Content area
Full Text
Token-ring switching presents some interesting technical challenges not present with Ethernet.
One promise of switches is to remove some of the latency associated with bridges and routers. A problem with token ring is that a sender must wait for a free token before sending the frame.
Once the frame is received, the switch must determine the forwarding port based on the destination address at the beginning of the frame (following the frame control and access control fields), then wait for a token on the destination ring. Thus, the token-passing mechanism adds latency.
The worst case scenario is a single station sending data from the source ring, because it must wait for a token for every frame, even though no other stations are contending for the token. Now imagine this in a network with a hierarchy or cascade of switches in a pure token-ring setting.
Token latency can add up, in a way similar to the effect of bridging in today's token-ring environments.
Several approaches to implementing token-ring switches help alleviate the effect of latency. These include cut-through operation, full-duplex operation or using a technology other than token ring to connect switches in a hierarchical (tiered) or cascaded network.
CUT-THROUGH VS. STORE AND FORWARD
The approach by vendors such as IBM and Madge Networks, Inc. is to minimize latency by operating the switch as close to cut-through as possible.
If the target LAN speed is the same as the...