Many Internet multicast applications have

stringent Quality-of-Service (QoS) requirements

that include delay, loss rate, bandwidth, and delay
jitter. In this paper, we present a Swarm
intelligence based on Ant Colony Optimization
(ACO) technique to optimize the multicast tree .

Solving QoS Multicast

Routing Problem Using
ACO Algorithm
Course of: High speed network

Abdulaziz Tagawy
Solving QoS Multicast Routing Problem Using ACO
Algorithm

Define the problem:

In IP multicasting, there are many applications (such as videoconferencing,
distance education, and online simulation, Access to Distributed Databases, and
Information Dissemination.) are sent messages from the source node to all
destination nodes.

To support these applications, it is necessary to determine a multicast tree of

minimal cost to connect the source node to the destination nodes subject to
delay constraints.

One of the most distinctive aspects of the network routing problem is the
nonstationarity of the problems characteristics. In particular, the characteristics
of traffic over a network changes all the time, and in some important cases (e.g.,
the Internet) the traffic can fluctuate in ways difficult to predict. Additionally,
the nodes and links of a network can suddenly go out of service, and new nodes
and links can be added at any moment.

These applications have stringent Quality-of-Service (QoS) requirements that

include delay, loss rate, bandwidth, and delay jitter. This leads to the problem of
routing multicast traffic satisfying QoS requirements. The above mentioned
problem is known as the QoS constrained multicast routing problem and is NP
Complete.

In order to meet QoS requirements an optimizing algorithm is needed.

Therefore; we introduce to use swarming agent based intelligent algorithm

which is a relatively new approach of Ant Colony Optimization (ACO) that
takes inspiration from social behaviors to solve optimization problem.
Literature Review:
Shortest-path routing has a source-destination pair perspective: there is
no global cost function to optimize. Its objective is to determine the
shortest path (minimum cost) between two nodes, where the link costs are
computed (statically or adaptively) according to some statistical
description of the trac flow crossing the links. Considering the different
content stored in each routing table, shortest-path algorithms can be
further subdivided into two classes called distance-vector and link-state
(Steenstrup, 1995).
Distance-vector algorithms make use of routing tables consisting of a set
of triples of the form (destination, estimated distance, and next hop),
defined for all the destinations in the network and for all the neighbor
nodes of the considered switch. In this case, the required topologic
information is represented by the list of identifiers of the reachable nodes.
The average per node memory occupation is in. The algorithm works in
an iterative, asynchronous, and distributed way.

The information that every node sends to its neighbors is the list of its last
estimates of the distances (intended as costs) from itself to all the other
nodes in the network.

It can be noted that this algorithm converges in finite time to the shortest
paths with respect to the used metric if no link cost changes after a given
time (Bellman, 1958; Ford & Fulkerson, 1962; Bertsekas & Gallager,
1992); this algorithm is also known as distributed Bellman-Ford.

Link-state algorithms make use of routing tables containing much more

information than that used in distance-vector algorithms. In fact, at the
core of link-state algorithms there is a distributed and replicated database.
This database is essentially a dynamic map of the whole network,
describing the details of all its components and their current
interconnections. Using this database as input, each node calculates its
best paths using an appropriate algorithm such as Dijkstras (Dijkstra,
1959), and then uses knowledge about these best paths to build the
routing tables.

In the most common form of link -state algorithm, each node acts
autonomously, broadcasting information about its link costs and states
 and computing shortest paths from itself to all the destinations on the
basis of its local link cost estimates and of the estimates received from
other nodes. Each routing information packet is broadcast to all the
neighbor nodes which in turn send the packet to their neighbors, and so
on. A distributed flooding mechanism (Bertsekas & Gallager, 1992)
supervises this information transmission, trying to minimize the number
of retransmissions.

In recent years, some meta-heuristic algorithms such as the ant colony

algorithm, genetic algorithm, particle swarm optimization, and Tabu
search, have been adopted by the researchers to solve the multi-
constrained QoS routing problems.
In mobile agents for adaptive routing an intelligent routing algorithm
ANTNET based on ant colony algorithm was proposed. Their algorithm
has some attractive distribution features and it can find a near-optimal
path from the source node to each destination node. It also provides the
required results in simulation. Although the ANTNET is a unicast routing
algorithm, it can be easily applied to multicast routing with some
modifications. In spite of the said merits of the ANTNET, it suffers from
a serious drawback i.e. the slow convergence rate.
The genetic algorithm (GA) was used to find a multicast tree satisfying
the constraints of bandwidth and delay with least cost. The GA has three
operators: selection, crossover, and mutation. The individuals are stored
in connective matrices by adopting the binary coding system.
The initial colony is generated randomly without considering QoS
constraints. The selection operation adopts Roulette wheel algorithm to
select the best individuals from the parent generation to pass onto the
child generation. Then, the crossover operation is used to find out the
fittest among the best. A penalty function is adopted to solve QoS
constraints in the multicast trees, which do not satisfy QoS constraints.
Although sometimes the algorithms performance is observed to be
satisfactory, still it encounters some faults, such as the local search
ability, premature convergence, and slow convergence speed. Further, the
genetic algorithm does not assure to find a global optimum. It happens
very often when the populations have a lot of subjects.

Researchers have proposed particle swarm optimization PSO algorithms

to solve QoS constraint routing problem. The PSO algorithm proposed
 solves the QoS multicast routing problem and can obtain a feasible
multicast tree by exchanging the paths. This algorithm can converge to
the optimal or near-optimal solution with lower computational cost.
Another algorithm based on the concept of quantum mechanics named as
Quantum-Behaved Particle Swarm Optimization (QPSO) was proposed.
Here, the proposed method converts the QoS multicast routing problem
into integer programming problem and then finally solves using the
QPSO. The QPSO finds the path from the source node to each destination
node and constructs the tree by merging the paths. A tree based PSO has
been proposed for optimizing the multicast tree directly. However, its
performance depends on the number of particles generated. Another
drawback of the said algorithm is the merging of multicast trees. The
elimination of directed circles and nesting of directed circles are also very
complex and are considered as some of the limitations of the PSO.
many researchers have solved the QoS constrained multicast tree using
Tabu search [18,19]. A Tabu search method was proposed in [18] to
search for the multicast tree with the least cost that satisfies the
constraints of bandwidth and delay. This algorithm obtains a complete
graph of all group members at the initial step and obtains the initial
Steiner tree via the generated tree of the complete graph. In this way, the
k-shortest paths replace the edges to find the chances of getting better
results. The method mentioned above is similar to the method of path
combination. However, it does not operate directly on the multicast tree.
This weakness makes it impossible to eliminate the constraints of
conventional multicast routing algorithms. Hence, there arises a need to
proceed further and do more amount of work in searching paths and
integrating the multicast trees.
Research Methodology:
Before sending or receiving IP multicast data, a network must be enabled for
multicasting, as follows:

Hosts must be configured to send and receive multicast data.

Routers must support the Internet Group Membership Protocol (IGMP),
multicast forwarding, and multicast routing protocols.

Working:

To support multicasting in an internetwork, the hosts and routers must be

multicast-enabled. In an IP multicast-enabled intranet, any host can send IP
multicast datagrams, and any host can receive IP multicast datagrams, including
sending and receiving across the Internet.

The source host sends multicast datagrams to a single Class D IP address,

known as the group address. Any host that is interested in receiving the
datagrams contacts a local router to join the multicast group and then receives
all subsequent datagrams sent to that address.

Routers use a multicast routing protocol to determine which subnets include at

least one interested multicast group member and to forward multicast datagrams
only to those subnets that have group members or a router that has downstream
group members.

ARCHITECTURE:
 Component Description

A host is any client or server on the network. A multicast-

enabled host is configured to send and receive (or only send)
Host (source or receiver) multicast data.

A multicast router is capable of handling host requests to join

or leave a group and of forwarding multicast data to subnets
that contain group members. A multicast router can be either
Router a non-Microsoft router that uses a multicast routing protocol
or a Windows Server 2003-based server running the Routing
and Remote Access service.

A Class D IP address used for sending IP multicast data.

An IP multicast source sends the data to a single
Multicast address multicast address, as described later in this section. A
specific IP multicast address is also known as group
address.

A multicast group is the set of hosts that listen for a

Multicast group specific IP multicast address. A multicast group is also
known as a host group.

The Internet multicast backbone, or the portion of the

MBone
Internet that supports multicast routing.

ANT COLONY OPTIMIZATION ALGORITHM:

Swarm intelligence is a relatively new approach to problem solving that takes

inspiration from social behaviors to solve optimization problem. The attempt is
to develop algorithm in computer technology to solve real life problems. Ant
colony optimization is a heuristic algorithm which has been proven a successful
technique and applied to a number of combinational optimization problem and
taken as one of the high performance computing methods.

It has wide range of applications with very good search capabilities for
optimization problems but it still remains a bottleneck due to high cost and time
conversion. ACO inspired by the forging behavior of real ants to find food from
their nest. The algorithm is basically used to find shortest path from nest to food
source and the path is then used by other ants this is all due to chemical name
pheromone deposited by ants on ground while searching for food.
Ant Colony Optimization technique has emerged recently novel meta-heuristic
for a hard combinational optimization problems. It is designed to stimulate the
ability of ant colonies to determine the shortest paths to food. Although
individual ant possess few capabilities, their operation as a colony is capable of
complex behavior. Real ants can indirectly can communicate though pheromone
information without visual cues and capable of finding shortest path between
food and their nests. The ants follow pheromone on trail while walking and the
other ants follow the trail with some probability dependent on the density of
pheromone deposited by the ants. The more the pheromone deposited the more
ants will follow that trail. Through this mechanism ants ultimately find the
shortest path.

Artificial ant can also imitate the behavior of real ants how they forage the food
but can also solve much more complicated problems than real ants can. A
search algorithm with such concept is called Ant Colony Optimization
Algorithm. ACO inspired by the forging behavior of real ants to find food from
their nest.

ALGORITHM:

Assuming S is source code and U= {U1, U2...Um} donated a set of

destination nodes.

1. Initialize network nodes.

2. Set LC=0 where LC is loop count.

3. Let be the shortest path for the destination node Ui.

4. The initial value of =0 as no ant has traversed any path so ant can
chose any path as probability of any path=0.
5. Ant chooses the path according to the probability of path.

6. Compute the pheromone update of the path and each edge selected by
the ant using

=

Where

R is any constant value.

Lk is total path traversed by the ant.

Update the local pheromone

= (1-) +

Where = (0 to 1) pheromone decay.

Compute the probability of each edge.

[]
[]
=
[]
[]

=1

Where k N

, are meta values.

heuristic value.

edge value.

9. Set LC=LC+1.

Repeat from step 5 update the value of and probability of paths.

11. Collect best paths to get the multicast tree.

Solution and Results:
We consider the multicast routing problem with bandwidth and delay
constraints from one source node to muti-destination nodes.

Find the shortest path between 1 and 5 ?

Solution:-

= 1 , =1 , =0.1 .

Initial k = 0, Q = 10

[]
[]
=
[]
[]

kN

Otherwise,
 =1

k = (1-)k + k

Initially fk for every path = 0 as k = 0

Ant chose path 1 2 4 5

C1,5 = d1,2 + d2,4 + d4,5

= 4+4+3 =11

Pheromone Value

k =10/11

= 0.90

As k = 0 + k

k = 0.90

2nd Ant chose path

P1,2 =[(0.90)(1/4)/(0.9)(1/4)+0]=1

P2,4 = 1 , f4,5 = 1

P1,3 = 0 , f2,3 = 0, f3,5 = 0

k = 0.90

k = (1-0.1).90 + .90

= 1.71

3rd Ant chose same path

ck = 11 , k = .90

k = (1-0.1)1.71 + .90

= 2.43

For path
1,2 = 2.43

2,4 = 2.43

4,5 = 2.43

4th Ant chose path 1 2 3 5

Ck = d1,2 + d2,3 + d3,5

= 4 + 5 + 6 = 15

k =10/15 =0.66 , k = 0.66

k 1,2 = (.9)*2.43 + 0.66 = 2.84

5th Ant on same path

k 2,3 = (.9)*0.66 + 0.66 = 1.25 = 3,5

1,2 = (.9)*2.84 + 0.66 = 3.21

6th Ant on same path

k 2,3 = 3,5 = (.9)*1.25 + 0.66 = 1.78

1,2 = (.9)*3.21 + 0.66 = 3.54

Path 1 3 5

ck = d1,3 + d3,5 = 5 + 6 = 11

k =10/11= 0.9

7th Ant

1,3 = .90

3,5 = (.9)*1.78 + .90

= 2.50

8th Ant

1,3 = (.9)*(.90) + .90 = 1.71

3,5 = (.9)*2.50 + .90 = 3.15

9th Ant

1,3 = (.9)*1.71 + .90 = 2.43

3,5 = (.9)*3.15 + .90 = 3.75

Destination Node The shortest paths
1245
1235
5
135
13245
The route set from source to destination.

The above table shows all the possible paths to the destination node.

Hence,

1,2 = 3.54, 1,3 = 2.43, 2,3 = 1.78

2,4 = 2.43, 3,5 = 3.75, 4,5 = 2.43

Finding shortest path

Ant 10th chose 1 2 4 5

ck = 11

k = .90

1,2 = (.9)*3.54 + .90 = 4.086

2,4 = (.9)*2.43 + .90 = 3.087

4,5 = (.9)*2.43 + .90 = 3.087

According to the above calculation path 1 2 4 5 has the highest probability.

Initially paths 1-2-4-5 and 1-2-3-5 were chosen by the ants. But the pheromones
on the path 1-2-4-5 were more, which lead to its higher probability to be chosen
by the ants.

Thus, 1-2-4-5 is the optimum path chosen by the ants. Given below is the
multicast tree obtained by the proposed Ant Algorithm.

CONCLUSION:
Hence by using the concept of ACO algorithm we found an optimum path to
send message from a source node to a destination node. The first ant moves
from the source to the destination leaving a trail of pheromones. The next ant
calculates the probability of the path based on the presence of pheromones on
that path. Higher the pheromone value on that path higher will be the
probability. And ant tends to choose higher probability path. Results show that
the proposed algorithm has features of well performances of cost, fast
convergence and stable delay.

NOTE:

1. I have some not about this paper that the researcher did make a
comparisons with other technique to prove what he deduced.
2. And also there is no explanation about who the Qos service improved by
using this technique and the old version that suffers from a serious
drawback i.e. the slow converbgence rate.