Using Software

  • If doing something that does not require much CPU time or memory, you can work on the login node. Otherwise, you should work on a compute node (see below).
  • If anything you're running involves a lot of input/output, you should use your /scratch/YourNetIdHere directory to store files rather than your /home/YourNetIDHere directory.

From Applications Menu

Several software packages with graphical user interfaces are available from the Applications Menu when you connect to BlueHive using X2Go. If the one you want to use is among them,

  1. Select the package you want from the menu. A JobLauncher window will come up.
  2. You can run on the login node by checking Run on Bluehive. Otherwise, leave unchecked to run on a compute node.
  3. Optional: select options for duration, number of nodes, etc.
  4. Click Submit Interactive job. If on a compute node you might have to wait for your job to start.

From an interactive session at the command line

  1. Open a terminal from the Applications menu (Applications -> System Tools -> Terminal)
  2. Commands typed in the new terminal window will run on the login node. To run on a compute node, type interactive. You might have to wait for your job to start. After it starts you will get a command prompt on a compute node.
  3. Find the software package you want to use from the Software Index. Then select the module name corresponding to the particular version you want. For example, for the R software package used for statistical analysis, you could select r/3.1.0/b2. If the package or version you need is not installed, email us and we will install it.
  4. Find the name of the executable or command to run. You might need to look at the documentation for the software package to find this.
  5. In the terminal window, type
    module load YourModuleNameHere
    YourCommandHere
    

    For example, for R, YourModuleNameHere would be r/3.1.0/b2 and YourCommandHere would be R.

On a compute node, non-interactively, using an sbatch script.

  1. First, make sure you know the command for running your software non-interactively (using an input file, etc.) It's a good idea to first run a small test case on the login node. For example, for R, a very simple input file is just the line 1+1. Say this line is saved in a file called test.R (to create the file you can use gedit from the Applications -> Accessories menu). The command to run R using that input file is then R --no-save < test.R.

  2. Create another file containing the lines

    #!/bin/bash
    #SBATCH --time=00:05:00 --output=test.out
    module load YourModuleNameHere
    YourCommandHere
    

    Again, for R, YourModuleNameHere would be r/3.1.0/b2 and YourCommandHere would be R --no-save < test.R. Say these lines are saved in a file called test.sbatch.

  3. Submit the job by typing sbatch test.sbatch. You can look at the status of this job (and all the jobs you are running) by typing squeue -u YourNetIdHere. In particular the fourth column with heading ST gives the status of the job: PD for pending, R for running, or CD for completed (these are the most common but there are others). After the job has completed the file test.out will contain the output.

  4. There are many options to customize how your job is run. See Running Jobs or type man sbatch in a terminal for more information.