1. WAN Technologies

2. Large Spans and Wide Area Networks
MAN networks: Have not been commercially successful
Consider a company that uses a satellite bridge to connect LANs at two sites (one PC and one printer at each site). Should the network be classified as a WAN or as an extended LAN? 2 2

3. Large Spans and Wide Area Networks
The key issue that separates WAN technologies from LAN technologies is scalability. A WAN must be able to
Grow as needed to connect many sites
Deliver reasonable performance for a large scale network
Thus, a satellite bridge that connects a pair of PCs and printers is merely an extended LAN 3 3

4. Forming a WAN
A packet switch provides local connections for computers at the site and connections to other sites.
A small computer system with a processor, memory, and I/O devices used to send and receive packets.
Earlier than LAN technologies and the switch used in LAN

5. Forming a WAN
WANs use leased data circuits from phone companies

6. Modern WAN (Internet) Architecture
Most WANs separate a packet switch into two parts:
A Layer 2 switch that connects local computers
Local connections
A router that connects to other sites
Global connections
Communication with local computers can be separated from transmission across a WAN. 6 6

7. WAN Store and Forward Paradigm
To allow as many computers as possible to send packets simultaneously, WANs can store and forward packets

8. WAN Store and Forward Paradigm
In order to correctly forward the packets, each packet switch needs to know the MAC address of every computer on the network
Not feasible. Billions of devices are connected to the Internet today  need better routing mechanism

9. Addressing in a WAN
WANs addresses follow a hierarchical addressing mechanism:
(site, computer at the site)
In practice, instead of identifying a site, each packet switch is assigned a unique number
First part of an address identifies a packet switch
Second part identifies a specific computer 9 9

10. Addressing in a WAN
When a packet arrives, a packet switch P examines the destination address in the packet and extracts the packet switch number Q of the targeted computer.
Packets targeted for a local computer (P = Q)
The switch sends the packet directly to the computer.
Otherwise (P ≠ Q)
The packet is forwarded over to another switch
Using only one part of a two-part hierarchical address  Much faster forwarding 10 10

11. Next-Hop Forwarding More generally, a WAN can be modeled as a graph.
Packet switches as nodes and connections between them as links.
(k, j): a link from node k to node j
Why a graph?
Abstraction
An old theory with rich algorithms (e.g., computing the shortest path) 11 11

12. Next-Hop Forwarding Forwarding table
A packet switch only needs to compute the next hop for a packet.
Forwarding table 12 12

13. Forwarding Table Computation
Forwarding table must guarantee the following:
Universal communication:
Contain a valid next-hop route for each possible destination address
Optimal routes
Point to the shortest path to the destination
Example: Computing the forwarding table for node 3:
100m
500m
300m
100m
700m
300m
(3,4) 13 13

14. Exercise 1 2 3 4 5 Computing the forwarding table for node 1:
In a network with millions or even billions of computers, efficient algorithms are needed
After class reading: Section 18.13

15. Dynamic Routing Updates in a WAN
Network changes/failures further complicate forwarding
A network manager cannot merely configure a forwarding table to contain static values that do not change
Most WANs use dynamic routing
A program builds an initial forwarding table when a switch boots
Program then alters the table as conditions in the network change
Question: How can a switch know the network condition of some location 2000 miles away? 15 15

16. Distributed Route Computation
In practice, WANs need to perform distributed route computation
Each packet switch must compute its own forwarding table locally
All packet switches must participate in distributed route computation
Example: Link-State Routing (LSR)
Each packet switches periodically broadcast messages that carry the status of a link between two packet switches
E.g., switch 3 says: “the link between 3 and 5 is up”
Each switch collects incoming status messages and build a graph of the network
Each switch compute its own forwarding table based on the graph 16 16

17. One More Problem
Suppose computer [1,2] wants to send a message to [3,5].
From previous chapters we know that NIC of computer [3,5] uses its MAC address (say, 78:DF:4C:84:BB:F2) to do addressing filtering.
But if we use the new hierarchical address [3,5] to forward the packet, the packet would fail the addressing filtering.

18. We Need Both Address [3,5] [1,2] I like you Layering: Compose a frame
Layer 3 (hierarchical) address: for routing
Layer 2 (MAC) address: for address filtering
Compose a frame
54:DD:91
11:CA:3B
E8:25:32
46:C2:8F
[3,5]
[1,2]
I like you
Layer 3 header
Layer 3 payload
Layer 2 header
Layer 2 header

19. From LAN/WAN to Internet
When there are several LANs or WANs wanting to connect with each other, we will get an internetwork.
The largest internetwork is the Internet.
Router: A hardware system dedicated to interconnecting networks

20. From LAN/WAN to Internet
Problem: there may be several computer [1,1] on the Internet, one for each WAN
Solution: for every router and computer on the internet, we need a unique hierarchical address for routing
Next class: IP Address