Computer

Networks

ROUTING ALGORITHMS

April 20, 2025 1

Routing Algorithms
 Main function of Network Layer:
 Routing of packets form the source machine to the
destination machine.

April 20, 2025 Veton Këpuska 2

Routing Algorithms
 Routing Algorithm:
 Network Layer Software responsible for
deciding which output line an incoming
packed should be transmitted on.
 Datagrams:
 require computation of decision making tables for each
packed
 Virtual Circuit:
 routing decision are made only when a new virtual circuit is
being set up.
 Session Routing:
 data packets follow the same routing for the entire session.

April 20, 2025 Veton Këpuska 3

  Routing vs. Forwarding:
 Routing:
 Filling and Updating routing tables
 Forwarding:
 making the decision which routes to use based on routing
tables.
 Adaptive vs. Non-Adaptive Algorithms.
 Non-Adaptive Algorithms:
 Routing decision is based on pre-computed measurements
or estimates and do not update the table based on current
traffic and topology
 Adaptive Algorithms:
 Change their routing decisions to reflect changes in the
topology and traffic.

April 20, 2025 4

Optimality Principle
 If router J is on the optimal path from
router I to router K, then the optimal
path form J to K also falls along the
same (optimal path) route.

April 20, 2025 5

Shortest Path Routing
 Optimization criterion:
 Distance,
 Bandwidth,
 Average Traffic
 Communication cost,
 Mean Queue Length,
 Measured Delay, …
 Algorithms:
 Dijkstra
 Flooding
 Selective Flooding

April 20, 2025 6

Distance Vector Routing
 Static Routing Algorithms
 Do not take into account actual network load.
 Dynamic Routing Algorithms
 Taking into account actual network load
 Distance Vector Routing: Each router maintain a
table with the best known distance to each
destination and which line to use to get there.
Tables updated by exchanging information with
the neighbors.
 Link State Routing

April 20, 2025 7

Count–to-Infinity Problem
 Slow Convergence to the correct
answer.
 “Good news” Propagate fast
 “Bad news” Propagate slowly:
 The core of the problem is that when X
tells Y that I has a path somewhere, Y
has no way of knowing whether it itself is
on the path.

April 20, 2025 8

Link State Routing
 Distance Vector Routing was used in the
ARPANET until 1979 – when it was replaced
by link state routing.
 Delay Metric was Queue Length thus did not
take into account line bandwidth when choosing
routes.
1. Problem when line bandwidth changed for
some bands from 56 kbps to 230 kbps or
1.544 Mbps.
2. Algorithm took to long to converge (the
count-to-infinity problem).
 Solution: Link State Routing

April 20, 2025 9

Link State Routing (2)
 Each router must do the following:

1. Discover its neighbors and learn their network

addresses.
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.

 Complete topology and all delays are experimentally

measured and distributed to every router. Dijkstra’s
algorithm can be run to find the shortest path to
every other router.

April 20, 2025 10

(1) Learning About the Neighbors
 “HELLO” packed send on each point-
to-point line from a booted router.
 Router on the other end must reply by
sending its globally unique “name”.
 Example of routers connected by a
LAN.

April 20, 2025 11

Learning About the Neighbors (2)
 Nine routers and a LAN:
E
D
F Router G H
B
G C
B D
E I A
F I
A C F

Routers connected to LAN Graph Model

 One way to model LAN is to consider

it as node (Graph model).

April 20, 2025 12

(2) Measuring Line Cost
 It is required by the Link State Routing algorithm that
each router should have a reasonable estimate of the
delay/cost to each of its neighbors.
 Send “ECHO” packet (ping) that the other side is
required to send back immediately.
 Measure Round Trip time; Divide by 2 to get an
estimate.
 More accurate estimate by repeating the process
several times and by averaging estimates.
 Assumes symmetric delay.
 Channel Load Issue when Measuring Delay
 To factor the load in: round trip timer must be started
when the ECHO packed is queued.
 To ignore the load: round trip timer must be started
when ECHO packed reaches front of the queue.

April 20, 2025 13

(2) Measuring Line Cost
(cont.)
 Including Traffic-induced Delays:
 If a router has a choice from 2 lines with
the same bandwidth, one of which is
heavily loaded all the time and one of
which is not, the router will regard the
route over the unloaded line as shorter
path. This choice in general will result in
better performance
 Problem with Oscillations in the choice of
best path.

April 20, 2025 14

(2) Measuring Line Cost
(cont.)
West East
G
C
B
F
H
A

D I

April 20, 2025 15

(3) Building Link State Packets
 Packet Format: A B C
 Identity of Sender Seq. Seq. Seq.
 Sequence Number
 Age Age Age
Age
 List of Neighbors B 4 A 4 B 2
 Corresponding Delay
E 5 C 2 D 3

F 6 E 1

B 2 C

4 3
D D E F

A 6 Seq. Seq. Seq.

1 Age Age Age

5 4 C 3 A 5 B 6

F 4 C 1 D 4

E F
Packets easily built –8problem with knowing when F 8 E 8
to built them.

April 20, 2025 16

(4) Distributing the Link State Packets
 Distributing Link State Packets Reliably is tricky:
 As the packets are distributed and installed, the routers getting
the first ones will change their routes before other routers in
the network update their routing tables.
 Different Routers may be using different versions of the
topology (inconsistencies, loops, unreachable machines, etc.)
 Basic Algorithm: Flooding
 Sequence Number (incremented for each new packet sent) is
used to keep the flood in check.
 Routers keep track of all the source router packets they have
been sent to.
 New link state packets is checked against the track list:
 If new/unseen (based on the sequence number) then it is
broadcasted to all neighboring routers with exception of the
sender.
 If duplicate, it is disregarded.
 If sequence number is lower than the highest one in the track
list, it is rejected.

April 20, 2025 17

(4) Distributing the Link State Packets (cont)
 Problems with basic algorithm:
1. Sequence Number wrap around.
 Make a long precision number (e.g., 32-bit)
2. Crash of a router: losing track of sequence number.
3. Corruption of sequence number.
 Solution: Include Age of each packet.
 Decrement this value once per second.
 When zero, this state information is disregarded.

 Normally a new packed is send every 10 sec.

 Router information times out when:
 Router is down, or
 A Number of (e.g., 6) consecutive packets have been lost.

April 20, 2025 18

(4) Distributing the Link State Packets (cont)
 Refinements of Distribution Algorithm:
 Holding the packet:
 Example of packed buffer for router B of subnet in
previous figure (Fig. 5-13 (a))
Acknowle
dged
Sent Flags Flags
Source Seq. Age A C F A C F Data
A 21 60 0 1 1 1 0 0
F 21 60 1 1 0 0 0 1
E 21 59 0 1 0 1 0 1
C 20 60 1 0 1 0 1 0
D 21 59 1 0 0 0 1 1

April 20, 2025 19

Hierarchical Routing
 Large Networks:
 Proportionally large routing tables are required
for each router
 More CPU time is needed to scan them
 More bandwidth is needed to send status
reports.
 At certain point network may grow so large
where it is no longer feasible for every router to
have an entry for every other router.
 Solution: Routing has to be done hierarchically.

April 20, 2025 20

Hierarchical Routing (cont.)
 Routers divided in Regions (as in telephone network):
 Each router knows how to route packets to destinations within
its own region.
 However, router does not have any information regarding the
topology of the network of other regions.
 When different networks are interconnected they are
regarded as a separate region in order to free the routers in
one network form having to know the topological structure
of the other ones.
 Huge networks will require more than two-level hierarchy.
 How many hierarchical levels are optimal.
 Kamoun and Kleinrock (1979): optimal number for an N router
subnet is ln(N), requiring total of e*ln(N) entries per router.

April 20, 2025 21

Hierarchical Routing (cont.)
 Example:
Berkley, California Router to
Malindi, Kenya.
 Berkley to Los Angeles router (out-of-
state traffic)
 Los Angeles to New York router
(international traffic)
 New York to Nairobi …

April 20, 2025 22

Two Level Hierarchical Routing
Example
Full table for 1A Hierarchical table for 1A
Dest. Line Hops Dest. Line Hops
2A 2B 1A - - 1A - -
1B 1B 1B 1 1B 1B 1
Region 2 1C 1C 1 1C 1C 1
Region 1 2A 1B 2 2 1B 2
1A 1C
2B 1B 3 3 1C 2
2C 2D
2C 1B 3 4 1C 3
2D 1B 4 5 1C 4
3A 1C 3
3B 1C 2
4A 1C 3
5B
4B 1C 4
3A 3B 4A 5A 5C 4C 1C 4
5A 1C 4
Region 3 Region 5
Region 44C 5B 1C 5
5C 1B 5
4B 5E 5D 5D 1C 6
5E 1C 5

April 20, 2025 23

Broadcast Routing
 Sending a packed to all destinations simultaneously
is called Broadcasting.
 Direct Method: Source sends a distinct packet to
each destination routers in the subnet:
1. Wasteful of the bandwidth.
2. It requires source to have a list of all destinations.
 In practice this may be the only feasible solution.
 Flooding:
 Ordinarily ill suited for point-to-point communication:
 Generates too many packets, and
 Consumes too much bandwidth.

April 20, 2025 24

Broadcast Routing (cont.)
 Multi-destination Routing
 Each packets contains:
 A list of designations, or
 A bit map indicating the desired destinations.
 When packet arrives at a router:
 The router checks all the destinations to determine the set
of output lines that will be needed.
 Generates a new copy of the packed for each output line to
be used and includes in each packet only those destinations
that are to use the line.
 After a sufficient number of hops, each packed will carry
only one destination and can be treated as normal packet.
 Multi-destination routing is like separately addressed
packets, except that when several packets must follow
the same rout, one of them pays full fare and the rest
ride free.

April 20, 2025 25

Broadcast Routing (cont.)
 Spanning Tree:
 It is a subset of the subnet that includes all routers
but contains no loops.
 Each router knows which of its lines belong to the
spanning tree, it can copy an incoming broadcast
packet onto all the spanning tree lines except the one
it arrived on.
 Makes excellent use of bandwidth (generates absolute
minimum number of packets necessary to do the job)
 Must have knowledge of some spanning tree for the
method to be applicable.
 Information available in some instances (e.g., link state
routing)
 Information not available (e.g., distance vector routing)

April 20, 2025 26

Broadcast Routing (cont.)
 Reverse Path Forwarding:
 Router checks if the broadcast packet arrived on
the line that is normally used for sending
packets to the source of the broadcast.
 If so, there is excellent chance that the
broadcast packet itself followed the best route
from the router and is therefore the first copy to
arrive at the router. The router forwards copies
of it onto all lines except the one it arrived on.
 If the broadcast packet arrived on a line other
than the preferred one for reaching the source,
the packet is discarded as a likely duplicate.

April 20, 2025 27

Broadcast Routing (cont.)
 Example of Reverse path Forwarding
B C
B C
A D
A
D F
F
E
E I G
I G

J
J H
H L
L N
O
N O
K
K

M
M

A subnet A sink tree for router I

April 20, 2025 28
Broadcast Routing (cont.)
 The tree build by reverse path forwarding. After 5 hops and 24 packets
the broadcasting terminates compared to 14 packets had the sink tree
been followed exactly
I

F H J N

A D E K G O M O

E C G D N K

H B L H

L B
April 20, 2025 29
Broadcast Routing (cont.)
 Reverse Path Forwarding in spite of not
being optimal procedure:
 Efficient and easy to implement Algorithm
 It does not require routers to know about
spanning trees.
 Does not have the overhead of destination list or
bit map in each broadcast packet (as multi-
destination addressing).
 It does not require any special mechanism to
stop the process as flooding does (hop counter in
each packed and a priori knowledge of the
subnet diameter, or a list of packets already
seen per source)

April 20, 2025 30

Multicast Routing
 Application that require separate processes (i.e., each
from separate location) access and ability to work on
the same data.
 Small group can use point-to-point messaging to
accomplish this task.
 Broadcasting can be used but communicating with
1000 “interested” machines out of million-node
network is inefficient.
 Need a mechanism that would send messages to well-
defined groups that are numerically large in size but
small compared to the network as a whole.
 Sending a message to a such a group is called
multicasting.
 Corresponding routing algorithm is called multicast
routing.

April 20, 2025 31

Multicast Routing (cont)
 Requirements:
 Create and Destroy Groups
 Nodes should be able to Join and Leave Groups, etc.
 Group Management.

 When a process joins a group it informs its host.

 Routers must know which of their hosts belong to
which group.
 Host must inform their routers about changes in group
membership, or
 Routers must query their hosts periodically.
 Information shared with Neighboring Routers
(propagation of information through the subnet).

April 20, 2025 32

Multicast Routing (cont)
 Each Router Computes Spanning Tree Covering all other routers.
 Example of a network with nodes belonging to two groups (1 & 2).
2 1
2 1
1, 1,
1, 1, 2 2 2
2 2 2
2 2
2 1
1 1
1 Spanning Tree for the leftmost
Network
Router

1 2

1 1 2 2
2

2
1
1
A Multicast Tree for group 1 A Multicast Tree for group 2

April 20, 2025 33

Multicast Routing Tree Pruning
 Link State Routing – each router is aware of the complete
topology, including which hosts belong to which groups.
 Pruning starting from leaf-node up toward root node.
 Distance Vector Routing – Basic pruning algorithm is based
on reverse path forwarding.
 Router with no hosts interested in a particular group and no
connections to other routers responds with PRUNE message to
a multicast message.
 Also when a router with no group members among its hosts
receives a multicast message it too will respond with a PRUNE
message effectively recursively pruning the subnet.
 Potential Problem:
 Scales poorly to large networks.
 Network with n groups,
 Each Group on average has m members.
 For each Group m – spanning trees must be stored; total of m*n
trees.
 Significant storage requirements for large number of groups.

April 20, 2025 34

Multicast Routing Tree Pruning
 Core-Based Trees – an alternative
algorithm:
 Uses one spanning tree per group,
 Root (core) node near the middle of the group.
 Host sends multicast message to core node;
which in turn sends the message along the
spanning tree.
1. Tree will not be optimal for every source,
2. Reduction is storage from m trees to one tree
per group.

April 20, 2025 35

Routing for Mobile Hosts
 Increasing number of users of Portable
Computers and Personal Computer Devices.
They require access to:
 E-mail
 File System, etc.
 In order to route a packet to a mobile host,
the network first has to find it.
 World Model of communication network:
 WAN consisting of routers and hosts,
 LAN’s connected to WAN, and
 MAN’s connected to WAN.

April 20, 2025 36

 Routing for Mobile Hosts
(cont)
 A WAN to which LAN’s, MAN’s and wireless cells are attached:

Wireless cell
Home Agent

Mobile
Host
Home LAN

Foreign
Agent

Foreign LAN
WAN MAN

April 20, 2025 37

Routing for Mobile Hosts
(cont)
 Stationary Hosts:
 Hosts that never move.
 Migratory Hosts:
 Stationary hosts who move from one fixed site to
another from time to time but use the network only
when they are physically connected to it.
 Roaming Hosts:
 Need to maintain their connections as they move
around.
 Mobile Hosts:
 Migratory and Roaming Hosts – that is all host that
are away from home and still want to be connected.

April 20, 2025 38

Routing for Mobile Hosts
(cont)
 All hosts are assumed to have:
 A permanent home location, and
 A permanent home address:
 Used to determine their home location
(analogous to telephone number; e.g., 1-212-
555-1212).
 Routing goal in systems with mobile hosts:
 To make possible to send packets to mobile
hosts using their home address, and
 Have the packets efficiently reach them
wherever they may be.
 Trick is off course to find them first.

April 20, 2025 39

Routing for Mobile Hosts
(cont)
 According to the sketch in previous slide world is divided up
(geographically) into small units – areas.
 Areas are typically LANs or wireless cells.
 Each area has one or more
 Foreign agents:
 Processes that keep track of all mobile hosts visiting the
area.
 Home agent:
 Keeps track of hosts whose home is in the area, but who
are currently visiting another area.

 REGISTRATION with Foreign Agent: When a new host

enters an area, either by connecting to it (e.g., plugging into
LAN), or wandering into the cell it must register itself with
the foreign agent of that area.

April 20, 2025 40

 Registration Procedure of Mobile
Hosts
1. Each foreign agent broadcasts periodically a packet announcing its existence and
address.
 Newly-arrived mobile host may:
1. Wait for one of these message, or if none arrives quickly enough
2. Can broadcast a packet “saying”: Are there any foreign agents around?

2. The mobile host registers with the foreign agent:

 Gives its home address,
 Current data link layer address, and
 Some security information.

3. The foreign agent contact the mobile host’s home agent and informs it about a mobile
host in his area. This message contains:
 The foreign agent’s network address,
 Security information (“to convince the home agent that the mobile host is really there”).

4. The home agent authenticates security information containing:

 Timestamp (to prove that it was generated within the past few seconds), etc.
 Acknowledges foreign agent by indicating to proceed if everything checks out.

5. Foreign agent registers and informs the mobile host.

 Checking out when done (typically mobile hosts just turn-off their computers).

April 20, 2025 41

 Packet Routing for Mobile Hosts
Example of sending a packed to a mobile host
that has been registered.

foreign agent 1. Pa
c
mob ket is se
packets are

ile h
ost’s nt to the
hom
e ad
3. Se dres
nder s
forei i s gi ve
gn a n
gent
’s ad
tunneled to the

dres
4. Subsequent

to the
t is t unneled
e
2. Pack gent
eigh a
for

April 20, 2025 42

Packet Routing for Mobile
Hosts
 Example: Sender wants to send a packet to a host in New York.

 Packets sent to the mobile host on its home LAN in NEW York are intercepted by
the home agent (step 1).

 Home agent looks up mobile host’s new (temporary) location and finds the
address of the foreign agent handling the mobile host (i.e., Los Angeles).

 Home agent does:

1. It encapsulates the packet in the payload field of an outer packet and sends the
latter to the foreign agent (step 2). This mechanism is called tunneling.
After getting the encapsulated packet, the foreign agent removes the original
packet from the payload field and sends it to the mobile host as a data link frame.

2. The home agent tells the sender to henceforth send packet to the mobile host by
encapsulating them in the payload of packets explicitly addressed to the foreign
agent instead of just sending them to the mobile host's home address (step 3.)
Subsequent packets can now be routed directly to the host via foreign agent (step
4.), bypassing the home agent entirely.

April 20, 2025 43

Routing in Ad Hoc Networks
 Ad Hoc networks refers to the extreme case of
mobility where not only the hosts are mobile but
routers as well. Examples:

1. Military vehicles on a battlefield with no existing

infrastructure.
2. A fleet of ships at sea.
3. Emergency workers at en earthquake that
destroyed the infrastructure.
4. A gathering of people with notebook computers
in an area lacking 802.11 (IEEE Wireless LAN
protocol).

April 20, 2025 Veton Këpuska 44

Routing of Ad Hoc Networks (cont)
 Each node consists of a router and a host –
usually on the same computer.
 Networks of neighboring nodes are called
ad hoc networks or MANETs (Mobile Ad
hoc NETworks).
 Features of such networks:
 Routers come and go or appear in new places all
the time.
 Consequently, Network Topology may be
changing all the time.
 Validity of paths can thus change spontaneously.

April 20, 2025 Veton Këpuska 45

Ad hoc On-demand Distance Vector
routing (AODV) Algorithm
 AODV takes into account limited
bandwidth and low battery life of
devices.
 It is a on-demand algorithm; that is it
determines a route to some
destination only when there is a
demand to send a packed to a
destination.

April 20, 2025 Veton Këpuska 46

Ad hoc On-demand Distance Vector
routing (cont)
 Route Discovery
 Ad-hoc network can be described by a graph of
the nodes (routers + hosts).
 Two nodes are connected (depicted via an
connecting arc in the graph) if they can
communicate directly using their radios.
 For simplicity connection is assumed symmetric
(e.g., case where node A has a more powerful
transmitter then node B thus node A can
communicate to B but not vice versa is not
considered).

April 20, 2025 Veton Këpuska 47

 Ad hoc On-demand Distance Vector
routing Example
Range of A’s
Broadcast

A A A A
B C B C B C B C

D D D D
E E E E
F G F G F G F G

H I H I H I H I
Range of A’s Broadcast After B and D have After C,F and G have After E,H and I have
received of A’s Broadcast received of A’s Broadcast received of A’s Broadcast

April 20, 2025 Veton Këpuska 48

AODV Algorithm
 Scenario: A wants to send a packet to node I.

 AODV algorithm maintains a table at each node keyed by

destination giving information about that destination
including which neighboring node to send packets to in
order to reach the destination.

 Suppose node A can not find the information about

destination node I. Thus it needs to discover a route to node
I.

 Algorithm Details.

April 20, 2025 Veton Këpuska 49

AODV Algorithm (cont)
 To locate an unknown node (I), source node A constructs a special
ROUTE REQUEST packet and broadcasts it.
 The packed reaches A’s neighboring nodes B and D. B and D at this
time are directly communicating with A (as opposed to F for example).
 Format of ROUTE REQUEST PACKET:

Source Request Destination Source Destination Hop

Address ID Address Sequence # Sequence # Count
 Source and Destinations Address (typically IP address).
 Request ID – local counter maintained separately by each node and
incremented each time a ROUTE REQUEST is broadcast.
 Source Address and Request ID uniquely identify the ROUTE REQUEST
packet. This allow nodes to discard any duplicate packets they may receive.
 Sequence Counters – Each node also maintains a second sequence counter
incremented whenever a ROUTE REQUEST is sent (or a reply to someone's
else’s ROUTE REQUEST) . It is used to tell new routes form old routes.
 Hop Count – keeps track of how many hops the packet has made.

April 20, 2025 Veton Këpuska 50

AODV Processing on a node (cont)
 When ROUTE REQUEST packet arrives at a node (B & D in this case), it is processed in the following
steps:

1. The Source Address, Request ID pair is looked up in local history table to see if this request has
already been seen and processed.


If duplicate – it is disregarded
If not duplicate – pair is entered in history table.
A
B C
2. The receiver node looks up the destination in its route table.
If a fresh route to the destination is know, a ROUTE REPLY packet is sent back to the source
D
informing it how to get to the destination. Fresh means that the Destination sequence
number stored in the routing table is greater than or equal to the Destination sequence
number in the ROUTE REQUEST packet. If it is less, the stored route is older than the previous
route the source had for the destination thus next step 3 is executed. E
F G

H I
After B and D have
received of A’s Broadcast

April 20, 2025 Veton Këpuska 51

 3. Since the receiver node does not know a fresh route to the destination, it:
 Increments the Hop count field,
 Rebroadcast the ROUTE REQUEST packet to its neighboring nodes with exception to
node from which it has received it.
 A
Extracts the data from the packed and stores it as a new entry in its reverse route
table. This information will be used to construct the reverse routeB so thatCthe reply
can get back to source later as specified with reverse arrows.
 A timer is started for the newly-made reverse route entry. If it expires the entry is
deleted. D
E
F G

H I
After C,F and G have
received of A’s Broadcast

April 20, 2025 Veton Këpuska 52

AODV Algorithm (cont)
 Since neither B nor D knows where I is, each creates reverse route
entry pointing back to A, and broadcasts the packet with Hop count
set to 1.
 Broadcast from B reaches C and D. A

B C
C makes an entry for it in its reverse route table and rebroadcasts it.
 D in contrast rejects it as a duplicate.
 Broadcast from D’s D
 is rejected by B as a duplicate, however, E
 it is accepted by F and G. F G

H I
After C,F and G have
received of A’s Broadcast

April 20, 2025 Veton Këpuska 53

AODV Algorithm (cont)
 After E, H, and I receive the broadcast, the ROUTE
REQUEST finally reaches a destination that knows
where I is, namely I itself. A
B C

 Note: Broadcast from each node are notD

E
coordinated in any way. F G

H I
After E,H and I have
received of A’s Broadcast

April 20, 2025 Veton Këpuska 54

Reply Packet in AODV
 In response to the incoming request, I builds a ROUTE
REPLY packet and:

Source Destination Destination Hop Life-

Address Address Sequence # Count time

 The Source address, Destination address, Hop

Counter are copied from the incoming request.
 Destination sequence number is taken from its counter.
 The Hop count field is set to 0.
 The Lifetime field controls how long the route is valid.
 This packed is cast back to the node that ROUTE
REQUERST packed came form, in this case node G.
 From G it follows the reverse path to D and finally to A.
At each node Hop count is incremented so the node can
see how far from the destination (I) is.

April 20, 2025 Veton Këpuska 55

Reply Packet in Intermediate nodes
 At each intermediate node the packed is inspected and
entered into local routing table as a route to I if one or
more of the following three conditions are met:

1. No route to I is known.
2. The sequence number for I in the ROUTE REPLY packed is
greater than the value in the routing table.
3. The sequence numbers are equal but the new route I s shorter.

 This ensures that all nodes on the reverse route learn the
route to I for free as a byproduct of A’s route discovery.
 Nodes that got the original REQUEST ROUTE packed but
were not on the reverse path (B, C, E, F, and H in this
example) discard the reverse route table entry when the
associated timer expires.

April 20, 2025 Veton Këpuska 56

Route Maintenance
 In Ad-hoc networks the nodes can move or can be switched
off.
 For example if G is switched off, A will need to deal with it
because the route that it was using (ADGI) is no longer valid.
 Idea: each node broadcasts a Hello message periodically. Each
neighbor is expected to respond to it. For the cases that there
is no response the information is purged for the routes that use
corresponding node.
 List of Active Neighbors for a destination: For each possible
destination, each node, N, keeps track of its neighbors that
have fed it a packed for that destination during the last ∆T
seconds.
 A node N does this by having a routing table keyed by
destination containing:
 Outgoing node to use to reach that destination,
 Hop count to the destination
 Most recent destination sequence number, and
 The list of active neighbors for that destination.

April 20, 2025 Veton Këpuska 57

Example of a Routing Table
Dest. Next Distanc Active Other
hop e Neighbors fields

A A 1 F,G A
B C
B B 1 F,G
C B 2 G D
E G 2 E
F G
F F 1 A,B
G G 1 A,B
H F 2 A,B
H I
I G 2 A,B Graph after G has gone
down

D’s routing table

before G goes down

April 20, 2025 Veton Këpuska 58

Route Maintenance (cont)
 When any of N’s neighbors becomes
unreachable, it checks its routing table to
see which destinations have routes using
this node. For each of these routes, the
active neighbors are informed that their
route via N is now invalid and must be
purged from their routing tables.
 The active neighbors nodes propagate this
information to their active neighbors, and
so on recursively, until all routes depending
on the unavailable node are purged from all
routing tables.

April 20, 2025 Veton Këpuska 59

Congestion Control Algorithms
 The situation when to many packets are present in
subnet it is called congestion.
Congestion Traffic Performance Chart as functino of Number
of Packets Sent

1.2

1
Maximum Carrying
capacity of subnet
0.8
Packets Delivered [%]

Desirable Performance
0.6 Congested Performance
Ideal

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Num ber of Packets in Mil.

April 20, 2025 Veton Këpuska 60

Congestion Factors
 Streams of inputs packets arriving from multiple lines (3-4 or more) needing
the same output line => queue buildup.
 Adding more memory may help to a point but – a study by Nagle suggest that
increasing memory to ∞ to accommodate larger queues congestions gets worse
not better due to time required to get to the front of the queue (timed out).

 Slow Processors:
 Queue build up due to routers slow CPU’s at performing bookkeeping task:
 Queueing buffers,
 Updating tables, etc.

 Low bandwidth lines.

 Typically the problem is that upgrading one part of the system shifts the
bottleneck to the other someplace else. The real problem frequently is a
mismatch parts of the system. The problem will persist until all components
of the system are in balance.

April 20, 2025 Veton Këpuska 61

Congestion Control vs. Flow Control
 Congestion Control has to due with making sure the subnets
are able to carry the offered traffic. Thus it is a global issue
involving the behavior of
 all the hosts,
 all the routers,
 the store-and forwarding processing within the routers, and
 all the other factors that tend to diminish the carrying capacity
of the subnet.
 Flow Control relates to point-to-point traffic between a given
sender a given receiver. Its job is to make sure that a fast
sender cannot continually transmit data faster than the
receiver is able to absorb it.
 It frequently involves some direct feedback from the receiver
to the sender to tell the sender how things are doing at the
other end.

April 20, 2025 Veton Këpuska 62

Congestion Control vs. Flow Control
Example
 Flow Control Problem: Consider a fiber optic network
with a capacity of 1000 Gbps on which a
supercomputer is trying to transfer a file to a personal
computer at 1 Gbps. Although there is not congestion
the supercomputer has to frequently stop in order to
allow PC to catch up.
 Congestion Control Problem: Consider a store-and-
forward network with 1 Mbps lines and 1000 large
computers, half of which are trying to transfer files at
100 kbps to the other half creating a traffic of
500x100 kbps = 50000 kbps = 50 Mbps. The problem
here is that this traffic exceeds the capacity of what
the network can handle.

April 20, 2025 Veton Këpuska 63

 General Principles of Congestion
Control
 Analogy with Control Theory:
 Open-loop, and
 Closed-loop approach.

 Open-loop approach
 Problem is solved at the design cycle
 Once the system is running midcourse correction are NOT
made.
 Tools for doing open-loop control:
 Deciding when to accept new traffic,
 Deciding when to disregard packets and which ones.
 Making scheduling decision at various points in the network.
 Note that all those decisions are made without regard to the
current state of the network.

April 20, 2025 Veton Këpuska 64

 General Principles of Congestion
Control
 Closed-loop approach
 It is based on the principle of feedback-loop. The
approach has three parts when applied to congestion
control:
1. Monitor the system to detect when and where
congestion occurs,
2. Pass this information tot places where action can be
taken
3. Adjust system operation to correct the problem.

April 20, 2025 Veton Këpuska 65

 General Principles of Congestion
Control
 Monitoring - Metrics to be used to monitor subnet congestion:
 Percentage of all packets disregarded for lack of buffer space.
 Average queue length
 Number of packets that time out and are retransmitted,
 Average packet delay
 Standard deviation of packed delay.

 Transferring congestion information

 Information send to the source(s) of the traffic. Undesirable because extra
traffic is initiated when opposite is needed.
 Another approach is to reserve a bit field that can be set when congestion
gets above some level of threshold.
 Routers/Hosts send periodically probe packets out to explicitly ask about
congestion. This information can then be used to route traffic around
problem areas.
 In general in all feedback schemes, the hope is that knowledge of
congestion will cause the hosts to take appropriate action to reduce the
congestion. Time constant in any adjustment scheme is critical and non-
trivial problem:
 To fast adjustment is responsive but can lead to oscillations.
 To slow adjustment is response time is sluggish and of no real value.

April 20, 2025 Veton Këpuska 66

 General Principles of Congestion
Control
 The presence of a congestion means
that the load is (temporarily) greater
than the resources (in part of the
system) can handle.
 Increase Resources, and/or
 Decrease the Load.

April 20, 2025 Veton Këpuska 67

 General Principles of Congestion
Control
 Increase of Resources:
 Use dial-up telephone lines to temporarily
increase the bandwidth between certain points.
 On satellite systems increasing transmission
power often gives higher bandwidth.
 Splitting traffic over multiple routes (instead of
using the best one) may effectively the
bandwidth.
 Use of spare routers that are normally used as
backups to give more capacity.

April 20, 2025 Veton Këpuska 68

 General Principles of Congestion
Control
 Decrease the Load
 It is not always possible to increase
capacity of the subsystem. Thus the only
other way to reduce congestion is to
decrease the load.
 Denying service to some users (AOL in early
days),
 Degrading service to some or all users,
 Having users schedule their demands in
more predictable way.

April 20, 2025 Veton Këpuska 69

 Congestion Prevention Policies
Open-loop Approach
 Open-loop system designed to minimize congestion in
the first place rather then reacting after it happened.
 Try to achieve this goal at various levels/layers:

 Data Link Layer:

 Retransmission Policy:
 Sender with quick time out and go-back-n retransmitting
packets vs.
 Selective repeat with slower time out.
 Out-of-order catching policy
 Acknowledgment policy
 Flow control policy

April 20, 2025 Veton Këpuska 70

 Congestion Prevention Policies
Open-loop Approach
 Network Layer:
 Virtual circuit versus datagram inside the subnet:
 Many congestion control algorithms work only with virtual circuit
subnets.
 Packet queueing and service policy:
 One queue per input line or output line or both.
 Processing the queue:
 Round-robin
 Priority based queue processing.
 Packet discard policy
 Bad policy can make the problem worse.
 Routing algorithm
 Spreading the traffic over all the lines can help, vs.
 Directing the traffic on the already congested line.
 Packet lifetime management – Deals with issue that determines how
long a packet may live before it is disregarded.
 To long a life time lost packets may clog up the works for a long time.
 To short on the other hand packets may time out before even given a
chance to reach the destination, thus inducing retransmission.

April 20, 2025 Veton Këpuska 71

 Congestion Prevention Policies
Open-loop Approach
 Transmission Layer – same issues like in Data Link Layer
with one addition determining the time out is harder
because the transit time across the network is less
predictable than the transit time over a wire between
two routers.
 Retransmission Policy:
 Out-of-order catching policy
 Acknowledgment policy
 Flow control policy
 Time out determination

April 20, 2025 Veton Këpuska 72

 Congestion Prevention Policies
in Virtual-Circuit Subnets
 Admission Control:
 Simple algorithm – once congestion has been
signaled, no more virtual circuits are set up until
the problem had gone away.
 Attempts to setup new transport layer
connections will fail.
 Alternate Approach to Admission Control:
 Allow new virtual-circuits but carefully route all
new ones around the problem area,

April 20, 2025 Veton Këpuska 73

 Alternate Approach to Admission
Control Example

New Virtual-Circuit

A A

B B

b) A redrawn subnet that

a) A congested subnet eliminates the congestion
April 20, 2025 Veton Këpuska 74
Negotiated Approach to Congestion
Control
 Agreement is negotiated between the host and subnet
when a virtual circuit is set-up.
 Agreement specifies:
 Volume and shape of the traffic,
 Quality of service requires, etc.
 Subnet guarantees the connection since it will make
all necessary resource available during the set.
Resource typically include:
 Table and buffer space in the routers, and
 bandwidth on the lines.
 Agreement can be done
 All the time as part of standard operating procedure
(wastes resources), or
 Only when the subnet is congested.

April 20, 2025 Veton Këpuska 75

 End

April 20, 2025 Veton Këpuska 76