What is wrong with this design? (After all, an attempt has been made to
include some resilience, with two border routers and HSRP)

1. Whole network is one large broadcast domain. Broadcasts packets
   are seen everywhere. (Windows NT servers are worst offenders)

2. All switches must learn all MAC addresses

3. Security: customer hosted servers and office machines can break your
   mail and web networks (by configuring a wrong IP address, ARP spoofing
   etc)

4. There's no such thing as a "layer 2 traceroute", so any network problems
   are very hard to locate

5. A broadcast storm in one part of the network will affect the whole
   network

6. Top switch/hub is a single point of failure. Reboot it and your
   whole network stops working for a while!

7. Switches form a tree. There are no backup links.

8. All traffic aggregates at the central switch.

9. What happens if you need to add more ports on the border routers, and
   you have run out of slots?