r5 - 25 Mar 2005 - 19:12:10 - JPMcCYou are here: TWiki >  Ltsp Web  >  Clients > HowIDidIt

DIY LTS Client

Why build your own LTSP client?

When I saw one of the purpose made LTSP workstations from DisklessWorkStations.com at a conference a couple of years ago, I was really impressed. However, the costs of shipping, customs duty, etc made a personal import into Europe prohibitively expensive.

On the other hand, it proved impossible to buy local alternatives without paying for a copy of Windows CE. Following a challenge on the LTSP discussion group I decided the only solution was to build my own from components. Until I did it, I didn't know just how easy this is.

Choose a Motherboard

Following advice on the list, I bought an VIA EPIA 5000 Mini-ITX Mainboard. This crams a full-spec PC onto a compact 17cm x 17cm board. Everything you need for an LTSP client is there - video, audio, ethernet, serial & parallel ports, USB ... The only thing you need to add is low profile (25mm) memory. I bought a 128Mb PC133 SDRAM module, which was the smallest capacity I could find (Kingston part number KVR133X64C3L/128).

Add a Case

The neatest case I could find was a Cupid 3677. This isn't ideal, as it's got space for a laptop style CD-ROM and hard disk, which is wasted space with a diskless workstation. It's also got a fan, which you don't need with the Via board.

Ready to start

When you open the boxes and lay them out on the table, this is what it looks like:

Ingredients of an LTSP client

Assembly

This is embarrassingly easy. You just clip the memory into the motherboard, put the motherboard into the case, and fix it with the bolts provided. You then have to plug in the various cables. The cables are all labelled. The motherboard isn't, and there's no documentation in the box, so you have to download the User Manual. Once you've done that, it's easy:

Assembled client

Note: the fan isn't necessary and I find it noisy - just leave it unplugged.

Software configuration

The final stage is to set up the BIOS on your new client. It's the usual process of working through the screens, disabling anything you don't need (IDE drives, etc) and enabling those you do (boot from the on-board NIC etc.). One word of caution - I had to disable all the power-saving features, as the LTSP kernel doesn't seem to support them. Any time the power saving kicks in, the current session dies.

On the LTSP side, PXE boot works 'out of the box'. You only need to add entries into lts.conf for Sound, Printers, etc - for example, I do the following to enable sound and printing to a locally attached USB printer:

    XSERVER          = auto
    SOUND            = Y
    MODULE_01        = usbcore
    MODULE_02        = printer
    PRINTER_0_DEVICE = /dev/usb/lp0
    PRINTER_0_TYPE   = S

The Finished Product

The photo below shows the final result. The LTSP client sits neatly on the left of the picture. The beige box next to it is a cable modem.

The finished product

Is it worth the effort?

IMHO, it is. For less than £150 (including shipping and taxes), I have a neat, low power, totally silent unit that works perfectly with LTSP. And I haven't added a cent to Microsoft's turnover. That sounds like a good buy to me.

-- JPMcC - 25 Mar 2005

Edit | Attach | Printable | Raw View | Backlinks: Web, All Webs | History: r5 < r4 < r3 < r2 < r1 | More topic actions
 
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback