Remote X

From CDOT Wiki
Jump to: navigation, search

The X Window System is a client-server graphics system:

  • the server manages the display hardware, each typically including one or more video cards and monitors, one or more keyboards, and one or more mice or other pointing devices.
  • the client software connects to the server through one of several network transports to draw on the display, get status information, and receive input from the keyboard and mouse.

In a typical Linux system, the server and clients run on the same computer, although one or more clients may also be remotely connected.


X servers are identified by a display specification or displayspec: [server]:display[.screen]


  • server is the hostname of the server: blank for local connections via the fastest avilable means; localhost for local connections via TCP/IP; or a hostname for remote connections via TCP/IP.
  • display is the display number. These are numbered starting at zero. The port number used for TCP/IP connections is (6000+display).
  • screen is the screen number if the display supports multiple screens (this is mostly obsolete; most multi-screen X servers now merge them into a single display, using the Xinerama protocol extension).


  • the default local display number is :0
  • display 0 on the host is

Displayspecs are communicated to client programs through the DISPLAY environment variable (or, less often, through command-line arguments).

Remote Access to an X Server with SSH

To run a client in a remote location from the server, you can run the raw X protocol over tcp/ip or other network transports. However, this is very insecure and is not recommended. Instead, current best practice is to use an ssh session to tunnel the X traffic.

Most standard ssh clients can set up the tunnel, X authentication, and other details necessary by simply including the -X or -Y command line parameters.

In these examples:

  • red is the local computer, running an X server
  • blue is the remote computer on which the client program will be run
  • user is the name of the user on blue

To connect to blue and start a shell, from which any program may be run:

ssh -XC user@blue

The -X argument sets up the X tunnel, while the -C argument turns on data compression for improved performance.

To connect to blue and start the program gnome-calculator:

ssh -XC user@blue gnome-calculator
Not for Benchmarking
This is not a good solution for benchmarking because the performance will vary with network conditions.

Using VNC

VNC is a remote display software that originated at Olivetti Research Labs in Cambridge UK. It is cross-platform and very lightweight. The underlying protocol is called rfb (for remote frame buffer). There are many different software packages that support various aspects of the vnc/rfb system.

When used with the X Window System, it is necessary to have a dual server that functions as both an X server and a VNC server. The easiest way to do this is to run the vncserver script from the tigervnc-server package, which configures and then start the Xvnc server. To connect X client, set and export the DISPLAY environment variable; to connect a VNC client, use this command: vncviewer -via blue localhost:display

Using Xvfb

Xvfb is a dummy X server that does not display output on any device. This is useful where a program needs to connect to an X server, perhaps for testing purposes, but the actual image drawn does not have to be viewed anywhere. Xvnb would typically be invoked with a command such as:

Xvfb :0 -ac

Where :0 is the display number and -ac disables access control to allow any local client to connect. You can then connect a client by setting and exporting the DISPLAY variable and starting the client:

export DISPLAY=:0


  • X Power Tools, a book by Chris Tyler published by O'Reilly Media and available through the Seneca Libraries on-line services.