This article describes the workflow and settings/considerations to run M-STAR CFD software on the Nimbix cloud.
To access M-STAR on the Nimbix platform, the following steps are required:
1. Select M-Star release from the Compute dashboard.
NOTE: If the option is not available in the first-page menu, press on “More” at the bottom of the page as shown in the image below:
2. A splash window will open. Select the M-Star Build option as shown below to start model set-up:
3. A set-up screen opens next, and here you must choose some of your settings by clicking on the Tabs on the top of the window (General, Optional, etc.) one tab at a time.
UNDER GENERAL TAB
1. Under Machine type, when you click on the caret on the right, you can select the type of machine you want to run your job on. The decision on machine type selection is based on the size and complexity of your model and the cost associated with the machine type (some machines will have higher RAM, others will only run the job on a single CPU, others will have better graphics and therefore higher cost, etc.).
NOTE: When running interactive based applications, you’ll find that selecting an NC9 or any NC* machine types should offer significant visual performance over not selecting an NC machine type. By selecting an NC machine, this places a GPU on your head-node and offers better visual performance. Another thing to keep in mind is that when running interactively, you can use a web browser, or in some cases, for large models, or you might consider using RealVNC.
M-Star NOTE: It is recommended to use a machine with 20 cores for running M-Star on NIMBIX with 1 node for optimal performance.
2. Select the number of cores:
The machine types you selected in the previous step will dictate the increment in the number of cores that you can choose/select. For M-Star applications, it is recommended to leave the default 20 cores (1 node) option for optimal performance:
NOTE: Do not confuse the number of cores with the number of nodes (nodes represent the number of increments of cores you selected. In the example above, 1 node represents 20 cores – recommended for simple M-Star applications).
UNDER OPTIONAL TAB
1. Assign a JOB LABEL (give a name that will help you keep track of your running jobs. For example, My_MStar_Project):
2. Leave blank the wall time limit, Casefile (for a new simulation), and the IP address. The Window size needs to be kept as default.
UNDER STORAGE TAB
Select vault type: The default vault is “Elastic_File.”
The "Elastic_File" vault is recommended for small to medium size jobs, such as M-Start projects, Icepak projects, simple linear Mechanical Analysis projects, some HFSS, and simple Fluent projects (not multi-phase). For any complex and computationally heavy jobs, and where partitioning the job over a number of cores becomes challenging, the “Performance_SSD” vault is strongly recommended. The “Performance_SSD” vault can be found in the drop-down under the "Select Vault" tab (NOTE: requires a subscription and extra monthly payment to have access to the “Performance_SSD” vault).
Before submitting your job for running, you can preview your settings under the PREVIEW SUBMISSION tab.
Start your M-Star NIMBIX job (pre-processor) by clicking on the SUBMIT button as shown below (your preview image may differ):
After M-Star has been successfully launched, you should be able to see the M-Star preview and can access the window by clicking on it (you can see details on the number of nodes, some information regarding memory, etc.). Click on the preview window in your browser to open M-Star in a new window/tab. You are now inside the M-Star environment.
NOTE: If the preview window does not show in your browser, click on the “Click here to connect” link to start the M-Star in a new browser window or tab.
M-STAR Setup and Solution On The Nimbix Cloud (abbreviated steps)
1. Start by setting up the new simulation: Set-up simulation parameters in the Model tree:
NOTE: Ensure the proper simulation parameters (mesh density, time step, run time, etc.) are set properly. This particular example is a single-phase, Newtonian, particle mixing in a tall, rectangular tank open at one end using a 6-blade propeller that rotates at 50 rpm CCW simulation.
Define your model, loads, and boundary conditions (preprocessing) using CAD imports or M-Star primitives. Save your work periodically.
NOTE: Ensure that each pannel is extended to its full width if the options in the panel are obscured from your field of view by pulling the panel edges to the right of the browser window:
2. Create the stationary domain, rotating domain, and particle domain by clicking the “Create” button and choosing the desired options:
3. You can add meshing operations (refinement), move/rotate your components, create cutouts (refer to M-Star website for tutorials and additional documentation pertaining to your specific problem)
4. Once the model is set-up and ready to run, save your project in the working directory of your choice: File > Save As.
NOTE: Choose a filename (see below My_Particle_Mix.msb for example). To create a new directory, click the “Create Folder” icon on the right side of the pop-up window, enter a folder name, and click the Save button (see below).
5. Set-up the model solver by clicking the Run button in the toolbar:
A “Select Case Directory” window opens. This is the location where M-Star will save the case file and output file. Click the “Open” button to accept the current directory.
NOTE: You may need to create a case directory to segregate your output file from any other M-Star projects you have in the current directory. To create a new directory, click the “Create Folder” icon on the right side of the pop-up window, enter a folder name, and click the “Create” button.
6. Click on the Start New option and monitor the simulation run. Check to ensure the number of CPUs is 20 (matches the number of CPUs chosen during NIMBIX setup) or less.
NOTE: Running the solver in GPU mode (click the check box next to GPUs) will result in a faster run than in CPU mode.
NOTE: You can also rename your output folder (default “out”) by choosing the “Advanced” option from the “Run” menu. The advanced option allows you to rename the output file in the current directory.
7. Monitor your system performance (as needed). There are two places to monitor performance:
a. Detailed Job Metrics in the NIMIBIX dashboard (useful when running in CPU mode)
b. M-Star output window (when running in GPU, CPU mode, or keeping track of solution statistical output and time marching scheme).
8. When the simulation is done, click “Continue” to finish writing the output (out) file (optional), then click “Exit”.
9. Save and Quit M-Star to proceed to post-processing (visualization of your results): File > Save followed by File > Quit.
Post Processing M-STAR Results Using ParaView
1. In the NIMBIX dashboard, open ParaView 7 (note that ParaView 5.4 or newer version is required to visualize M-Star results without the M-Star Post add-on). Slices or numerical output files are also available in the “out” folder for off-line processing using Python scripts:
2. Use .vtk (open source) files to visualize the regions of interest (recall you can find your results such as slices and animation files in the “out/Output” folder in the case directory unless you used the “Advanced” solver option to rename your output folder):
NOTE: The .vtk and .stl files can be rendered using any general-purpose visualization software, including ParaView. The .pvd files are used for visualizing particle flow and motion of rotating bodies, and .stl files can be used to render stationary and rotating bodies.
3. Post process per your requirements (for example, you can animate the particle flow as shown below):
NOTE: Particle flow can be visualized using “Glmph” mode in ParaView to render the particle as points (in this case, constant solid color with uniform 5000 points distribution). Click on the green “Play” button to play the animation such as particle dispersed by the 6-blade impeller as they are dropping from about 25cm above the impeller as shown below:
4. Close Paraview when done to save processing time.