**Update** Just asked out IT guy about it and he said we don't use this kind of set up due to the cost and the trade off in processing power.
It doesn't have to cost that much or consume much processing power, TBH. Depends how you implement it. In some cases performance will be boosted because hard drives are the biggest bottleneck in a server and if you've got multiple drives with the same data on, you can read from them in parallel.
Hard drives are the one thing in a server that have a very finite life so it does make sense to build-in some fault tolerance. In the simplest case, you install 2 drives and "mirror" them. Anything that gets written to disc goes to both drives. You effectively maintain 2 drives with exactly the same data on. When one dies, the system switches to using the remaining working drive. You replace the failed drive and the system then rebuilds the "mirror" by copying everything onto the new drive. The server can even be up and running, albeit with reduced performance, during this whole process.
You can get hardware controllers that do all this for you, so it doesn't take any resources from the CPU of the server itself, or you can do it all in software, which means it's much cheaper but does use some resources of the CPU itself.
I really would look into implementing some kind of fault tolerance. A single hard drive failure is too common an occurrence for it to take your server down.
Kevin