I had a lively exchange with another consultant today that included a discussion of imaging client machines. This other fellow subscribed to what I think is an old school philosophy that there should be a separate volume for the system and the data on client machines. I still subscribe to this view in regard to servers – but in that case we are usually talking about different disk arrays (i.e. a two-drive RAID 1 container and a N-Drive RAID 5 or RAID 10 container)

So his primary argument in favor of a separate system volume (or partition in his case as he is an adherent to FAT32 – which is a whole other can of worms) was that he can keep a “ghost” image on the data partition and if there is ever a problem with the system partition or if the OS won’t boot, he can walk the end-user through booting to DOS and running a script to re-image machine. Now I like the creativity of this solution, but here are my reservations:

  1. Using FAT32 is probably causing as many blue screens as it solves. Because it’s not a journaling file system, you are more likely to run into problems whenever Windows is ungracefully shut down.
  2. Having users self image a machine risks overwriting data that might have been unknowingly saved to the system partition.
  3. It can be a waste of disk space if you allocate too much space for the system partition, but you risk filling it up and bringing the system down if you allocate too little.

I think it’s simplest to just keep one partition:

  1. You don’t need to worry about resizing partitions or wasting disk space.
  2. It simplifies administration (no need to document or train admins on this aspect of a client build).
  3. You can save a base “ghost” image to the fileserver or on a USB drive somewhere. If you do want to risk having the end-user re-image their machine, you can have them first backup their current image to USB.

Of course client data backups are a must..