This page shall explain how to setup a simple network boot scenario.
For network booting, you will need two machines:
The client will not need any disks, so it can be built noiseless. Most “thin clients” on the market do not have any moving parts. If your network adaptor does not have a network boot ROM though, you will need either a floppy drive or a CD-ROM to load the network boot code (i.e. Etherboot) from. You could also burn Etherboot onto a ROM chip, or put it into your system BIOS, but that would be a bit difficult right at the beginning.
Assuming your client machine has a floppy drive, go to Rom-O-Matic, the on-the-fly ROM image generator. Choose the most recent Etherboot version available as “stable” (5.4.2 at the time of writing), select the network chipset in your client machine, choose format Floppy disk (.zdsk). Get ROM, and save that file to disk (on any machine, all that). You will now have to write it to the floppy disk boot sector, which on Linux is done by typing (as root):
cat etherboot.zdsk > /dev/fd0
Check the client machine BIOS and make sure it will try to boot from the floppy drive.
For the network boot process to work, two services will have to be setup: DHCP and TFTP, with the latter being the method for file access. Depending on your Linux distribution (sorry, this is possible with *BSD, other Unices etc., but I will only cover Linux here) install the packages for “
dhcp-server” (ISC version 3.0++) and tftp-hpa.
The TFTP service will in most cases be setup to serve files only from the directory
/tftpboot on the server, which is a sensible thing. You will have to move the client machine kernel and initial ramdisk into that directory, but that later.
** to be continued