Revision as of 11:23, 12 June 2017

Introduction

There are two visualisation nodes on Viper, visu01.hpc.hull.ac.uk and visu02.hpc.hull.ac.uk each with dual NVidia GeForce GTX 980 Ti graphics cards.

There are two main ways of using the visualisation nodes, X2go or VNC. VNC will tend to provide better performance on a fast network (i.e. when connected directly to the university network) while X2go will provide better performance on lower bandwidth connections (i.e. connecting remotely via the VPN).

Please note if you are using X2go off campus you must be connected to the university VPN.

Usage Example

A typical work flow to use the visualisation nodes is as follows:

VNC

• SSH into one of the visualisation nodes with X11 forwarding [ ]
• Start a VNC session
• Connect to the VNC session
• In VNC Session: Start a terminal in the VNC session and load the relevant application module
• In VNC Session: Start the application using VirtualGL
• In VNC Session: Close application
• Disconnect from VNC session
• Stop running VNC session

X2go

• Connect to one of the visu01.hpc.hull.ac.uk
• Start a terminal in the VNC session and load the relevant application module
• Start the application using VirtualGL

Connecting to the Visualisation Nodes

VNC

vnc-start

The first time you start a VNC session, you will be prompted to set a VNC password,. This is not linked to your university credentials, but should follow standard security advice so that it is a complex password. You will need this password to connect to future VNC sessions.

[username@rvisu01 ~]$ vnc-start

You will require a password to access your desktops.

Password:
Verify:
Would you like to enter a view-only password (y/n)? n

Desktop 'TurboVNC: rvisu01:3 (username)' started on display rvisu01:3

Starting applications specified in /home/username/.vnc/xstartup.turbovnc
Log file is /home/username/.vnc/rvisu01:3.log

If you forget the password you set, you will need to stop your running VNC session (see vnc-stop) then delete the directory ~/.vnc and run vnc-start again to set a new password.

When you next start a session you will see a message to say which session you have open:

[username@rvisu01 ~]$ vnc-start

Desktop 'TurboVNC: rvisu01:3 (username)' started on display rvisu01:3

Starting applications specified in /home/username/.vnc/xstartup.turbovnc
Log file is /home/username/.vnc/rvisu01:3.log

If you try and start a session while you have an existing session open, you will see the following message:

[username@rvisu01 ~]$ vnc-start
Please use the existing VNC session available on rvisu01:3

vnc-check

[username@rvisu01 ~]$ vnc-check
username rvisu01:3

vnc-stop

[username@rvisu01 ~]$ vnc-stop
Terminating VNC session on localhost:3
Killing Xvnc process ID 4718

vnc-viewer

[username@rvisu01 ~]$ vnc-viewer
Opening VNC session on localhost:3
libjawt.so path: /usr/lib/jvm/java-1.8.0-openjdk-1.8.0.77-0.b03.el7_2.x86_64/jre/lib/amd64
CConn: connected to host localhost port 5903
CConnection: Server supports RFB protocol version 3.8
CConnection: Using RFB protocol version 3.8

You should then see a new window open titled "Standard VNC Authentication [VncAuth]@rvisu01" with a password prompt. Enter your VNC password (not your university credentials) and you should then see a MATE desktop.

X2go

(Currently only available on visu01.hpc.hull.ac.uk )

• In the "Session" tab:
  • In "Session name:" to visu01.hpc.hull.ac.uk
  • In "Host:" enter visu01.hpc.hull.ac.uk
  • In "Login:" enter your username
  • Select Session type to MATE (or custom desktop with command MATE)
• In the "Connection" tab
  • When connecting from a university connection, set connection speed to LAN for best performance
  • When connecting from outside the university ADSL should be adequate
• In the "Media" tab
  • Uncheck the "Enable sound support" box
  • Uncheck the "Client side printing support" box

There are two main ways to use X2go, which are configured in the "Session" tab by using "Session type"

X2Go macOS

For info on using X2Go on macOS visit: X2Go macOS

xterm

To use a standard xterm terminal, in the "Session type" drop down menu select "Single application" and set command to /usr/bin/xterm

Full Desktop Environment

To use a full desktop environment, in "Session type" drop down menu select MATE. See further information about the MATE desktop environment below.

MATE Desktop

The MATE desktop environment, based on the popular GNOME2 environment and is available in both VNC and X2go. MATE is a relatively lightweight desktop environment.

When connected to your MATE session, start a terminal by either going to Applications > System Tools > MATE Terminal or Applications > System Tools > XTerm at the top of the screen, or using the black icon on the task bar at the top of the screen.

Note: If you leave your desktop session inactive for a period of time, the screen will lock. To open this you need to use your normal university credentials

Note: While the MATE desktop contains applications such as Firefox and PDF viewer, these are only available to support research use of visualisation tools and the visualisation nodes should not be used as a desktop replacement.

NVidia SMI

The command nvidia-smi can be used to see what is currently running on each of the GPU cards on the visualisation nodes. The nvidia-smi command can be used to access all sorts of information about the GPU cards, but just running the command with no options will provide basic information about the utilisation of each card:

[username@rvisu01 ~]$ nvidia-smi
Mon Feb 13 12:46:42 2017
+------------------------------------------------------+
| NVIDIA-SMI 361.42     Driver Version: 361.42         |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 980 Ti  Off  | 0000:02:00.0     Off |                  N/A |
| 22%   49C    P0    79W / 250W |   5662MiB /  6143MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 980 Ti  Off  | 0000:82:00.0     Off |                  N/A |
| 22%   26C    P8    15W / 250W |     28MiB /  6143MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0      1206    G   .../9.2.0/bin/arch-LinuxAMD64-Optimize/Avizo  5530MiB |
|    0      2785    G   /usr/bin/X                                      46MiB |
|    0     12468    G   ....1/bin/arch-LinuxAMD64-Optimize/AvizoLite    46MiB |
|    1      2785    G   /usr/bin/X                                       6MiB |
+-----------------------------------------------------------------------------+

Virtual GL

With VirtualGL, OpenGL commands and 3D data are redirected to a 3D graphics accelerator (each visualisation node has 2x Nvidia GTX 980Ti) before the rendered 3D images are sent to the client machine.

vglrun

Using the nvidia-smi command you can see what processes are running on card 0 and card 1. If there are any processes listed on either card other than /usr/bin/X then you need to ensure you use the other card (see the example above). For example, to do this you use the vglrun command as shown below:

[username@rvisu01 ~]$ vglrun -d0.0 command

or

[username@rvisu01 ~]$ vglrun -d0.1 command

vglwrapper

The command vglwrapper will round robin tasks between the two cards on the visualisation nodes:

[username@rvisu01 ~]$ vglwrapper command