Operating system

Configuration data

Application data

User data

Hard disks

Audio devices

Printers

USB devices (memory sticks, WLAN cards, digital cameras, etc.)

Bluetooth devices (cell phones and many others)

PCMCIA cards (GPRS, ISDN, WLAN, and other cards)

Smart card readers

Firewire devices (hard disks, video cameras, etc.)

Except for the hardware connected to the clients locally (like audio devices, 

printers, USB sticks, Bluetooth devices, GPRS cards, smart card readers, and 

firewire devices), all other resources can be either on the client or server side, 

depending on what kind of storage or memory devices are built into the system 

(like hard drives, RAM, Flash RAM, etc.)

A classic diskless X terminal can boot its operating system (hopefully Linux) from 

Flash RAM or over the network (that is, with PXE boot), and use all other 

resources from the server, which means all applications except the ones 

available in the Flash ROM (that is, Mozilla, Java JVM, etc.) are started on the 

server and use the memory there. A diskless client can also mount certain or all 

parts of the file system (for example, the home directory, application directory) 

from a server and run the application using the local memory (some call that a 

Linux Net PC). 

Since modern thin client hardware is quite powerful, the second approach seems 

appropriate today and you do not waste local computing resources. If you can 

migrate your Linux applications on the fly from one machine to the other (with 

virtual machines (JVM, Mono

2

), openMosix

3

, or other single system image 

technologies) you could start your applications on some server, migrate them to 

you local client, and park them back to maybe another server when you 

disconnect. This allows you to continue your work later at exactly the same place 

where you left. Saving a complete application (state) to hard disk and restarting it 

would be needed here, so applications not being used would not consume any 

resources, but this is not possible right now with Linux.

You could use VMWare images loaded on demand on the server instead and 

use them remotely with Nomachine, VNC, Citrix, or Tarantella so you would not 

park single applications but the whole operating system. There are some 

2  

http://www.mono project.com

3  

http://www.openmosix.org

110 

Linux Client Migration Cookbook   A Practical Planning and Implementation Guide for Migrating to Desktop