SMC Networks SMC6608T, SMC6608M manual Switch Architecture, Switching Methods

ABOUT THE TIGERSWITCH 100

Switch Architecture

Switching Methods

Adaptive Cut-Through

SMC’s TigerSwitch 100 models employ a dynamic forwarding architecture that enables them to support adaptive cut-through frame forwarding mode. This approach allows the switch to automatically alternate between three different switching methods to provide the best possible performance in response to current network conditions.

Cut-Through

Also known as “on-the-fly” switching, the cut-through switching technique involves forwarding each data packet as soon as the

6 byte destination address in the header has been read. It does not wait until the entire packet has been received. This method, which takes up the least processing time, reduces the latency of each packet to twenty microseconds (20 µsec) or less. Cut-through switching mode is the default setting of each port on the TigerSwitch 100.

Fragment-Free

This switching method is similar to cut-through, but it requires that the first 64 bytes of the data packet be received before it is forwarded. This enables the switch to discard runt packets - smaller than legal size packets which are collision byproducts - and effectively clean up the data stream. This method provides the greatest benefit when the collision rate is high (e.g., when the switch is used to interconnect several shared segments, each having a large number of end stations.)

Store-and-Forward

In store-and-forward switching mode, the entire packet must be received into a buffer and checked for validity before being forwarded. This prevents errors from being propagated throughout the network.

1-10