The MAC of a recived frame will always be that of the last router. When a frame is received on "port a" of a router, the L2 frame is stripped and discarded. The router looks at the packet contained in the (discarded) frame and makes a routing decision. The router then re-encapsulates the packet in a new frame (with the transmission interface's MAC address) and sends it to the destination host on "port b".
A router connected to a switch will use a straight-through cable. Routers like PCs, are considered "end" systems. Cross-over cables are used to connect like-systems (PC-PC, Router-Router, switch-switch)
You should erase the NVRAM to clear the previous configuration so you can "start from scratch"
ARP is more of an L2 protocol than L3 (though it is not strictly in either layer). If you examine an ARP packet, you'll see that there's no IP (network) layer. ARPs are initiated as a local broadcast - since there is no L3 information, the ARP is restricted to it's broadcast domain.
NTP is UDP, NNTP is TCP (no, they're matched backwards)
DLCI is a layer two address. It's between the the local router and the Frame Switch. There will also be another DLCI between the far-end router and the switch, but it's not the same DLCI. DLCIs are "locally significant" they are only used for the specific like that they label.
Multicasting COULD cause congestion, but it generally thought of as a mechanism to reduce congestion. In a situation where 20 hosts are trying to view a six megabit video stream through a typical network, if all 20 used unicast, that would create 120 megabits of traffic. Using multicast, all 20 hosts "tune in" to the same multicast stream, using only six megabits.
10. True. Switches use (now, usually, 802.1q but older may use ISL) trunks. Routers use sub-interfaces that are enabled for 802.1q (or ISL) encapsulation.
You can have many / multiple loopbacks, but I don't remember the max number.
Good Luck on your test
Scott
