Scott: T oday, we're switching to "switching"--a new name for a transparent bridge with more than two ports.

Bill: Yeah, and next it's IP switching, a new name for a router. Dýjý vu, haven't we been here before?

Scott: People jumped on the switched-network bandwagon because of the lure of "low-cost per-port" switching--thinking it was "plug-and-play."

Bill: With the ol' "You can pay me now or you can pay me later," IP users are paying for believing the "no design needed" marketing story.

Scott: Let's look at the fundamental problem.

Bill: Right. A Data Link Control (DLC, Layer 2) switch, whether it be Source Route Bridging or Transparent Bridging, can't perform Layer 3 routing between two different IP subnets.

Scott: Therefore, if a source IP station and dest ination IP station are on different subnets, the packet must be sent to the default gateway (router), which in our client's case, was located on the backbone.

Bill: The router forwards the packets back onto the same wire and over the backbone to the appropriate destination...

Scott: ...which causes the traffic to double up on the backbone for all subnets that are switched on one side of the router.

Bill: Analysis of packet traces showed this to be precisely the case. The router's CPU also began to strain, as it took on this additional load of locally routed packets.

Scott: Our client's desire to assign a subnet for each Token-Ring segment meant that a zillion--OK, more like several dozen--secondary addresses must be assigned to the router port attached to the backbone.

Bill: Assigning one subn et per Token-Ring (or Ethernet for that matter) makes a lot of sense in a routed environment.

Scott: But this configuration does not make much sense in a switched environment, since all the segments attached to the switch now look like one big, logical or flat subnet.