Statistics Department

IT/Computing

Due to the popularity of LINUX clusters, you have a couple of different clusters available to you. Which one will work the best/fastest is really up to you. The department has put together a small cluster designed to meet the needs of the STATISTICS Department only. We have created tools which hopefully will make it easy to submit your program. We are willing to work with you personally on your initial job submission to our cluster, to answer questions or problems you may encounter.

The files stored on the cluster are NOT backed up. The disk space is available for running simmulations only. It's not uncommon to loose your files due to poor passwords and/or the cluster being hacked. Please keep your results and important code backed up elsewhere.

[Statistics Cluster]

Hopefully to run a job is as easy as the following steps. I'm sure there are more details that should be explained, however this is a basic guideline which will be modified as I assist more students.

1. Request Access
   Email help@stat.ncsu.edu stating you wish to have login access to the beowulf cluster.
2. Transfer your programs to cluster.
   {from mac} $ scp [program] unityid@beowulf.stat.ncsu.edu:
   {from pc} winscp beowulf.stat.ncsu.edu
3. Login to Cluster
   {from mac} ssh unityid@beowulf.stat.ncsu.edu
   {from pc}putty -ssh beowulf.stat.ncsu.edu
4. Submit job
   $ bwsubmit [program] [filename]
   {Examples}
   $ bwsubmit sas program.sas
   $ bwsubmit r sim.r
   $ bwsubmit exec ./fortran.exe
   $ bwsubmit matlab prog.m
5. Log Off Cluster
   $ exit
6. You'll receive an email when your program has finished running.
7. Login to Cluster
   {from mac} ssh beowulf.stat.ncsu.edu
   {from pc}putty -ssh beowulf.stat.ncsu.edu
8. Transfer your results from cluster.
   {from mac} $ scp beowulf.stat.ncsu.edu:[output] ~
   {from pc} winscp beowulf.stat.ncsu.edu
9. Log Off Cluster or return to Step 4.

Other usefull commands

bwcheck - Check the status of the jobs/queues.
bwkill {jobnum} - To stop/kill a submitted job.
bwload - To see the cluster load/status.
nano {filename} - Allow you to make easy modifications to your files.

Other clusters you can use have information displayed at the following sites:

• PAMS HPC
  
• NCSU HPC

  These clusters may have more nodes and possibly faster CPUs, however the ease of use and the compatibility with Statistics applications are determined by your needs.  

[Back]

webmaster@stat.ncsu.edu