Rolling Out Linux - page 2
First let's look at how ordinary installation works and how it can be applied to a cluster
One common Linux installation method is from a CDROM. The CDROM is taken physically to each machine, and after booting from an install floppy the install proper occurs. The person doing the install selects various device drivers to match the machine, selects the timezone and a few other related things. Next, some packages are selected, copying from the CDROM and decompressing happens and soon the system is installed. Thanks to a lot of effort on the part of distribution makers this process can be fairly well streamlined. With experience and the right hardware it can be completed in around 30 minutes for a basic install.
An FTP install is similar, but the install time is very much affected by the bandwidth back to the server holding the package files. It is usually slower than a CD install. The advantage is, that you can usually get access to the very latest versions, which you may need to support hardware on the very newest machines.
Most distributions also support installation of packages from a remote server via NFS or SMB. These methods offer a speed improvement over FTP because the protocols are slightly more efficient. It is much easier to set up a local mirror with better network bandwidth.
Now, in the case of installation on several machines, two things are apparent. If a CDROM install is used then to install several machines at once, several CDROMs are needed. Also, all the machines in question must be equipped with CDROM drives. In the case of FTP or NFS installs sufficient network bandwidth must be available, or else the time taken will be excessive.