Get Started

Getting an Account

  • Eligibility: you must hold principal investigator status at OSU

    • Students, post-doctoral fellows: work this out with faculty mentor. You can be added to an existing project or a new project can be created.
  • Application: Sign up for an OSC account


  • Types of Accounts
    • Startup – requires only CV/resume and a very brief description of the project
    • Classroom – class syllabus is required (account lasts for one semester)
    • Standard – requires proposal (3-5 pages)
    • OSC’s help page


    Initial allocation of computing time for Standard account is twice as much as Startup account.
    But Startup account will be ample for most users.

File Management

  • Up/downloading files is easily done using OSC’s OnDemand website


  • Click the Files menu and select the Home Directory option
    • file management is entirely intuitive at this point

Overview

Permission to use Stata

  • Send an email to OSC Help () stating you would like to use Stata

  • OSC Help will ask you to sign a software license agreement

  • When PI adds a student or colleague to a project, the new users will need to sign the software license agreement.

Using Stata

  • There are two modalities:

    1. interactive session

    2. submit a batch job

  • Both types are simple to launch through OSC’s OnDemand website where you can log in and access the supercomputer


Interactive Session

Interactive Stata sessions

  1. From the OnDemand Dashboard, click Interactive Apps and select the Stata option.

  2. Fill in the Number of hours and the Number of cores, then click Launch.

A Stata session will start in your web browser when the computing resources become available.

Batch Jobs

Submit a Stata Batch Job

Row

Step 1
create do-file

To illustrate, I uploaded a new do-file called simpleScript.do with:


capture log close
log using simpleScript.log, replace

sysuse auto, clear
ds

gen gphm = 100/mpg
su gphm

regress gphm weight
log close

Step 2
create batch script

  • Create a new script (to submit the batch job) by clicking on New File (I named mine: runStataBatch.sh).

  • Select your new script and click Edit. Copy and paste the following text into the new file

Be sure to replace path/to/folder/with/do_file and name_of_do_file.do with the path and do-file name that you created.

  • This job will run for 1 hour (walltime) on 1 node with 8 cores using 4GB of RAM
  • For more on changing these settings see Configuration.

Step 3
create new job

  • Return to the OnDemand Dashboard (with the File Explorer) and click on the Jobs menu.

  • Click on the Job Composer option.

  • From the New Job button, select the From Specified Path option.

Step 4
configure job

In the new tab/window, fill in

  • the path to your script file (created in step 2)
  • the name of your new job
  • the name of your script file
  • select the Owens cluster
  • click the Save button

Step 5
submit job

  • In the Job Composer page, run your job by clicking on the job name (from step 4).

  • Click Submit to enter your job into the queue.

  • This same page will indicate when the job has completed.

Interactive Session

Interactive session with RStudio

  1. From the OnDemand Dashboard, click Interactive Apps and select the RStudio Server option.

  2. Choose the R version, fill in the Number of hours and the Number of cores, and then click Launch.

An RStudio session will start in your web browser when the computing resources become available.

Submit R Batch Job

Row

Step 2
create batch script

  • Create a new script by clicking on New File (I named mine: runRBatch.sh)
  • Select your new script and click Edit.
  • Copy and paste the following text into the new file

Be sure to replace path/to/folder/with/Rscript and name_of_Rscript.R with the path and file name you created.

  • This job will run for 1 hour (walltime) on 1 node with 8 cores using 4GB of RAM
  • For more on changing these settings, Configuration.

Step 3
create new job

  • Return to the OnDemand Dashboard (with the File Explorer) and click on the Jobs menu.

  • Click Job Composer.

  • From the New Job button, select the From Specified Path option.

Step 4
configure job

In the new tab/window, fill in

  • the path to your script file (created in step 2)

  • the name of your new job

  • the name of your script file

  • click Save

Step 5
submit job

  • On the Job Composer page, run your job by clicking on the name of the job (from step 4).

  • Click Submit to enter your job into the queue.

  • This same page will indicate when the job has completed.

Configuration

Configuring Batch Jobs

  • #PBS -l walltime=01:00:00 the amount of time (hours::minutes::seconds) allocated to your job (your job will be terminated – without saving – when this time is up)
  • #PBS -l nodes=1:ppn=8 the number of nodes (1) and cores (8) your job will use (note, Owens only has 28 cores per node)
  • #PBS -l mem=4GB memory needed across all nodes
  • #PBS -j oe join the two log files (standard output and error messages) into a single file
  • module load stata load the software into your computing environment
  • cd path/to/folder/with/do_file move to the folder with your do-file
  • stata-mp -b do name_of_do_file.do Stata command for submitting a batch job

Misc

About OSC

The Ohio Supercomputer Center (OSC) offers high-performance computing. Use of these resources requires an OSC account. Accounts are project-by-project, not person-by-person. You may set up one project account for all your work, or you may be a member of multiple projects.

Additional information

  • Publications that contain results obtained using the Supercomputer should cite OSC. Examples of citation formats can be found on this web site: https://www.osc.edu/citation. A bibliographic reference also should be emailed to .

  • Terminology
    • Research Units: Every project starts with an amount of computing time available, measured in research units (RUs). An RU is roughly 10 core-hours (e.g., one core running for 10 hours or 10 cores running for 1 hour). Startup and Classroom accounts start with 5,000 RUs; Standard accounts start with 10,000 RUs.
      Link to details on how RUs are calculated

    • Portable Batch System (PBS): A computer program that schedules when BATCH jobs begin to run. The wait time is based on the number of batch jobs currently in the queue, and the amount of resources (nodes, cores, memory, and computing time) requested. Link to more information about PBS commands