Cisco LAN Switching Methods

All the switching methods used by Cisco switches provide increased throughput in comparison to bridges. The hardware-based architectures of switches allow them to make decisions at wire speeds. The primary difference between the various methods is the process each uses to switch frames .

Frames have been consistently used to represent Layer 2 data messages in this chapter. The term cells can also be used to identify Layer 2 data messages. This term is used when referring to data traffic using the Asynchronous Transfer Mode (ATM) technology. The CCNA exam does not cover this technology, so only frames have been presented in this chapter to simplify the concepts.

Store-and-Forward Switching

In store-and-forward switching mode, the switch reads the entire incoming frame and copies the frame into its buffers. After the frame has been completely read, the switch performs the Layer 2 cyclical redundancy check (CRC) to determine whether an error occurred during transmission. If the frame has an error, the switch drops the frame. If no error is identified, the switch checks its forwarding table to determine the proper port (in the case of a unicast) or ports (in the case of a multicast) to which the frame must be forwarded.

Store-and-forward switches have the highest latency of any switching mode, because the switch must read the entire frame before making a forwarding decision. The added error checking of store-and-forward switching, however, reduces the number of erred frames that are forwarded.

Cut-Through Switching

Cut-through switches introduce a lower level of latency during the switching process than store-and-forward switches do, mainly because the frame is forwarded as soon as the destination address and outgoing interface are determined. They achieve increased performance by eliminating the error checking and making forwarding decisions based only on the first six bytes of the incoming frame. (These first six bytes contain the destination MAC address of the frame.) Cut-through switches read the destination address of the incoming frame and immediately check the forwarding table to determine the proper destination ports. This increased performance does, however, allow erred frames to be forwarded more often than store-and-forward switches do.

Fragment-Free Switching

Fragment-free switching is a modification to the cut-through switching method. Like cut-through switches, fragment-free switches read only a portion of the frame before beginning the forwarding process. The difference is that fragment-free switches read the first 64 bytes, which is enough to check the frame for collisions. This allows for better error checking than with cut-through switches, with a minimal loss in latency.

Cisco has incorporated switching technology into devices operating at layers other than Layer 2. Be very careful when answering exam questions to determine how the term "switch" is used. The features and functions we are presenting in this chapter apply to Cisco layer two switches exclusively.