Setup of Wireless LAN for thin client network

Jim Carter jimc at math.ucla.edu
Sat Nov 15 04:21:32 EST 2003


On Fri, 14 Nov 2003, Stephen Liu wrote:
> I am running a thin client system - LTSP/K12osn.  Each diskless desktop
> workstation is without hard drive.  They are connected to the server via
> a switch and running X server (X window) from the latter on their screen
> and booted with a rom on NIC.  Notebook connection is also possible
> booted from floppy with a booting image installed.
>
> Now I am prepared to convert the system to Wireless LAN.  According to
> suggestion for diskless desktop workstation it is possible.  For
> notebook chance is remote.

A number of years ago UCLA-Mathnet used a similar arrangement to good
effect (on wired Ethernet).  However, as the cost of discs and
general-purpose PCs went down we decided it was more cost-effective to put
US$600 PC's on people's desktops, and to keep them updated by
locally-written scripts.  This is particularly easy for Linux.  For Windoze
we have engaged "automatic update" and it is working out pretty well.
Plus we can get people to do much of their work, and to occupy swap space,
on their own machines rather than on the servers.

I encourage users to do on the local machine:
    xterm -e "slogin -X serverhost"
(using an alias or a menu item for the principal servers).  This puts most
of the resource usage, for the xterm, on the local machine.  But when they
use graphical apps on the server such as maple, matlab or mathematica,
X-windows service is available.  But we tell them to run matlab on the
local host too; our license scheme allows that.

On the wireless issue, the kinds of setup info (ESSID, WEP key) that's
needed could easily be included on a boot floppy, but I doubt that a boot
PROM would have the flexibility to handle this.  But who knows, you may
have found a vendor with a smart boot PROM.

If only a few people will use the system at a time, it should work OK.
Certainly at home and at work I use my laptop with X-windows across the
wireless net (though when possible I use the local xterm method) , but if
you have 50 to 100 workstations I'm afraid that they would fight for
bandwidth.  In comparison, your wired network switch can handle multiple
connections simultaneously, so the total throughput is several times the
nominal network bandwidth (10 MHz or 100 MHz), provided the users are
distributed over several servers.

One possibility is a hybrid system: run wires to most of the clients, but
for two or three that are impossible to reach, use wireless.  On the
notebook computers, I would recommend using a general-purpose operating
system so the users have freedom to do general-purpose work on them.
I have a user support page for setting up a wireless card:
    http://www.math.ucla.edu/computing/user_support/hardware/wiresetup.html

Hope this helps!

James F. Carter          Voice 310 825 2897    FAX 310 206 6673
UCLA-Mathnet;  6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90095-1555
Email: jimc at math.ucla.edu    http://www.math.ucla.edu/~jimc (q.v. for PGP key)



More information about the wireless mailing list