Visit is a scalable data analysis and visualization application that provides server-based computation with local workstation visualization. Visit server is installed on Matilda, and can be accessed by users who install the Visit desktop client on their local machines. In this configuration, Visit allows users to open data sets located on Matilda, and perform analysis and data transformation on the cluster, ultimately displaying graphical results on their local workstation. This permits Matilda's resources to be used to do the "heavy lifting" while displaying results quickly and efficiently on your local workstation. It is also convenient for users who have performed data analysis on Matilda, as it allows data visualization without having to move those sets back down to the local workstation.
This document presents the steps and configuration required to connect to Visit on Matilda.
Who is Eligible?
Active Faculty or Staff
Client Installation
It is first necessary to install the client/desktop version of Visit that matches the version installed on Matilda (currently version 3.3.3). Download Visit for your workstation platform and follow the installation instructions.
Linux workstation installations have been tested by UTS. The recommended procedure for a Linux install is as follows:
- Download the appropriate "tgz" tarball for your platform
- Download the "visit-install3_3_3" installation script
- Place the tarball and installation script in the same directory (do not extract the tarball)
- Run the visit-install3_3_3 installation script in the form: ./visit-install3_3_3 <version> <platform> <installation dir> (Eg. sudo ./visit_install3_3_3 3.3.3 linux-86_64-fedora31 /usr/local/visit)
- Add the Visit exectuable directory to your PATH (Eg. export PATH=/usr/local/visit/bin:$PATH)
The "visit-install3_3_3" script should only be necessary for Linux installations. Mac installation packages are available in "dmg" format, and Windows packages are available as an "exe" executable.
Client Configuration
After installing Visit on your local workstation, start the application and you should see two windows - a "control" window and a blank "display" window (used for displaying plots).
In the control window, select "Options->Host Profiles". A new window will popup that has a "Hosts" sub-window, and a window with tabs labeled "Host Settings" and "Launch Profiles".
We will first need to create a new host profile for Matilda, so click the "New Host" button - this will open the "Host Settings" tab.
Fill-in the "Host Settings" fields as follows:
- Host nickname: Matilda (or whatever you like)
- Remote host name: hpc-login.oakland.edu
- Host name alias: <leave blank>
- Path to Visit installation: /cm/shared/apps/Visit/3.3.3
- Username: Your OU NetID
- Click the box "Tunnel data connections through SSH"
- Click the boxes for "SSH command" (ssh) and "SSH Port" (22)
After you are done with the settings, click the "Launch Profiles" tab, and click the "New Profile" button.
Here you will see 3 tabs: Settings, Parallel, and GPU Acceleration. For now, we are going to deal with just the first two. For the "Settings" tab, you only need to fill in the name (in this case we used MPI). Next click the "Parallel" tab.
Under Parallel, we see two sub-tabs: "Launch" and "Advanced". You only need concern yourself with the "Launch" tab.
Fill in the Launch settings as follows:
- Click the box "Parallel launch method" and in the dropdown select "sbatch/mpirun"
- For "Number of processors" we used "4" in this example, but you can use whatever you like. These will be the default settings, and can be altered when launching the job.
- Click the "Number of nodes" box and fill in the field. Here, we used "2", but again set this to whatever you like. This will be the default value, but can be changed when launching the actual job.
- Click the "Time Limit" checkbox. This is the default walltime that will be used, and can be altered when launching a job. In this example we used 1 hour (1:00:00)
When you are finished, click the "Apply" button (lower left of Host Profiles display), then "Dismiss" to close the window.
IMPORTANT: In order to retain the Host Profile you just created, select "Options->Save settings" from the main control window. This will retain the profile between sessions.
Running a Job
Here we present an example of an "Ensight" data set that we will load from Matilda. We will go through the process of creating a plot (using Matilda for computation) and displaying the result on our local workstation.
To start, from the main control menu click "File->Open file". A window will popup that initially shows the filesystem on our local workstation. Click on the dropdown and select the name of the Host Profile you just created - in this case, "Matilda".
It may take several seconds for the window to respond, and in between you will likely see another popup indicating a connection is being established.
You should now see "hpc-login.oakland.edu" in the "Host" box, and the "Path" should correspond to your home directory on Matilda. You can navigate to scratch or projects if desired, by changing the path manually, or by clicking in the "Directories" sub-window. In this example, we are going to open an "Ensight" files named "arts.case", making sure to select "Ensight" in the dropdown labeled "Open file as type". Click the "OK" button to proceed.
It may take several moments for the file to open. In some cases, you may receive a message stating that the GUI isn't responding. If that happens, click "Wait" as this will usually resolve. When the file is open, a new popup will appear labeled "Select options for 'hpc-login.oakland.edu'". You should see a list of Launch Profiles - in this case the "MPI" launch profile we created previously. Feel free to adjust the number of nodes, processors, and or walltime from the defaults you entered, and click "Ok".
You will again see the launch engine popup as a batch job is created on Matilda.
After that dialog disappears, if you check on Matilda you will see a job running that has reserved the nodes and processors you requested. If you now look at the "Selected files" window (top of control window), you should see "hpc-login.oakland.edu", and underneath the file we just opened.
Now that the file is loaded we are going to add a Plot. Go to the "Plot" window and click "Add".
When we click "Add" we will see a series of plot choices. In our example, we are going to select "Pseudocolor->Mesh->EPSF". Now to render the plot in our Display window, we will click the "Draw" button.
It may take several moments for the drawing to render. You may get a warning message that the GUI is not responding. Click "Wait" once or twice (as-needed), and you should see the plot display in the Display window.
We can see our available "Compute Engines" by selecting "File->Compute_engines". A popup will appear which essentially shows the resources dedicated to our job.
