Networking

The compromise worked well for 10 Mbps, but it was a problem for higher data-rate Ethernet developers. Fast Ethernet was required to provide backward compatibility with earlier Ethernet networks, including the existing IEEE 802.3 frame format and error-detection procedures, plus all applications and networking software running on the
10-Mbps networks.

Although signal propagation velocity is essentially constant for all transmission rates, the time required to transmit a frame is inversely related to the transmission rate. At 100 Mbps, a minimum-length frame can be transmitted in approximately one-tenth of the defined slot time, and any collision that occurred during the transmission would not likely be detected by the transmitting stations. This, in turn, meant that the maximum network diameters specified for 10-Mbps networks could not be used for 100-Mbps networks. The solution for Fast Ethernet was to reduce the maximum network diameter by approximately a factor of 10 (to a little more than 200 meters).

The same problem also arose during specification development for Gigabit Ethernet, but decreasing network diameters by another factor of 10 (to approximately 20 meters) for 1000-Mbps operation was simply not practical. This time, the developers elected to maintain approximately the same maximum collision domain diameters as 100-Mbps networks and to increase the apparent minimum frame size by adding a variable-length nondata extension field to frames that are shorter than the minimum length (the extension field is removed during frame reception).

Figure 7-7 shows the MAC frame format with the gigabit extension field, and Table 7-1 shows the effect of the trade-off between the transmission data rate and the minimum frame size for 10-Mbps, 100-Mbps, and 1000-Mbps Ethernet.

Figure 7-7 MAC Frame with Gigabit Carrier Extension