In some of the older, more expensive servers, it didn't load the operating system straightaway. It had a ROM chip with a much simpler OS built into it, and that simpler OS would boot the more complicated, production one. This may have added complexity, but there was a good reason for it.
The good reason was that sometimes administrators did really stupid things. Sometimes they erased large, seemingly useless file that were actually critical to the OS's operation. Sometimes the disk screwed up. Sometimes the main OS didn't work, and the big mainframe didn't exactly come with a rescue CD. The "Boot Monitor" ROM OS may have been simpler than the production one, but it was good enough to, say, rescue files from a failing disk, copy them to the other computer in the next town over via the network, and load the replacement OS into place.
Desktop computers don't typically have this features because of the expense and added complexity this would add. But frankly, I think we should have USB-based boot monitors. If you mess up the computer, you can pop in the USB-Key device and it'll help make things all better. It can't save files that you forgot to save during the power outage, but it can rescue files from a failing disk, or help you find that note with your lost password. You know, that one you've asked helpdesk to help you find 21 times now.
I could also use it when putting together a new OS. I'm often left with this chicken-and-egg problem of being unable to make the filesystem because I need the writing program, but I can't build the writing program with no filesystem. Here, the writing program would be on the USB-key, the filesystem would be made, and then the OS installed on it. Easy peasy. Way better than my current method of using the swap-partition that'll get erased later as a temporary fake filesystem.
To non-computer people who asked me what in the flying hell filesystems and partitions are, they have to do with storing information on a hard drive. First you have to cut the hard drive into partitions, which decides what section gets what. Windows people can make the whole drive one partition if they like, but the advantage to having several is that they tend to fail one at a time. So if you have an OS partition and a Data partition, and the OS partition fails, you might need to reinstall but you still have your data. If the data partition fails, I hope you had a backup. If you had one partition, you're reinstalling and restoring from backup, which is a bit of a pain.
A filesystem is like installing a cabinet in the partition. It says what files are where, and must be in a form that the OS can understand. Newer ones can store more files in less space, thanks to research into block-sizes and whatnot. The filesystem determines that your letter to your boss is at, say, D:\letters\howaboutaraise.doc . It allows organization, so that your business letters can be separate from your letters to your mother. With no file system, any information would be like a big pile of papers on the ground. That get stepped on and mangled. No one could make sense of that.