![]() |
|
||
| |
|||
DMOL on BiowulfDMol3 is a Density Functional Theory (DFT) quantum mechanical code that allows users to study problems in chemistry and materials science with high accuracy and reliability. It is developed by Accelrys and can be used to simulate processes in gas phase, solvent, surface and solid environments.DMOL v3.2 is the default. Users can choose to run v 3.1 for comparison or to complete ongoing projects by following the instructions further down this page. [Setting up your environment] [Submitting a parallel job]
For bash shell users: eval `/usr/local/msi/MaterialsStudio/Licensing/Setup/lic_setup.sh -s sh` eval `/usr/local/msi/MaterialsStudio/share/bin/ms_setup.sh -s sh` DMOL3_DATA=/usr/local/msi/MaterialsStudio/Data/Resources/Quantum/DMol3; export DMOL3_DATA DMOL_TMP=/scratch; export DMOL_TMP For csh or tcsh users: eval `/usr/local/msi/MaterialsStudio/Licensing/Setup/lic_setup.sh -s csh` eval `/usr/local/msi/MaterialsStudio/share/bin/ms_setup.sh -s csh` setenv DMOL3_DATA /usr/local/msi/MaterialsStudio/Data/Resources/Quantum/DMol3 setenv DMOL_TMP /scratch These commands will run the default version of DMOL, v 3.2 Users who wish to specifically run v 3.1 should replace MaterialsStudio with MaterialsStudio3.1 in the lines above.
This example assumes that the user's environment has been set up as above, and that the required DMOL input files have been previously created from Cerius, Visualizer, or another program. Create a simple batch command file along the following lines:
This command file would be submitted with: qsub -v np=4 -l nodes=2,dmol=4 dmol_run.shThis job will use 4 processors (np = 4) on 2 nodes (-l nodes=2). In our experience DMOL runs do not scale to more than 4 processors (2 nodes). Users who run jobs on more than 2 nodes should justify this decision by running their own benchmarks. (We would be very interested in seeing the results of DMOL jobs that scale better than 4p; please email staff@helix.nih.gov). Note: DMOL resources The PBS Batch system on the Biowulf cluster keeps track of the DMOL licenses via the DMOL "resources". Jobs requiring a DMOL license will be started or remain in the queue depending on the availability of DMOL licenses. All DMOL jobs must therefore specify the appropriate resource when the job is submitted. For example, qsub -v np=4 -l nodes=2,dmol=4 my_dmol_scriptspecifies that the job requires 4 dmol licenses, one for each processor. If a job appears to be queued for a long time when there are nodes available, you can check its status via: qstat -f job_numberThe output from this qstat command should show the dmol resource required.
[susanc@biobos test]$ qstat -f 64402.biobos
Job Id: 64402.biobos
Job_Name = DMOL_CHA
[...]
Resource_List.dmol = 4
Resource_List.ncpus = 2
[...]
comment = Not Running: Waiting for dmol resource
The 'comment' line near the end of the qstat output will explain why the
job is queued. In this case the batch system is waiting for DMOL licenses
to become available before starting the job.
Submitting jobs to the Altixfirebolt.nih.gov, the Altix, is considered a "fat node" on the Biowulf cluster, and jobs are submitted via the PBS v7.1. To specify that a DMOL job should run on the Altix, use altix as the resource, and specify the number of processors with ncpus=#.qsub -v np=6 -l nodes=1:altix:ncpus=6,dmol=6,mem=6gb /data/username/large/dmol.run.altixnp=6 is an environment variable passed to your batch script. ncpus=6 tells PBS how many Altix processors to allocate to this job. dmol=6 tells PBS how many 'dmol' resources (= DMOL licenses) are required by this job. mem=6gb tells PBS that 6GB of memory is required for the job.
| |||
This document is available as http://biowulf.nih.gov/apps/dmol.html Biowulf home page | Helix Systems | NIH |
|||