Changes between Initial Version and Version 1 of vnc3d/details


Ignore:
Timestamp:
05/03/16 23:58:23 (8 years ago)
Author:
Jens Henrik Goebbert
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • vnc3d/details

    v1 v1  
     1\\
     2=== VNC - more details
     3
     4VNC follows a very general approach, where a remote graphical desktop on one node of the remote cluster is started. \\
     5On the user's workstation only a lightweight VNC viewer has to be installed. \\
     6The graphical screen of the remote desktop is send as an image from the cluster to the viewer on the fly.
     7
     8The user can work with this remote graphical desktop in the usual way, just by interacting with keyboard and mouse. \\
     9This is a very convenient way to work on a remote machine, not only for data visualization.
     10
     11VNC offers the ability to detach from a session (running in a VNC server) and then attach back at a later time. \\
     12That means, that the user can close the VNC viewer on his/her computer and any application started in the VNC session on the server keeps on running - they do not pause or even be killed. \\
     13Later on the user can connect back to the same VNC session (even from a different computer) and keep on working. \\
     14
     15[[Image(Trac_Setup_VNC.png, 380px)]]\\
     16
     17==== hardware accelerated rendering (OpenGL)
     18[[Image(https://upload.wikimedia.org/wikipedia/commons/thumb/6/67/X11transport.png/600px-X11transport.png, 200px, align=right, margin=0)]]
     19Whenever an OpenGL capable visualization software, like !VisIt, is started on the remote cluster node, \\
     20OpenGL commands can be redirected to the GPU of this node with the help of [http://www.virtualgl.org VirtualGL]. \\
     21* ... using '''vglrun''' (e.g. vglrun paraview)
     22This way the hardware accelerated rendering capabilities of a cluster node (if available) can be exploit for remote rendering. \\
     23(attention: __without__ VirtualGL software, rendering using the CPU instead of the GPU must be used (if available), which is much slower).