 |
 |
 |
 |
|
|
|
|
Ethernet began as a research project at Xerox PARC by Robert Metcalfe in the late 1970's. Originally, it operated at 3 Mbps (megabits per second). Now, it is most commonly operated at 10 Mbps. The fundamental problem was how to supply a shared medium of communication to a set of computers. If one attempts to directly connect 100 computers to each other, one gets 99 wires hanging out of the back of each computer and a total of 9900 wires: clearly not viable.
Ideally, one might attempt to do without wires altogether which is exactly what was done at the University of Hawaii. Their "Aloha" system was used to link computers on different islands via radio signals on a shared frequency. Due to the distances between the nodes, it was not possible to know if one of the other nodes had begun a transmitting a data frame when it was time to transmit one's own data frame. However, when two frames overlapped in time, they interfered with each other and were garbled. This kind of event is called a "collision" and could be detected after the fact, a processes called "collision detection." When a collision was detected, both nodes would need to retransmit and hope to avoid the collision on the next try. Of course, by then, some third node may be transmitting.
At Xerox, they were working with coaxial (wires with a common axis, similar to what is now used for cable television) cable instead of radio, and realized that they could detect transmissions in progress and prevent transmitting when this was the case. Doing so, however, requires limiting the distances involved to limit the amount of time it will take the signal to propagate down the cable. This technique, called "carrier sensing," reduced the collision rate (the percentage of the time a collision occurs) significantly. Collisions result in a loss of efficiency, so this can become quite important on a busy network.
The length of the cable is also limited by how far an analog signal (all signals are fundamentally analog signals, we just choose to interpret some of them as bit patterns) will remain recognizable as a sequence of bits. To preserve the signal integrity, a thick coaxial cable was used with limited flexibility and a solid metal core. One needed to "tap" into the core and connect to the shield wire to connect a computer to the cable. Taps could be placed only at intervals of 5 meters, the cable was marked at these locations to prevent random signal reflection at the tap points. A "transceiver" (transmitter/receiver unit) was connected to the tap to isolate ether from electrical noise from the computer. Today, we call this 10BASE-5 or "thicknet."
The transceivers need to be located at the tap point on the
cable which is buried in the floor or hidden in the walls of the building. "Transceiver
cable" is used to connect the transceiver to a computer. This cable is also
thick and well-shielded, but is a bundle of separate wires not a single coaxial
pair. It is easy to mistake the transceiver cable for thicknet wiring.
 |  |
Figure 1: Thicknet (orange cable) with a tap (black unit in front of cable), a transceiver (black unit in back of cable), and transceiver cable (gray cable plugged on right side and looping over top). The photo to the left shows the unit in operation in 1994, the photo to the left shows it out of service in 1996 when the coaxial cable was replaced with fiber optics. The electrical outlet and plug above the unit in the second photo provides a sense of the size of the unit.
Eventually, the thicknet wiring dwindled popularity and was
replaced by other forms of media all using the same mechanism for carrier
sensing and collision detection.
|
|
|
|