ATK On-Demand On Sabalcore

This guide describes the steps you need to follow to activate your account on Sabalcore, configure your local VNL installation, and run ATK On-Demand jobs. For more information about prices visit the Atomistix ToolKit On-Demand page.



Create an ATK On-Demand account

Create an account on Sabalcore. Sign up at this link to create your personal account and get 150 free core-hours of ATK On-Demand license. Sabalcore will assign you a specific username, e.g. “atkuser01”.

Generate password-less SSH public keys

You will need password-less SSH communication between your local computer and the Sabalcore cluster. Therefore, generate a password-less SSH public key on your computer and email it to support@sabalcore.com along with your username.

Please refer to the tutorial SSH keys for more details on how to create the SSH key.

Important

Please DO NOT send your private key, it must remain secure on your computer.

Set up the Job Manager in VNL

Note

It is assumed that you have VNL installed on your local machine so you can submit jobs directly from the Job Manager in VNL on the remote cluster. Check the Installation guide for VNL and ATK 2017 for more details on how to install VNL. If you are an academic user, you can request a free perpetual VNL license. Please refer to this page for more details.

Once Sabalcore has confirmed that your public SSH key is uploaded to your account, you can set up the job_manager_icon Job Manager in VNL for remote execution of your ATK jobs on Sabalcore.

  1. Download the Sabalcore machine settings file.

  2. Start VNL and open the Job Manager. Click the “Machine Manager” icon and add a new Remote PBS machine:

    ../../_images/machine_manager.png
  3. Click the “Import” button and browse your file system to select the downloaded Sabalcore_settings_atk2016 machine file:

    ../../_images/settings_tab.png
  4. Edit the following fields to match your personal settings:

    • Settings Tab

      ../../_images/settings_tab_edit.png
    • Environment Tab

      ../../_images/environment_tab.png

      If you need to load a specific version of ATK, you can edit the “Modules to load” field and use the corresponding module, e.g. “atk/2015”.

    • Notification Tab

      ../../_images/notifications_tab.png

      Add your e-mail address if you want to receive notifications about your job status (optional).

    • Diagnostic Tab

      ../../_images/diagnostics_tab.png

      Click on “Run Diagnostics” button to run the diagnostic tools. These tools check if the options in the Settings and Environment tabs allow your local computer to connect to the remote cluster and execute the commands needed for job submission and management.

  5. Press OK and close the Machine settings window

Important

Currently, “Maximum wall-clock time” settings are not working on this machine. Notice that you can always monitor your log file and kill your job at any time if needed by clicking the Stop execution of job button:

../../_images/kill_job.png

Test the ATK On-Demand setup

Finally, you should test the Job Manager settings and run a simple ATK job:

  1. Download the ATK Python script sic_bandstructure.py and save it in your project folder.

  2. Drag and drop the script from the Project Files list onto the job_manager_icon Job Manager icon

    ../../_images/drag_and_drop_job.png
  3. Select the newly created “Sabalcore – ATK 2016” machine:

    ../../_images/sabalcore_machine.png
  4. Open the Job settings:

    ../../_images/edit_job_settings.png
  5. Setup your job to run in parallel with several processes, e.g. 4 MPI processes per node (see Job Manager for remote execution of ATK scripts for more details on how to set up your MPI parallel runs):

    ../../_images/run_job.png

    The examples below show how to set up the Resources settings to use 16 MPI processes on a single node (left) or 64 MPI processes on four 16 cores nodes (right).

    ../../_images/resources_processes.png
  6. Click OK and click the Play button to run the simulation. In the central panel you can monitor the status of your job. For more information visit the Job Manager for remote execution of ATK scripts tutorial.

    ../../_images/play_job.png
  7. Once the job is done (it will take about 30 seconds) the Job Manager will copy the output .log and .nc files from the Sabalcore cluster to your local machine.

  8. Check that the file sic_bandstructure.nc is loaded correctly in the LabFloor and the band structure analysis object is loaded as well.

  9. Finally, open the sic_bandstructure.log file and verify that the job ran on four cores:

    Node list
    n716017
    Core list
    n716017
    n716017
    n716017
    n716017
    

with four MPI processes:

+------------------------------------------------------------------------------+
|                                                                              |
| CPU Information                                                              |
|                                                                              |
+------------------------------------------------------------------------------+
|  Process ID 0 at n716017                                                     |
|  Process ID 1 at n716017                                                     |
|  Process ID 2 at n716017                                                     |
|  Process ID 3 at n716017                                                     |
+------------------------------------------------------------------------------+

...

| Total number of processes: 4                                                 |

You are now ready to run your ATK On-Demand simulations!

Tip

You can see the list of available nodes and corresponding details by logging ini via SSH to Sabalcore and use the command upnodes

Tip

To check your account balance login using your SSH key and run the mybalance command. The mybalance command has a lot of useful options. Check also the myusage command to see a list of jobs that have completed.