputs, or 
Enter <y,n> to submit/discard this job. <1,2,3,4,5,y,n,q(quit)>: Y <Enter>

Job Monitor

serial | parallel | monitor

You can monitor the progress of your job with the following LSF commands: