Network Layer:

Delivery, Forwarding,
and Routing
Desirable properties of Routing Algorithms:
Correctness (applicable to all)
Simplicity (applicable to all)
Robustness: able to cope up with
 changes in topology, load.
 hardware and software failures
Stability (hard to achieve)
 Converge to equilibrium
Fairness (conflicting)
Optimality (conflicting) see next fig.
 Types of Routing Algorithms:
Non-Adaptive: Static. Routing decisions computed in advance, off-line and
downloaded.
Adaptive: Dynamic. Adaptive to the changes in
topology and load. Issue here is how to get the information? Locally,
From adjacent routers, from all routers?
22.6
The smaller the administrative distance value, the more reliable the
protocol. For example, if a router receives a route to a certain
network from both Open Shortest Path First (OSPF) (default
administrative distance - 110) and Interior Gateway Routing Protocol
(IGRP) (default administrative distance - 100), the router chooses
IGRP because IGRP is more reliable. This means the router adds the
IGRP version of the route to the routing table.
The best path selection process by dynamic protocol –
If a router receives the same advertised routes from more than one source for a
remote network, then
first the AD value is checked. The advertised route having the least AD value will get


preference.

If the AD value of the advertised routes is the same, then the metrics of advertised
routes are checked. The advertised route with the least metric will be placed in the
routing table.

If both AD and metric are the same then load balancing is done i.e the traffic will
traverse through different routes. The load balancing can be equal or unequal. In
equal load balancing, the same amount of traffic will
22.10
 Figure 22.12 Autonomous systems

22.12
 Figure 22.13 Popular routing protocols

22.14
 Distance Vector Routing:
 Distance Vector Routing Algorithm:
 At each step within a router:
 Get routing tables from neighbours
 Compute distance to neighbours
 Compute new routing table
1. Router transmits its distance vector to each of its neighbors.
2. Each router receives and saves the most recently received distance vector
from each of its neighbors.
3. A router recalculates its distance vector when:
a. It receives a distance vector from a neighbor containing different
information than before.
b. It discovers that a link to a neighbor has gone down (i.e., a topology
change).
The DV calculation is based on minimizing the cost to each destination.
The distance vector routing algorithm is sometimes called by other names, the
distributed Bellman-Ford routing algorithm and the Ford-Fulkerson
algorithm.
 Distance Vector Routing

(a) A subnet. (b) Input from A, I, H, K, and the new

routing table for J.
 Drawback of Distance Vector Routing:
 Slow Convergence-Count to Infinity Problem:

It reacts rapidly to good news,

But, leisurely to bad news.

Updates value fast when neighbor is down, but not
when neighbor is again up. How?

Lie to neighbour about distance
if routing via neighbour

The core of the problem is that when X tells Y that it
has a path somewhere, Y has no way of knowing
whether it itself is on the path? This is how problem
is created.
 It does not take bandwidth into account.
 Take too long to converge changes in one
node to all other nodes.
 Solution?
 Link State Routing.
 Figure 22.14 Distance vector routing tables

22.18
 Figure 22.15 Initialization of tables in distance vector routing

22.19
 Note

In distance vector routing, each node

shares its routing table with its
immediate neighbors periodically and
when there is a change.

22.22
 Link State Routing
Each router must do the following:
1. Discover its neighbors, learn their network address.

2. Measure the delay or cost to each of its neighbors.

3. Construct a packet telling all it has just learned.

4. Send this packet to all other routers.

5. Compute the shortest path to every other router.

A complete topology is developed. Then Dijkstra’s

Algorithm can be used to compute the shortest path.
Following 5 steps are followed to implement it.
1. Learning about the Neighbors

2. Measuring Line Cost.

3. Building Link State Packets.

4. Distributing the Link State Packets.

5. Computing the New Routes.

Learning about the
Neighbors

(a) Nine routers and a LAN. (b) A graph

model of (a).
 Step 1: Learning about the Neighbours:
 Upon boot of router,

Send HELLO packet on each point-to-point line

Routers are supposed to send reply with a globally unique
name
 Step 2: Measuring the Line Cost:
 Measure round-trip delay using ECHO Packet and
wait for its reply
 Take load into account? Yes. Arguments both ways:
when choice is given to router having same number
of hops from S to D.

Yes! preference for unloaded line as shortest path.

No! where oscillations are possible.
 Better Solution? Distribute Load over multiple lines.
 Enhanced IGRP (EIGRP)

22.26
What does it mean to be the shortest (or optimal)
route?
a. Minimize mean packet delay
b. Maximize the network throughput
c. Minimize the number of hops along the path
Dijkstra algorithm:
 Each node is labeled (in parentheses) with its distance
from the source node along the best known path.
 Initially, no paths are known, so all nodes are labeled
with infinity.
 As the algorithm proceeds and paths are found, the
labels may change, reflecting better paths.
 A label may be either tentative or permanent.
 Initially, all labels are tentative.
 When it is discovered that a label represents the
shortest possible path from the source to that node, it
is made permanent and never changed thereafter.
Now lets see the algorithm which describes this
procedure to develop shortest path from source to
destination.
 Figure 22.22 Dijkstra algorithm

22.29
 Link State Distance Vector
link states algorithm is an algorithm using global the distance vector algorithm is iterative, asynchronous,
information and distributed
each node talks with all other nodes, but tell them only each node talks to only its directly connected neighbors,
the cost of it's directly comparison of some of their but provides its neighbor with least cost estimates from
attribute itself to all the nodes.
Message complexity: With link state, every node has to Message complexity: with distance vector algorithm,
keep the information about the cost of each link within message is exchanged between two hosts which are
the network. directly connected to each other.
very times, if any of the link cost is changed, all the nodes change of cost in the link which is belong to the least cost
are updated. path for one of the nodes, the DV algorithm will update
the new value. But if the change doesn't belong to the
least cost part between 2 hosts, there will no updating.

Speed of convergence: can converge faster in comparison Speed of convergence: can converge slowly and have
of later. routing loops while the algorithm is converging.

Such probability is less. DV algorithm also suffers from the count to infinity
problem.
Robustness: For LS, when a router is down, it can Robustness: DV, the wrong least cost path can be passed
broadcast a wrong cost for the closest one. LS node is to more than one or all of the node so the wrong
computing for its own forwarding table and other node do calculation will be process in the entire net work. This
the calculation for themselves. Better than DV. problem of DV is much worse than LS algorithm.
 Interior G/W Protocols Comparison

22.31
 EGP: Path Vector Routing……..Border Gateway Protocol(BGP)

22.32
 Figure 22.31 Stabilized tables for three autonomous systems

22.33
 Figure 22.32 Internal and external BGP sessions

22.34