Switching Fundamentals and Network Design
A switching expert must reason from the frame up: MAC learning builds the forwarding table, unknown unicast is flooded within a VLAN, and broadcast/multicast are flooded to all ports in the VLAN except the ingress. Know the difference between store-and-forward and cut-through, and how oversubscription and buffer behavior affect real throughput. Design questions test your ability to size uplinks, avoid single points of failure, and choose between collapsed-core and three-tier topologies.
Traps to watch: a switch never forwards a frame back out the ingress port; a MAC learned on one port moves to a new port immediately (no aging wait needed); and "wire speed" claims require bidirectional, all-port, minimum-frame analysis. Candidates also confuse collision domains (per port) with broadcast domains (per VLAN).
Rule of thumb: for line-rate non-blocking, total uplink bandwidth should meet or exceed the sum of edge-port bandwidth you expect to be simultaneously active. A 48-port 1G access switch feeding 10G uplinks is 4.8:1 oversubscribed at the edge but non-blocking if realistic utilization is under ~20%.
Worked example: 24 access ports at 1G, average 15% utilization = 3.6G aggregate; two 10G uplinks (20G, minus one for redundancy = 10G usable) comfortably cover it with headroom.