This short video tutorial demonstrates how to run simultaneous concurrent simulations on single cloud HPC server using Kogence Stack Builder GUI interface.
- To follow this tutorial along, please make a copy of public model Multiple simultaneous KOSS runs through Stack tab GUI by clicking on the Copy button at the top right corner. Users that are not explicitly added as Collaborators on a model cannot make changes to or run that model even if it is a public model. Users can make a personal copy of public models and then run their personal copy as it is or edit as per their needs. Copying an existing model that has been executed or copied many times before by other users on the platform is easiest and quickest way to get started.
- Optical scattering problem as defined in the
*.prbfile. This input file is used to define geometry, material properties, properties of incident light as well as parameters that control the accuracy of simulations. In this tutorial we want to study the convergence of optical simulation results as we choose to retain different number of spatial hamonics -- 5x5, 7x7, 9x9 and 11x1. We would like to run these simulations in parallel.
- Go to the Files tab. Open the
ProblemDefinition.prbfile by double click on it, select all the content (shortcut Ctrl-A) and copy to clipboard (shortcut Ctrl-C). We will make multiple copies of the file. Go back to your root folder. Click on Add/Upload Files button on top navigation bar. Click on New File and then give it a name such as
PRB7x7.prb. Past the copied content in this file (shortcut Ctrl-C). Change the values of
REMS_HP.NYto 7. Also change the value of
'testing7x7'. In KOSS all the data is save in files with the name of the project name. As we will run several simulation in parallel we do not want the same file to get overwritten multiple times. Save the file.
- Follow same steps to create 2 more files
- We will invoke KOSS using its 2 argument syntax
koss *.cmd *.prbto simulate this 2D photonic crystal geometry. For more invocation options, please check the documentation page for KOSS. Go to the Stack tab of your Model. Click on the + button to add KOSS software container in the software stack of our workflow. On clicking the + button, you should get a popup on which you can either search for KOSS or browse to find KOSS. Then click the + button next to the KOSS entry in that popup. This will add KOSS to your workflow. You will see a dropdown menu that allows you to select an entrypoint binary to run inside the container. Please select
kossbinary. You will then see a command text box next to it. Please enter
YourCMDFile.cmd YourPRBFile.prb. Do not forget to save the Stack stack tab settings before navigating away form this tab otherwise your settings would be lost.
- Follow the same steps to add KOSS multiple times in the workflow. Change the name of the PRB file in the command text box.
- In the Stack tab, lets check mark the box labeled "Run With Previous" next to the last 3 entries of KOSS. This allows all simulations to start together. So, for example, the second simulations will start while first one is still running. If we do not check mark this box then second simulation will only start after the first simulation ends. Lets save the Stack tab settings.
- Now we are ready. Lets click the Run button to start the simulation on our selected cloud HPC server. First time you request to boot up a server, it takes upto approximately 2 minutes. Once cloud HPC server boots up, you will see that Run button turns into Stop button. You will also see that the Visulaizer button becomes active. Clicking on the Visualizer button will connect you to your cloud HPC server. This is a full interactive graphical connection.
- You will see a few windows here. You can move them around for better visibility. You will see couple of log file that will show you if you are KOSS simulation is running or not or if it threw any errors. You will also see a system performance monitor window. Monitor has three tabs. If you go to the "Resources" tab, you can see how efficiently KOSS is using CPUs in the cloud HPC server. Under the "Process" tab you can see multiple instances of
- Generally, your batch mode simulations end automatically as soon as all the jobs complete. All the model files sync back to the Files tab. Billing stops irrespective of whatever time limit you selected in the Cluster tab.
- After simulation stops, we can go back to Files tab and explore or download the results. All graphically results from your simulation are also automatically pulled into your model's wiki page. Wiki pages come with a WYSIWYG Visual Editor. Feel free to move the graphical figures to appropriate locations and add some text in your model's wiki make to clearly document your modeling project and results so that it can shared with your collaborators and reviewers.
- After stopping your simulation if you want to make some changes and restart the simulation, this time your HPC server should come up instantly.