First off, why design a new protocol when there are so many well refined ones already in existence? I've never understood the need to re-invent the wheel unless you have a solid reason that requires something so completely different than what already exists. Take a look at existing options first.
Second, if you're going to have all the clients on a 10/100 port switch, your limit is 100mbps minus overhead per client. It doesn't matter what the switch they connect to is capable of, that is the fastest you'll get, period, but that should be no problem to pull 1GB in an hour. If you plan on having 48 clients pushing that much traffic all at once, the switch will require a gigabit connection at the very least, to whatever it talks to further upstream, and even then, you'll never handle all 48 clients at once on anything less than an enterprise grade switch. This cisco would work, but keep in mind the cost of this device is about $1400. It is also stackable so you can add more as you need more.
I've always liked the idea of fiber, but cables are expensive and cleaning and troubleshooting fiber problems is expensive too (standard tester and cleaning equipment is around $2000, more if you want anything other than simple db loss). If you don't know anything about how fiber works or how to clean a fiber ferrule, don't go with fiber, gigabit ethernet will be fine for most people's needs. A gigabit backbone to the server that will be holding the data will suffice.
What you really need to pay attention to is the ability for the server to read off its hard drives for 1GB/hour from 48 machines and the CPU power as well. Your limitations will be on the hardware of the server, not your networking gear.
((2**30*8)/(3600))*(1/2**20)*48
. A single 100Mib/s uplink is almost enough assuming traffic is constant.