It took another 10 years for scientists and engineers to figure out – and agree on – how to make Ethernet networks run 10 times faster. In 1995, the IEEE set the 802.3u standards for 100 Mbps "Fast Ethernets."It only took the IEEE three years to approve yet another tenfold speed increase to 1,000 Mbps – one gigabit – under the 802.3z specifications. The era of Gigablt Ethernets had arrived – well, mostly.In 1998, the IEEE approved the standards for fiber-optic and coaxial cables for Gigabit Ethernet networks. The institute, however, held off on certifying the 802.3ab gigabit specs for unshielded twisted-pair (UTP) copper wire. The IEEE scheduled approval of standards for l000BaseT lines for March 1999.Category 5 UTP wire is the installed base of a vast number of existing computer networks. The cost of rewiring those networks could be prohibitive. Thus, it is extremely important that Gigabit Ethernets be compatible with legacy 100Base and10Base networks as well as the operating systems and application software used on those nets.
Non-Switched and Switched Networks
A traditional 10Base or 100Base network can be loosely termed a "non-switched" network, because it incorporates a shared, "half-duplex" path where only one node can transmit at any given time. When many nodes attempt to transmit, the probability of packet collisions increases. To evade the degradation in performance caused by packet collisions, the network must be segmented. One segmentation technique is to replace a I0/100BaseT or coax cable hub with a central hub-switching matrix. This is very expensive.By having the segmented, half-duplex subsystems transmit independently, and then combining the multiple transmissions in a switching matrix, the various subsystems can communicate with each other with a reduced probability of collisions. This "collision evasive" system, however, is quite expensive, complex and requires elaborate software control. Even then, CSMA/CD collision switches employ a methodology that merely attempts to compensate for the shortcomings of CSMA/CD at high-speeds.In contrast, because the SMAC design is a "collision-less" protocol, it can easily be deployed at gigabit bandwidths on a non-switched network. And when SMAC technology is applied in a switched environment, the switch matrix becomes low-cost, elegantly simple, and easy to control with a user-friendly interface.
In order to understand why SMAC technology is superior to CSMA/CD, it is necessary to understand CSMA/CD.Carrier sense multiple access with collision detection is a MAC – medium access control – protocol. The function of a MAC is to enable nodes on the Ethernet to transmit information on a shared path without undue interference from other nodes.In simpler terms, a "node" is a computer or some other device attached to the network, such as a printer. A "shared path" is any transmission medium (for example, coaxial