They said it couldn't be doneââ,¬"running Gigabit Ethernet over standard Category 5 twisted-pair cables. Well, it has been done, and many people are excited about the possibilities of running gigabit speeds over their existing wiring. The question at hand is, at what point is it appropriate? The answer seems to be, in the server room, or when the existing cabling plant can support it.
The exciting thing about gigabit over copper is the price. Both switch ports and network interface cards (NIC) are less expensive, because they don't need the expensive optical transceiver to convert electrical signals to optical and back. If existing wiring can be used, that also may offer substantial cost savings. However, as with many other items in the IT industry, acquisition cost does not tell the whole story. Nevertheless, interest in Gigabit Ethernet is picking up. Market watcher Dell'Oro Group estimates that the Gigabit Ethernet segment will nearly double by this time next year, reaching about US$4 billion. Not exactly chump changeââ,¬"and that's only for the switches themselves.
Time to terminate?
The disappointing thing is that getting actual gigabit speeds may require re-termination, if not new cabling. Network administrators also may find they need to upgrade patch panels and wall jacks, not to mention adding new patch cords, before they can get full speeds. The good news here is that it will take an infrastructure expert (like you) to make the call.
Buying a server based on price is not the best way to goââ,¬"it's a crucial part of the network and any cost savings can be quickly negated by higher administrative costs, down time, etc. Likewise, no one would buy a server and not put a good UPS on it (we hope). Anyone looking to run lots of Gigabit Ethernet should be looking at costs from the same perspective.
Running gigabit to the desktop for a normal LAN environment would be pretty silly. Even if costs were not a couple of orders of magnitude higher than for switched 100Mbps ports, the stringent requirements for the wiring would increase costs, even if just to test the wiring to ensure it'll handle the bandwidth. When you add the fact that a normal desktop running Windows 98 will never see even 500Mbps from a gigabit adapter, the whole idea doesn't make much sense.
