Over the decade of the 1990’s, the networking marketplace saw dramatic increases in desktop computing power. As application programs grew in complexity and sophistication, the need to send large quantities of data as quickly as possible grew proportionally. The shared-media environment forced all of these communicators to compete with each other for the use of the media. This proved to be an inadequate solution. To facilitate the demands of these increasingly complex networks, the industry experienced an evolution from shared media to switched network infrastructures. Today star-wired LANs using switches as the central connecting points are pervasive, creating large meshed network topologies.
While switched networks provide part of the solution for efficient use of the network media and infrastructure, they bring with them some inherent restrictions and limitations to the protocol analysis engineer. By their nature, switches do not forward all packets to all stations. Of course, broadcast and multicast packets continue to be forwarded out all ports of a switch and, therefore, reach all the stations in the broadcast domain. This is identical to the shared-media model. Directed frames, however, are forwarded in a much more intelligent manner. A “directed frame” is one with a specific Ethernet address as the destination target address. It is intended for only one recipient. The switch evaluates the Ethernet destination address on all incoming packets and forwards them only through the single port to which the intended target machine is attached.
As a result of this behavior, the network benefits from a reduction in contention for network bandwidth and a corresponding reduction in Ethernet collisions and the resulting retransmissions. This can easily be seen if one considers a simple topology in which a single switch has two file servers and sixty workstations attached to it. At the same time that Workstation #1 is sending a packet to File Server #1, it is possible for Workstation #2 to send a packet to File Server #2. Neither workstation is required to wait for the other, as would have been the case in the older shared-media networking model.