Routing Algorithms

Suggested Reading:
Bertsekas and Gallager: 5.1 (5.1.1, 5.1.2), 5.2.3 (on Bellman-Ford and
Dijkstras)

Kurose and Rose (2001 Edition): 4.1, 4.2, 4.3, 4.4

Also http://www.stanford.edu/class/cs224a: routing protocol

Outline

Flooding
Distributed Bellman Ford Algorithm
Dijkstras Shortest Path First Algorithm

Problem
Given a topology, link costs, and a source-destination (SD) pair,
determine a route from S to D so that the route has the minimum
cost (i.e., is the shortest).

Example network
The shortest route A to B:
R1, R2, R5, R8

R1

R2

R4

R3

R6

R5

R7

R8

Routing Metrics

Metrics
Link state: up or down (stability)
Delay to send an average size packet (Make high speed links
attractive, but closeness counts)
Bandwidth
Link utilization

For simplicity, our examples assume that the cost for the path is
additive.

Example network
In this simple case, solution is clear from inspection
A

R1

R2

R4

R3

R6

R5

R7

R8

Flooding
Routers forward packets to all ports
except the ingress port.
R1
Advantages:
Simple.
Every destination in the network is reachable.
Disadvantages:
Some routers receive a packet multiple times.
Packets can go round in loops forever.
Inefficient.

Spanning Trees
Find the lowest cost route from each of
(R1, , R7) to R8 (the same as from R8 to the rest for this case)

R1

R2

R4

R3

R6

R5

R7

R8

A Spanning Tree
1

R1

R2

R4

R3

R5

R6

R7

2
3

R8

The solution is a spanning tree with R8 as the source of the tree.

Tree: There are no loops.
Spanning: All nodes included.
Will see two algorithms that build spanning trees automatically:

The distributed Bellman-Ford algorithm

Dijkstras shortest path first algorithm

Bellman Ford Algorithm*

Finds the shortest paths, from a given source node, say node 8, to all other
nodes (i is the index for destinations).
General idea:
First find the shortest single arc path,
Then the shortest path of at most two arcs, etc.
Let dij be the cost between node i to j; and dij= if (i,j) is not directly linked

Let Di(h) be the shortest distance from 8 to i using at most h hops/arcs. (h

can be regarded as time also.)
Di(1) = d8i for i8, D8(h) = 0 for all h.
Di(h+1) = min {j} [Dj(h) +dji] for i8.

If all weights are positive, algorithm terminates in N-1 steps, where N is the
no. of nodes in the network.
*Notations are similar to those used in Bertsekas and Gallager.

Bellman-Ford Algorithm
R1

Example

R3

R1

R2

R4

R3

R5

R6
2

R8

R7

R2
4

R4

R5

R7

R6
2
3

R8

Bellman-Ford Algorithm
6

R1

R2

R5

R3

R4

6
3

R6

R7

R8

Solution
5

R1

R2

2
4

R3

R4

R5

R6

R7 3

2
R8

Routing Table for R8

Destination

R1

R5

R2

R5

R3

R3

R4

R5

R5

R5

R6

R6

R7

R7

Next Hop

Distributed Bellman-Ford
Let N(i) be a set of neighbors of node i:
Di(h+1) = min {j in N(i)} [Dj(h) +dji] for i8; D8(h+1) = 0
Only need to keep/eachange distance information from/with neighbors.

Bellman-Ford Algorithm
Questions:
1.
2.

3.

How long can the algorithm take to run?

How do we know that the algorithm always
converges?
What happens when link costs change, or when
routers/links fail?

A Problem with Bellman-Ford

Bad news travels slowly
R1

R2

R3

R4

Consider the calculation of distances from/to R4:

Time
0
1
2
3

R3
R1
R2
3,R2
2,R3
1, R4
3,R2
2,R3
3,R2
3,R2
4,R3
3,R2
5,R2
4,R3
5,R2
Counting
to

infinity

R3
d34=inf.

R4 fails

Counting to Infinity Problem

Solutions
1.

2.

3.

Set infinity = some small integer (e.g. 16). Stop

when count = 16.
Split Horizon: Because R2 received lowest cost path
from R3, it does not advertise cost to R3
There are many problems with (and fixes for) the
Bellman-Ford algorithm.

Dijkstras Shortest Path First Algorithm

Routers send out update messages whenever the state of a

link changes. Hence the name: Link State algorithm.
Each router calculates lowest cost path to all others, starting
from itself.
At each step of the algorithm, router adds the next shortest
(i.e. lowest-cost) path to the tree.
Finds spanning tree routed on source router.

Dijkstras Shortest Path First Algorithm

Example
Step 1:

Shortest path set, S = {R8}. Candidate set, C = {R3 , R5 , R7 , R6 }

Step 2:

S = {R8 , R5},

C = {R3 , R7 , R6 , R2 }.
Step 3:

S = {R8 , R5 , R6},

C = {R3 , R7 , R2 , R4 }.
Step 4:

R5

R8
R6

R5

R8

S = {R8 , R5 , R6 , R7 },

C = {R3 , R2 , R4}.

R6
R5

R7
R8

Dijkstras SPF Algorithm

Step 8 : S = {R8 , R5 , R6 , R7 , R2 , R1 , R4 },
C = {}.

R1

R2

R4

R6
2

R3

R5

R7

3
R8