I just recently transferred to the IT Department in my company and one of the tasks on my plate is to review their backup strategy and make sure it's "good". Although I have a lot of IT experience with previous employers, I've never really been involved with backup and restore much beyond the theory level.
They are performing nightly backups to a 400GB tape and a 1TB USB hard drive. The hard drive is backing up absolutely everything on all five servers. The tape is backing up just the essential stuff: web sites, databases, exchange, departmental shares, users' home folders. Both jobs start at 11:00 PM. The tape finishes around 8:30 AM; the hard drive finishes in the middle of the afternoon.
The selection lists were set up by a third-party company. The IT Manager is questioning whether it is really necessary to back up absolutely everything; for example, do we need to back up the WINNT folder? If a server dies, is the tape backup sufficient to bring it back to life, or do we need more?
Backups are not about having a copy somewhere else but being able to restore what's needed in case it's lost. You would only need to back up an operating system's files if you needed to restore that OS. It's certainly possible to only back up your data and plan on rebuilding the server as long as you are willing to accept the time required to rebuild the system and you have a way of keeping track of what patches you have put on the server so you can rebuild it at that point in time. In addition you have to accept the risk that any applications that are installed may have issues when being rebuilt. Once you've tried a restore this way you'll know if this approach can be successful for you
By your admission that the tapes only have "essential" data (which is in quotes because to me the essential data is what's required to make the service work- not just user data) your tapes do not contain what's required to rebuild a server.
There are several questions on serverfault about backup strategies but in a nutshell, you need to figure out what your time to recover the entire service is and work backwards from there before you can get meaningful answers to this question