Contents 1 FLUENT on the ITS HPC Cluster 1.1 Running a serial FLUENT job 1.2 Running a parallel FLUENT job 1.3 Documentation 1.4 Licensing 1.5 Performance

[edit] FLUENT on the ITS HPC Cluster

FLUENT is a software package for computational fluid dynamics. Version 6.3.26 is currently installed and available on the cluster.

To use FLUENT, you must modify your environment using the following shell command:

   module load fluent

For interactive use, type this command to the shell prompt and your environment will be modified appropriately to permit debugging/testing use in the current login session. If you logout and login again, you will need to load the FLUENT module again. For batch use, put this command in your batch script before the command that runs FLUENT.

[edit] Running a serial FLUENT job

To run a serial FLUENT batch job on the cluster, you need to create a PBS script. The sample script below includes a request for one processor in one compute node and imposes a limit of 10 hours of wall time. Note that the script includes the command to load the FLUENT module.

  #PBS -N name
  #PBS -l walltime=10:00:00
  #PBS -l nodes=1:ppn=1
  #PBS -j oe
  # cd to directory where execution will occur
  cd $TMPDIR
  # Copy FLUENT input files to this directory
  cp ~/test/test.tar .
  tar -xf test.tar
  # Load the FLUENT module
  module load fluent
  # Run FLUENT
  fluent 2ddp -g < input.txt > output.log 2>&1
  # Copy files back to home directory
  cp * ~/test

In this example, input files are copied to a local temporary directory on the node where FLUENT is running. This improves I/O performance of the FLUENT job while it is executing. The temporary directory is created automatically in /tmp and will have a name of the form pbstmp.NNNN.master.priv.cwru.edu on the compute node allocated to the job, where NNNN is the one to five digit job number. While the job is running, you can examine files in the temporary directory to confirm that the job is executing correctly by typing qstat -f NNNN to determine which node is allocated to the job and then using rsh to login to that node and examine the files.

If no FLUENT subroutines are used, only an input file (e.g., input.txt) and a file containing the problem statement (*.cas) are required. If user-defined subroutines are being used, the libudf directory should also be copied over.

[edit] Running a parallel FLUENT job

To run a parallel FLUENT batch job on the cluster, you need to make a few modifications to the instructions given above for running a serial FLUENT job. The sample script below includes a request for two processors in each of two compute nodes (for a total of four processors) and imposes a limit of 10 hours of wall time. Note that the script includes the command to load the FLUENT module.

  #PBS -N name
  #PBS -l walltime=10:00:00
  #PBS -l nodes=2:ppn=2
  #PBS -j oe
  # cd to directory where execution will occur
  cd $TMPDIR
  # Copy FLUENT input files to this directory
  cp ~/test/test.tar .
  tar -xf test.tar
  # Load the FLUENT module
  module load fluent
  # Run FLUENT
  fluent 2ddp -t4 -cnf=$PBS_NODEFILE -g < input.txt > output.log 2>&1
  # Copy files back to home directory
  cp * ~/test

This example is identical to that given for in the serial case with the exception that the '-t4' option tells FLUENT how many processors to use and the '-cnf=$PBS_NODEFILE' option tells FLUENT which nodes to use. In this example, input files are copied to a local temporary directory on each node where FLUENT is running. This improves I/O performance of the FLUENT job while it is executing. The temporary directories are created automatically in /tmp on all assigned nodes and will have names of the form pbstmp.NNNN.master.priv.cwru.edu, where NNNN is the one to five digit job number. While the job is running, you can examine files in the temporary directories to confirm that the job is executing correctly by typing qstat -f NNNN to determine which nodes are allocated to the job and then using rsh to login to those nodes and examine the files.

If no FLUENT subroutines are used, only an input file (e.g., input.txt) and a file containing the problem statement (*.cas) are required. If user-defined subroutines are being used, the libudf directory should also be copied over.

[edit] Documentation

Complete documentation in the form of PDF files is available on the cluster in the directory $FLUENT_HOME/help/pdf. Load the fluent module to define FLUENT_HOME properly. See Section 31.3 for specific information on running parallel FLUENT jobs in a Linux cluster environment.

[edit] Licensing

The cluster is licensed for FLUENT as follows:

FLUENT/GAMBIT: 12 Process Research License (includes 4 serial 8 parallel only processes)
FIDAP 12 Process Research License (includes Fieldview, 4 Serial, and 8 parallel only processes)
T-Grid ADD-ON Module
G-Turbo ADD-ON Module

[edit] Performance

Some investigation has shown that improved FLUENT performance can be gained for parallel jobs, at least in some cases, by minimizing the number of nodes used through the use of multicore processors. Prior to extensive use of FLUENT for a large number of similar long-running jobs, some experimentation should be carried out to determine the best type of node(s) to use.

This article is a stub. You can help by adding to it.

ITS High Performance Cluster Articles
This article is part of the ITS Cluster series of articles
FAQ | Intel compilers | GNU compilers | Portland Group compilers
Intel Math Kernel Library | MINPACK | ScaLAPACK | GSL | FFTW3 | MPICH | NAG
R | Mathematica | NAMD | GROMACS | Amber | MATLAB
FLUENT | GAMESS | Gaussian | MOLCAS | LAMMPS | APBS