Running VSimComposer On a Remote Computer System

Just as on a local workstation or laptop, the computational engine (Vorpal) may be invoked through the graphical interface or from the command line on a remote system. On high performance computing clusters the command line approach may be required in order to submit a job to a resource management system. These are documented separtely here: - Running Vorpal from the Command Line - Running Vorpal from a Queueing System

In this section we discuss alternatives for setting up, running via the GUI, and visualising output.

In the present version we offer the following capabilities for running VSim remotely:

Note

Prior to starting up VSim, it may be necessary to set the environment variable export LIBGL_ALWAYS_INDIRECT=1 in order for the visualization stage to work correctly. Some users using VNC on ubuntu 14.04 have also reported adding the system installation of mesa to the start of the LD_LIBRARY_PATH environment variable, has helped overcome their issues.

export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/mesa:$LD_LIBRARY_PATH

  • Remote Desktop: We have tested two different methods for setting up a remote desktop on a server. One method uses VNC such as Tight VNC and Turbo VNC. If a VNC server is set up on the remote machine, one may try to connect a local client to the remote machine using VNC. Another method is through the use of NoMachine. In order to log in to a remote server, the same software needs to be installed on both the server and the client. Please work with your local support staff to ensure any one of these software solutions is correctly installed on both the server and client.

  • Virtual GL: If one is using a remote machine that has virtualGL server or DCV one may run using the hardware acceleration on the remote machine. This may provide the best performance but also the most system administration work. Many HPC centers are already set up for this kind of access. There is a super-accelerated virtualGL client, but it is more common to find virtualGL set up like DCV such that the remote machine is running a VNC server, which you may connect to with any VNC client on your local machine.

  • X Windows: If one does not have hardware acceleration on the remote machine one may forward X using an ssh client (ssh -Y) or use accelerated X forwarding using software like NoMachine NX. As of this writing, a good discussion is at (https://www.hoffman2.idre.ucla.edu/access/x11_forwarding). Briefly,

    • Linux users running X: edit /usr/bin/Xorg as described at the above link.

    • OS X users running XQuartz: execute

      defaults write org.macosforge.xquartz.X11 enable_iglx -bool true

    • Windows users: many options described at the above link. In addition, you can install an X server (such as XMing) on your Windows machine. You will also need to install Cygwin. Follow these steps to open Composer on a remote Linux machine. (1) Start the XMing server on your local computer. (2) Open Cygwin and type the following commands in the Cygwin terminal: - startxwin & - export DISPLAY=:0.0 - xterm & - ssh -XY address.to.your.remote.machine - navigate to location of installed VSim software and issue the command ./VSimComposer.sh

      If X11 forwarding is enable on the remote machine, then this should open a Composer window.

Visualizing Remote Data

The following capabilities are recommended for visualizing remote data if the previous recommendataions do not work for you:

  • One may use an external utility to copy the remote files back to a local machine (scp, sftp or winSCP are likely options). Providing the .pre or .sdf file is in the same directory as the data to be viewed, it should be possible to visualize the output locally. This is the most appropriate solution for those dealing with small datasets.
  • Use standalone VisIt and it’s remoting feature. VisIt may be downloaded from (https://wci.llnl.gov/simulation/computer-codes/visit/downloads). Users of native VisIt should be sure to make sure their local version number matches up with the version running remotely. This option offers the full flexibility of VisIt, such as the ability to make all the individual images needed to make movies in a single go and the ability to fully Python script your visualisation, but there is a corresponding learning curve. There are many tutorials and resources at The visitusers website (http://www.visitusers.org).
  • Use matplotlib or alternative software on the remote machine. The VsHdf5 module is provided with VSim and may be used to read and manipulate remote datasets for this purpose.
  • If you are using VNC or NoMachine, you can open any data set through Composer just as you would on your own desktop.