The first Cisco switches did not have any Layer 3 routing capability. They merely switched frames at Layer 2 between hosts on the same VLAN. All inter-VLAN switching was forwarded to the router. The switch was many times faster in switching packets than the router. A router had greater latency when it had to forward packets at Layer 3. To get around this problem, as mentioned in earlier chapters, VLANs were extended throughout the LAN campus, minimizing the role of routers in the LAN network.
The actual connection between the switch and the router came in two forms. The first was one physical link per VLAN that was configured on the switch. The second method involved one physical cable between the switch and router with trunking enabled. The former was widely deployed because most LAN designs did not require trunking between a router and switch. After the packet arrived on a router's interface, the router did the Media Access Control (MAC) rewrite and software switched the packet out of one of its other interfaces. A popular router at that time, 4000 router, could only fast switch packets at a rate of 14,000 packets per second (pps). The enterprise router, 7500 family with an RSP2 card, could only do 150,000 pps. A Catalyst 5000 switch can switch packets at millions of pps (Mpps).
The multilayer switching mechanism enabled the use of Catalyst 5000 hardware to switch packets between different subnets in hardware, which translated into higher performance of packet handling on the network.
Newer hardware for Catalyst 6500 with Cisco Express Forwarding (CEF) implementation can switch packets at very high rates. For example, Supervisor II with a Switch Fabric Module (SFM) card can switch packets at rates as high as 210 Mpps. The Supervisor 720 can go as high as 400 Mpps. The discussion at the moment is with MLS and its role in Cisco switches.