1. Network Redundancy Multiple paths may exist between systems. Redundancy is not a requirement of a packet switching network. Redundancy was part of the design for the ARPANET. Today’s Internet core is highly redundant. Redundancy in any network will cost more.

2. The Routing Decision, Part I If the network destination of a packet matches one of your connected networks, use the data link layer to deliver the packet to the destination host. If the destination is on a different network, send the packet to a router (using the data link). If you have only one default route, use that router.

3. The Routing Decision, Part II But what if there is a choice of routers for a packet leaving a network node? (Then that node is equivalent to a router, even if it is a host computer) The network node (computer or router) must select the best router (the “next hop”) for the destination network of the packet.

4. Network Matching In classful networks, a network mask wasn’t needed. The host 192.168.1.2 and the host 192.168.1.5 are on the same network (send packet to destination, using the data link) The host 192.168.1.3 and the host 192.168.2.15 are not on the same network. (send packet to a router)

5. Subnet masks IP Subnetting defined in 1984 Useful in large, bridged class B nets and to conserve IP addresses Expressed as dotted decimal or /nn notation 128.186.121.1 /24 is the same as 128.186.121.1 255.255.255.0 192.168.1.1 /28 is the same as 192.168.1.1 255.255.255.240

6. Network Matching Use the network mask Bit-wise compare the two networks, using the mask length 192.168.1.2 /28 and 192.168.1.250 /28 are on different networks (or different subnets, if you prefer that terminology). 192.168.1.2 /28 and 192.168.1.3 /28 are on the same subnet. (use the data link)

7. Network Masks Note that network masks work for classful network matching also. 128.186.234.2 /16 and 128.187.123.4 /16 must be on different networks when the first 16 bits of the addresses are compared. Modern IP routing tables include a mask for every network route.

8. Host Addresses per Network The lowest address refers to the network itself and can’t be used for a host. Eg. 192.168.1.0 /24 The highest address is reserved for the IP broadcast address. Eg. 192.168.1.255 /24 So in this network there are 254 available IP addresses for hosts and routers.

9. Host Addresses in a Subnet The address 192.168.1.16 /28 refers to the network and can’t be used for a host. The address 192.168.1.31 /28 is reserved for the IP broadcast address. This network has 14 IP addresses available for hosts and routers.

10. Using the Data Link Layer Ethernet uses Data Link Layer addresses in frames for host addressing. Each network interface card has a 6 byte MAC (Medium Access Control) address; the first 3 bytes specify manufacturer, last 3 are unique for each card by that company. The special FF:FF:FF:FF:FF:FF address is used as the broadcast address.

11. Address Resolution Protocol ARP is used to discover the MAC address for an IP address A broadcast frame is prepared, asking “What is the MAC address for this IP address?” The reply is unicast back to the sender Can be used for other protocols than IP

12. Proxy ARP A router replies to an ARP request for IP address X with the router’s MAC address, even though X is not an address on any router interface.

13. Ethernet Bridges The bridge “learns” by remembering on which interface a given MAC has been seen as a SOURCE MAC address in a frame. The bridge examines the DESTINATION MAC in every frame, forwards if it knows the destination address, or floods the frame out every other interface if the DESTINATION is unknown.

14. Switches and VLANS Ethernet bridging is the basis for modern Ethernet switches, which typically have many ports. Frames are forwarded only when necessary, but broadcasts are always forwarded. VLANs prevent broadcasts (or any other traffic) from forwarding between VLANs.