Categories of network technology

WAN technologies and routing

Packet switches and store and forward
Hierarchical addresses, routing and routing
tables
Routing table computation
Example WAN technologies

Packet Switches
In order to grow, WANs use many switches
Basic component is the packet switch that
can connect to local computers and to
other packet switches

Local Area Network (LAN)

Metropolitan Area Network (MAN)
Wide Area network (WAN)
Key distinguishing feature is scale:
geographic distance AND
number of connected computers

WAN topology
Chosen to accommodate expected traffic
and to provide redundancy

Physical addressing in a WAN

Store and forward
Each switch receives packets, queues
them in its memory and then sends them
out when possible (i.e., when the
destination is available)
Many computers can send packets
simultaneously

A WAN defines a frame format and assigns

physical addresses to its computers
Hierarchical addressing, e.g.,
first part identifies packet switch
second part identifies computer on this switch
C [2,2]

[1,2] A

address

[1,5] B

D [2,6]
switch 1

switch 2

Next hop forwarding

A packet switch determines the destination
for each packet from the destination address
local computer or
another packet switch

Only has information about how to reach the

next switch - next hop
This is held in a routing table

Example routing table

[3,2]
C

[1,2]
A
S3

S1
B
[1,5]

D
[3,5]
S2
E
[2,1]

Routing table for S2

Destination Next Hop
[1,2] interface 1
[1,5] interface 1
[3,2] interface 4
[3,5] interface 4
[2,1] computer E
[2,6] computer F

F
[2,6]

Hierarchical addresses and routing

Source independence
The next hop depends upon the destination
address but not on the source address

Routing in a WAN
Large WANs use interior switches and exterior
switches
Their routing tables must guarantee that
Universal routing - there must be a next hop for
every possible destination
Optimal routes - the next hop must point to the
shortest path to the destination

Routing is the process of forwarding a

packet to its next hop
Hierarchical addresses simplify routing
smaller routing tables
more efficient look up

Default routes
A routing table may be simplified by including
(at most) one default route

Switch 1
Destn next hop
1
*
interface 3

Switch 2
Destn next hop
2
4
interface 4
*
interface 3

Switch 3
Destn next hop
3

interface 2

interface 3
4
interface 4

WANs and graphs

Routing table computation

Manual computation is not feasible for large
networks
Static routing - program computes and installs
routes when a switch boots; these routes do
not change.
Dynamic routing - program builds an initial
table when the switch boots; this alters as
conditions change.

A Wan can be modelled as a graph

Nodes are switches: 1, 2, 3, 4
Edges are connections between switches: (1,3) (2,3) (2,4) (3,4)
Weights are distance along a connection

Example of the shortest path

Computing the shortest path
Dijkstras algorithm - find the distance from
a source node to each other node in a
graph
Run this for each switch and create nexthop routing table as part of the process
Distance represented by weights on
edges

Dijkstras algorithm
S = set of nodes, labelled with current distance
from source, but for which minimum distance
is not yet known
Initialise S to everything but the source
Iterate:
select the node, u, from S that has smallest current
distance from source
examine each neighbour of u and if distance to
the neighbour from source is less via u than is
currently recorded in S then update

Implementation
Data structure to store information about the
graph (nodes and edges)
Integer labels for the nodes [1..n]
Three data structures:
current distance to each node - array - D[1..n]
next hop destination for each node array - R[1..n]
set S of remaining nodes - linked list - S

weight(i, j) function (infinity if no edge)

Given:
a graph with nonnegative weight assigned to each edge
and a designated source node
Compute:
the shortest distance from the source node to
each other node and a next-hop routing table
Method:
Initialise the set S to contain all nodes except
the source
Initialise D so that D[u] = weight(source, u)
Initialise R so that R[v] = v if there is an edge
from the source to v and zero otherwise

while ( set S is not empty ) {

choose a node u from S such that D[u] is minimum
if ( D[u] is infinity ) {
error: no path exists to nodes in S; quit
}
delete u from set S
for each node v such that (u, v) is an edge {
if v still in S {
c = D[u] + weight (u,v);
if c < D[v] {
R[v] = R[u]
D[v] = c;
}
}
}
}

Example
S = { 1, 2,
1

Distance D
2 3 4 5 6

3,
7

4, 5,

Example
S = source = 6
6,

7 }

Next hop route R

1 2 3 4 5 6 7

8X
5

Distributed route computation

Dijkstras algorithm requires each switch to
hold a complete description of the network
In distributed route computation
computation, each switch periodically
computes a new local table and sends it to
its neighbours
After a while, the switches learn shortest
routes or adapt to changes in the network

S = { 1, 2,

3,

4, 5,

Distance D
3 4 5 6

2
X
13

6,

7 }

Next hop route R

2 3 4 5 6 7

Given:
a local routing table, a weight for each link that connects
to another switch and an incoming routing message
Compute:
an updated routing table
Method:
maintain a distance field in each routing table entry
initialise routing table with a single entry that has the
destination equal to the local packet switch, the next hop
unused and the distance set to sero

Repeat forever {
wait for the next routing message to arrive over the network;
let the sender be switch N
for each entry in the message {
let V be the destination in the entry and D the distance
compute C as D plus the weight of the link over which
the message arrived
examine and update the local routing table {
if ( no route exists to V ) {
add an entry to the local table for V
with next hop N and distance C
} else if ( a route exists with next hop N ) {
replace existing distance with C
} else if (route exists with distance > C )
change next hop to N and distance to C
}
}
}

Link state routing

Each switch periodically broadcasts state of
specific links to other switches
Switches collect these messages and then
apply Dijkstras algorithm to their version of
the state of the network