Nâng cao hiệu năng mạng MANET sử dụng kỹ thuật định tuyến cân bằng tải đảm bảo chất lượng truyền dẫn tt tiếng anh

MINISTRY OF EDUCATION
AND TRAINING

VIETNAM ACADEMY
OF SCIENCE AND TECHNOLOGY

GRADUATE UNIVERSITY SCIENCE AND TECHNOLOGY
———————————-

LE HUU BINH

IMPROVE THE PERFORMANCE OF MOBILE
AD HOC NETWORK USING LOAD BALANCING
ROUTING TECHNOLOGY ENSURING
QUALITY OF TRANSMISSION

Major: Information System
Code: 9480104

SUMMARY OF INFORMATION TECHNOLOGY
DOCTORAL THESIS

HA NOI - 2019


The thesis has been completed at Graduate University of Science
and Technology, Vietnam Academy of Science and Technology

Supervisor 1. Assoc. Prof. Dr. Vo Thanh Tu
Supervisor 2. Assoc. Prof. Dr. Nguyen Van Tam

Review 1:
Review 2:
Review 3:

The dissertation is defended at Graduate University of
Science and Technology - Vietnam Academy of Science
and Technology, 18 Hoang Quoc Viet street, Hanoi, at
. . . . . . .on. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

This thesis could be found at:
- National Library of Vietnam.
- Library of Graduate University of Science and Technology.


INTRODUCTION
1. The necessity of this research
With the development trend of communication network technologies, the wireless
communications is one of the decisive solutions for the transmission technology of
the telecommunications network in general, and the computer network in particular. In the era of the fifth generation wireless network (5G) and Internet of things
(IoT), there are several wireless network models to provide the practical applications such as MANET, wireless sensor networks (WSN), wireless mesh networks
(WMN), and hybrid wireless networks. Among these types, MANET is becoming
more and more widely used in many fields [66].
To be able to expand the scope of application of the MANET, it is necessary to
improve the transmission speed, increase the radio range, expand the area of the
network. However, this can lead to some technical difficulties. For example, if the
transmission speed and the radio range increase, the physical effects that happen on
the routes also increase, reducing the network performance [26, 29, 30, 61, 65]. To
improve the network performance, it is necessary to find solutions to ensure QoT in
the network. The QoT of the channels depends on the route, meanwhile, the route
is determined by the routing algorithm. Therefore, the study of QoT constraint

routing algorithms in MANET is very necessary. This issue has been interested
by many research groups recently [5, 24, 33, 35, 46, 51, 53]. These published
works have proposed routing algorithms that take into account the constraints of
some QoT parameters, where the proposed algorithms attempt to find out the best
QoT route. This therefore improves the QoT in the network. However, for the mesh
topologies such as MANET, the routing technique with the best QoT can increase
the bottlenecks due to unbalanced traffic load. To reduce the bottlenecks in the
networks, the load balancing routing is often used [34, 39, 41, 44, 67, 70]. However,
it can be the cause of the QoT impairment due to the route passing through many
intermediate nodes.
We have the following comments based on the analysis above, it is necessary to
investigate the routing algorithms that take into account both QoT and load balancing, especially in the case of a wide MANET, high bit rate and heavy node
density. This is the research motivation of this thesis. The author focuses on studying the load balancing techniques, while ensuring QoT of data transmission routes
to improve the performance of MANET.
1


2. Research purpose
The thesis focuses on the analyzing and evaluating QoT of the data transmission
routes and its effect on the performance of MANET according to different routing
algorithms. Thence, propose improved routing algorithms to balance the traffic
load, meanwhile ensuring QoT on data transmission routes, improving the performance of MANET networks.
3. Research scope and object
The research object of the thesis focuses on the load balancing routing algorithms
and QoT aware routing in MANET. The research scope of the thesis is the DSR
and AODV routing protocols.
4. Research contents
The thesis focuses on studying the following contents: (i) Constructing and developing QoT constraint conditions according to different routing algorithms. (ii)
Analyze and evaluate QoT in MANET network in the cases DSR, AODV routing
protocols and load balancing routing are used. (iii) Proposing improved routing

algorithms of DSR and AODV protocols in order to balance traffic load in the
network, meanwhile ensuring QoT of the data transmission routes, improving the
performance of the MANET network .
5. Layout of thesis:
The thesis consists of following sections:
The introduction section focuses on analyzing the necessity of the research topic,
thence determining the research purpose, the object and scope of the research as
well as the research methods of the thesis. Chapter 1 presents the overview of
MANET and the factors that affect on the network performance. Chapter 2 focuses
on evaluating the quality of transmission of MANET network in the cases using
the on-demand routing and load balancing routing protocols. Chapter 3 proposes
a load balancing routing algorithm ensuring the quality of transmission based on
the traffic load that offers to each route. Chapter 4 proposes the Source-based load
balancing ensuring quality of transmission based on the characteristics of dynamic
source routing protocol in MANET. The conclusions section presents the new contributions of the thesis and proposes the contents of studying for the future. Finally,
there are two appendixs. Appendix A presents in detail the calculation for illustrative examples in the thesis. Appendix B presents the source code of some main
modules in the simulation software based on OMNeT ++.
2


CHAPTER 1
OVERVIEW OF MANET AND FACTORS AFFECTING ON NETWORK
PERFORMANCE
1.1. The basics of MANET
This content presents the principles, characteristics of MANET network and the
factors affecting on the network performance.
1.2. Routing in MANET
In MANET, the routing protocols need to perform two tasks, one is to create routing information, that is, discover the route from source node to destination node to
update into the route cache. The second is to maintain updated routing information
to determine the route information in route cache is still fresh or not.

1.3. Related researches in the fields of routing in MANET
1.3.1. QoS routing
QoS routing is the routing technique in which QoS parameters such as packet
blocking probability, latency, and throughput are considered during the route discovery process to ensure QoS of the network system [1, 14, 62].
1.3.2. QoT routing
QoT routing is the routing technique in which QoT parameters are considered during route discovery. Recently, QoT routing technique has been implemented by
some research groups. There are two methods currently used to determine the
constraint conditions of QoT in the routing algorithms. Firstly, constraining QoT
through the weight function. This method is done by constructing weight functions
that contain the parameters of QoT, the routing algorithm based on this weight
function to select the route [46, 35, 29]. Second, constraining QoT through control
packets. This method is done by using control packets such as RREQ and RREP
to exchange QoT information between network nodes, thence determining the constraint conditions of QoT for the route selection [5, 24, 51, 58].
1.3.3. Load balancing routing
The load balancing routing technique in MANET has been implemented by several
research groups recently. The authors of [44] have proposed a load balancing routing protocol for MANET namely LMP-DSR (Load balanced Multi-Path Dynamic
Source Routing). LMP-DSR protocol is modified from original DSR protocol by
using multiple paths routing instead of single path routing. In [39], a multi-level
routing algorithm (MRA) has been proposed to balance the traffic load in wireless
3


ad hoc network. MRA uses an efficient method of selecting the intermediate nodes
which have the enough resources and capability to reach the destination node. In
[34], the authors have proposed a routing protocol called LBCAR. This protocol
uses two metrics, traffic load density and link cost associated with a routing path
in order to determine the congestion status, the route with low traffic load density
and maximum life time will be selected for data transmission.
1.3.4. Some comments and evaluations
• Proposing routing algorithms taking into account QoT has been implemented.

However, most of the proposed algorithms check the QoT constraint conditions
after the route set has been found. Therefore, there are some cases where the
found route is not the best route with QoT, even does not satisfy the constraint
conditions of QoT.
• Regarding the network models are used for the performance evaluation, most
of works only evaluate the network models with low bit rate, using the channels
with the bandwidth of 20 MHz. In the case of a broadband network, the effects
of physical effects need to be considered, because the wider the bandwidth, the
greater the interference on the channel.
• Load balancing routing in MANET has also been devoloped by some research
groups. The results have demonstrated that network performance improves in
terms of the packet blocking probability and network throughput. However, the
constraints of QoT have not been considered in the balanced routing algorithms.
1.4. The new contributions of the thesis
(i) The thesis has proposed a new method to determine the constraint conditions
of the quality of transmission based on the cross-layer model. This method is
used for discovering the route of the on-demand routing protocols in MANET.
(ii) Based on traffic load offers each route, the thesis has proposed Load Balancing
Routing algorithm ensuring Quality of Transmission (LBRQT) for MANET.
(iii) Based on the characteristics of DSR protocol, the thesis has proposed Sourcebased Load Balancing ensuring Quality of Transmission in DSR for MANET.
1.5. Conclusion of chapter 1
Chapter 1 has presented the basics of MANET and the factors affecting on the
network performance, in which routing techniques were analyzed in depth. The
author also carefully analyzed the published works related to routing techniques
in the MANET. Thence the author determines the research problem and the new
contributions of the thesis.
4


CHAPTER 2

EVALUATE QoT OF MANET IN THE CASE USING ON-DEMAND ROUTING
AND LOAD BALANCING ROUTING PROTOCOLS

2.1. Physical effects happen on the data transmission routes
2.1.1. Related technical factors
The physical effects happen on the data transmission routes depend on the technical
solutions used at the physical layer and data link layer, such as modulation formats,
wireless communication standards.
2.1.2. Path loss [20]
Lf =

2

4πd
λ

=

4π fc d
c

2

(2.2)

where fc is the frequency of the carrier, c is the speed of the light and d is distance.
2.1.3. Noise accumulates on the transmission routes
There are four noise components generated during data transmission, thermal noise,
interference noise, crosstalk and impulse noise. For MANET, the noise component
that most affects on QoT is thermal noise with the power is given by:

Pn = K × T × B

(2.5)

where K is Boltzmann constant, T is temperature and B is channel bandwidth.
2.2. Performance of MANET
In a general sense, the network performance is the efficiency, capacity and quality
of a network. The evaluation of the network performance is the determination of
measures that reflect the effectiveness, capacity and quality of a network system
by methods of simulation, analytical or experimental. In MANET, metrics commonly used to evaluate performance include packet blocking probability, delay,
throughput, signal-to-noise ratio and error bit rate.
2.2.1. Blocking Probability of Data packet (BPD)
BPD = Nb /Ng

(2.6)

where Ng and Nb are number of data packets are generated and are blocked, respectively. Nb includes blocking due to congestion and QoT constraint unsatisfactory.
2.2.2. Delay end to end
Delay end to end is the summation of time taken by a data packet to travel from
source to destination.
5


2.2.3. Signal to Noise Ratio (SNR)
In MANET, SNR depends on the relay type of the intermediate nodes. There are
two relay types which are amplify and forward (AF) and decode and forward (DF).
SNR of a route depends on these forward types, is determined by [9, 65]:

min(βh1 , βh2 , .., βhn−1 )
if DF

(2.8)



−1
n−1
βn =
1

otherwise
(2.20)

∑ βh

i
i=1
where βn is the SNR at the destination node and βhi is the SNR of the i hop.
2.2.4. Bit Error Ratio (BER)
BER is the number of bit errors per number of transmitted bits. Dependence of
BER versus SNR according to modulation formats is determined by [11].
2.3. QoT of the routes when using on-demand routing protocol
2.3.1. The basic principle of on-demand routing protocol
The principle of on-demand routing protocol is that routes will be discovered according to the requirement [3]. When a node requests a new route, it must initiate
a route discovery process. This process is only completed when a new route is
found or all possible routes have been checked. There are two on-demand routing
protocols in common research, which are DSR [22] and AODV [16].
2.3.2. QoT of the routes when using on-demand routing protocols
Node Next Hops

A

A
1
Node Next Hops
According to the principle of onA
B
2
B
demand routing protocols, there are
24
D
Node Next Hops
H
E
3
28
some cases that the route found
32
35
A
Node Next Hops
does not satisfy the constraints of
A
A
1
24
H
C
2
29
QoT. Considering an example as

C
31
32
28
E
shown in Figure 2.16 with AODV
29
Node Next Hops
A
E
2
H
protocol is used. Assuming that A
H
H
1
29
F
want to discover a route to H. For
32
Node Next Hops
Node Next Hops
32
A
C
3
I
AODV, the found route is A → E
A
A

1
31
G
Node
Next
Hops
RREQ
Node Next Hops
→ C → H. Assuming the relay type
A
G
3
A
E
2
RREP
of the nodes is AF. According to
Figure 2.16. An example of the route discovery
(2.20), SNR of route A → E → C
using AODV routing protocol
→ H is 23.87 dB. Considering that
minimum required SNR is 24 dB, this route does not satisfy the constraint of QoT.
For the topology as shown in Figure 2.16, from A32 to H
can use the route A → E
32
32

6

10

32

32


→ G → I → H. Although hopcount of this route is 4, SNR of that is 24.1 dB. This
value is better than SNR of the route A → E → C → H that AODV found.
2.4. QoT of the routes when using load balancing routing protocols
2.4.1. The principle of load balancing routing technique
Load balancing routing is the routing technique in which the route selection criterion is the uniform load traffic distribution
across all connections in the network.

B

24

D

28

A

32

35
24

2.4.2. QoT of the routes

29
31

C
E

32
28

29
Consider an axample of the route discovery as shown in Figure 2.17 with the FMLB
29
F
32
load balancing routing algorithm [70] used,
32
I
31
G
K is set to 3. Considering case A wants to
RREQ is continued to broadcast
transmit data to H. According to the princiRREQ is discarded
ple of route discovery by broadcasting the
RREP is replied to source node
RREQ packets, three routes found are A →
Figure 2.17. An example of load
E → C → H, A → E → G → I → H and
balancing routing in MANET network
A → B → D → H. SNR of the routes are
23.86, 24.04 and 20.2 dB, respectively. Thus, only the second route satisfies QoT

constraint. Meanwhile, all three routes are used. Therefore, data packets are transmitted on the first route and the third route with non-guaranteed QoT.

2.5. Evaluate QoT and network performance using simulation method
2.5.1. Simulation scenarios
To evaluate QoT of the data transmission routes and its effect on the MANET
performance, the author has simulated based on OMNeT++ [10].
Table 2.5. Simulation parameters
Parameters
Network Size
Modulation format
MAC protocol
Number of nodes
Transmit Power
Receiver Sensitivity

Setting
1000m × 1000m
256-QAM
802.11ac
From 20 to 50
19.5 dBm
-68 dBm

Parameters
BER threshold
Required SNR
Noise model
Temperature
Transmission Range
Speed of nodes

Setting
10−6
23.5 dB
Thermal noise
3000 K
250 m
5 - 20 m/s

2.5.2. Simulation results of DSR protocol
The result in Figure 2.19 shows the SNR at the receiver of the destination node.
There are many routes that does not satify the constraint of QoT since its SNR is
less than required SNR. This is the cause of the increasing BPD in the network.
7

H


26.00

26.00

Required SNR
DSR

25.00

SNR nhỏ nhất (dB)

Minimum SNR (dB)

25.00

24.00

23.00

22.00

24.00

23.00

22.00

21.00

21.00
20

25

30

35

40

45

50

20

Network size (nodes)

Figure 2.19. SNR of the routes in case of
DSR protocol

Figure 2.21. Minimum SNR in case
of DSR protocol

The existence of many routes that do not satisfy QoT constraint has increased BPD as
shown in Figure 2.24. BPD due to QoT is not
satisfied to account for nearly 50% of the total BPD.

BPD overall

0.04

BPD

BPD

0.03

0.02

0.01

0.01
0.00

0.00
0.6

0.7

0.75

0.8

0.85

0.9

0.95

1

Figure 2.24. BPD versus traffic load
in case of DSR protocol

0.08

BPD toàn phần

BPD overall
0.07

BPD do QoT

0.06

0.06

0.05

0.05

BPD

0.03

0.02

0.02

0.01

0.01

0.00
15

20

Tốc độ di chuyển (m/s)

Figure 2.29. SNR of routes in case of

AODV protocol

BPD due to QoT

0.04

0.03

10

0.65

Traffic load (Erlang)

0.04

0.00
5

10

15

20

Mobility speed (m/s)

Figure 2.31. BPD versus mobility
speed of AODV protocol

2.6. Conclusion of chapter 2
Chapter 2 presents the research results about the physical effects happening on
the data transmission routes and its impact on MANET network performance. The
simulation results have proved that, these effects is the cause of BPD increase, leading to the reduction of network performance. Therefore, it is essential to improve
routing algorithms to ensure QoT and improve network performance.
8

0.03

0.02

0.08

5

0.05

0.04

For AODV, SNR of the routes as shown in
Figure 2.29. There are many routes that does
not satify the constraint of QoT (is less than
23.5 dB). This is the cause of increasing
BPD, this is clearly visible from Figure 2.31.

BPD

BPD due to QoT

0.05

2.5.3. Simulation results of AODV

0.07

0.06

0.06

0.6


CHAPTER 3
LOAD BALANCING ROUTING ENSURING QUALITY OF TRANSMISSION
BASED ON TRAFFIC LOAD OFFERS TO EACH ROUTE

3.1. Introduction
The research results in Chapter 2 have shown that, load balancing routing can be resolved traffic bottleneck in the network. However, it can decrease QoT because the
routes may pass through multiple hops. To ensure the QoT of the data transmission
routes, several works have proposed routing algorithms that take into account the
constraints of some QoT [5, 24, 46, 58], where the proposed algorithms attempt
to find out the best QoT route. This therefore improves the QoT in the network.
However, for the mesh topologies such as MANET, the routing technique with the
best QoT can increase the bottlenecks due to unbalanced traffic load.
Thus, one problem to consider is how to combine harmony between QoT constraint
routing and load balancing routing, to find a set of routes that load traffic distribute
balancedly for all links, while satisfying the constraint of QoT as shown in Figure
3.2. For this idea, the author proposes a load balancing routing algorithm, while ensuring QoT of the routes. The load balancing route is chosen based on information
about probability of blocking packets from source to destination. The proposed
algorithm is called LBRQT (Load Balancing Routing ensuring QoT).

Shortest path
or best QoT
routing

Load balancing
routing

Traffic load distributes
unbalancedly to all
connections

There are some long
routes (pass through
multiple hops)

Bottlenecks
Load balancing
routing under
constrain of QoT
Decreasing QoT

Figure 3.2. The idea of proposing load balancing routing under QoT constraints

3.2. Relevant theory
3.2.1. Analyze the blocking probability of data packet using queue theory
Considering a hop j (hi j ), assume that the data packet arrivals follow Poisson distributions, the packet transmission times are exponentially distributed. Thus hi j is
modeled as M/M/1/L queuing [6, 63]. By solving the steady-state balance equation,
we determine BPD on hi j as follows:
 L


 ρi j (1 − ρi j ) if ρi j = 1
(h)
1 − ρiL+1
Bi j =
(3.4)
j

 1
if ρi j = 1
L+1
9


where λi j and µi j are the arrival and service rates of data packets, ρi j = λi j /µi j is
(r)

traffic density distributed to hi j . Let Bsd is BPD of route rsd , we have
(r)

Bsd = 1 −

∏

(h)

(1 − Bi j )

(3.7)

∀hi j ∈rsd

3.2.2. Analyze end-to-end delay based on the queue theory
End-to-end delay (EED) of a route is determined by:
(r)

τsd =

∑

(h)

(3.9)

τi j

∀hi j ∈rsd
(h)

where τi j is delay of hi j , consists of four components which is the processing
(i)

(i)

(i j)

delay (τ p ), queuing delay (τq ), transmission delay (τt
delay

(i j)
(τr )

[18]. Since

(i)
τp

and

(i j)
τr

) and radio propagation

are small enough, they are able to ignore,

(i j)
τt

(i)

is determined based on the bit rate of the channel and data packet size, τq is
determined based on the queue mechanism at the network nodes. As analyzed in
(i)
Section 3.2.1, M/M/1/L queue mechanism is used, so τq is determined by [19]:
L

(i)

τq =

(h)
λi j (1 − Bi j )

+

1
µi j

(3.11)

where L is the average length of the queue, determined by [19].
3.3. The idea of the proposed algorithm
3.3.1. Analytical model
The idea of proposing LBRQT algorithm is to combine balancing routing and QoT
constraint routing. To implement this idea, the objective function is to minimize
BPD on each route. The constraint are defined including QoT and EED.
In order to formulate LBRQT routing algorithm, the author defines a matrix Xsd =
(sd)
xi j n×n which is the matrix denoting the links of the route rsd , where each ele(sd)

ment xi j

is determined by
(sd)

xi j

=

1 if rsd passes through ci j

0 otherwise
(sd)

Therefore, the equation (3.7) is denoted according to xi j
(r)

n

n

(sd) (h)

Bsd = 1 − ∏ ∏ (1 − xi j Bi j )
i=1 j=1

10

(3.12)

as follows:
(3.13)


Thence, LBRQT algorithm is modeled to nonlinear programming problem:
(r)
Miniminze (Bsd )
(3.19)
Subject to the following constraints due to:



if j = s
−1
(sd)
(sd)
x
−
x
=
(3.20)
if j = d
∑ i j ∑ jk  1
 0
i∈N
k∈N
otherwise
N

N

∑∑

(sd) (h)

x i j τi j

≤ τth

(3.21)

i=1 j=1


N N

1



∑
∑
(h)

i=1 j=1 βi j xi(sd)
j

≤



(h)


min β
≥ βreq

x(sd) =1 i j

1
βreq

if AF is used
(3.22)
otherwise

ij

(sd)
(sd)
(xi j − 1)xi j

=0
(3.23)
The constraint conditions of (3.20), (3.21), (3.22) and (3.23) are the flow conservation, EED delay, QoT and integer constraints, respectively.
3.3.2. The idea of implementing LBRQT algorithms ussing cross-layer model
3.3.2.1. Modify the node structure using cross-layer model
To be able to use information about
Node j
QoT for routing constraints, the netTransport
Predicting the
Update the
work layer must be able to directly
parameters of
database of
performance
traffic
density
access to the information of the
Network
physical layer. This can only be performed by using cross-layer model
[2, 5, 26]. In LBRQT algorithm,

MAC
SA
the cross-layer model is proposed as
shown in Figure 3.6, where an staData
Physical
RREQ
tionary agent (SA) is used for the
SA: Stationary Agent
exchange of the information of QoT
Figure 3.6. Cross-layer model uses for the
between physical and network layLBRQT algorithm
ers. The tasks if the SA includes: (i)
updating traffic load for the connections in the network, and (ii) predicting the performance parameters which include the blocking probability of the data packets,
SNR of a route and EED. The information of QoT and EED are used for routing
constraints according to (3.21) and (3.22). The information of BPD is used for the
11


criteria of selecting the load balancing route according to the objective function
(3.19) by source node.
3.3.2.2. Improve the processing RREQ and RREP at each node
(i) RC of the intermediate node does not have a valid route to destination
S
This idea is illustrated as Fig. 3.7.
SA at node I predicts QoT, EED and
K
BPD from S to each neighbor of node I
When node I receives an RREQ
packet of route discovery request
L

RREQ
RREQ
from S to D, SA at I predict the
I
….
.
.
.
.
Data Packet
measurements of QoT and EED (a)
.
.
RREQ
M
from S to each neighbor of I. Then,
SA at I statistics the load traffic offering
to link from I to the next node
SA determines the set Qi is a set of
The set of all neighbors of node I
P
neighboring nodes of I that satisfy
The set of all neighbors of node I satisfies the
the QoT constraints. Thence node I
constraint conditions of QoT and EED (Set Q )
only broadcast RREQ to the nodes
Figure 3.7. Principle of process RREQ when RC
S
anddestination
EED from S to D

andIEED
to D
D
of set Qi . In addition, after deterof QoT
node
hasfrom
noS route
toQoT
the
don’t satisfy the given
satisfy the given constraint
constraint
conditions
conditions
mining set Qi , SA at I also predicts
BPD from S to each node of set Qi . This BPD is used for source node
to select a
RREQ
….
(b)
RREP
I
….
load balancing route. The set Qi is determined by Algorithm 3.1.
i

RREQ

RREQ

Algorithm 3.1: Finding set of neighbors of I satisfying constraints of QoTRREQ
(Set QLi )
(1)
(2)
(3)
(4)

(r)

(r)

SA at I predicts QoT, EED and BPD from S to

D along the route S  I joins I  D
Read the information of (βsi and τsi ) in RREQ;
Qi ← 0/ ;
for ((each node J is the neighbor of node I) do
(h)
Collect the information SNR from I to J (βi j ) at physical layer;

(h)

(5)

Predict EED from I to J (τi j ) according to (3.9);

(6)

τs j ← τsi + τi j ;
if ((Relay type of the nodes is DF) then

(r)
(r)
(h)
βs j ← min(βsi , βi j );
else

(r)

(r)

(7)
(8)
(9)

(r)

(h)

(h) −1

(r)

βs j ← 1/βsi + 1/βi j

(10)

;

(11)

end

(12)

if ((τs j ≤ τth ) and (βs j ≥ βreq )) then
(r)
Read information BPD from S to I (Bsi ) in RREQ;

(h)

(13)

(h)

(h)

(14)

Predict BPD of hop from I to J (Bi j ) according to (3.7);

(15)

Bs j = 1 − (1 − Bsi )(1 − Bi j ); Qi ← Qi ∪ J;

(r)

(h)

end

(16)
(17)

(r)

end

12

M


The set of all neighbors of node I

P

The set of all neighbors of node I satisfies the
constraint conditions of QoT and EED (Set Qi)

(ii) RC of the intermediate node has a valid route to destination
Figure 3.8 illustrates the idea of imS
QoT and EED from S to D QoT and EED from S to D
don’t satisfy the given
satisfy the given constraint
proving RREQ processing at each node
constraint conditions
conditions
when the intermediate node’s RC has
RREQ
….

a valid route to the destination node. (b)
RREP
I
….
Assuming the current node is I, in this
RREQ
RREQ
L
case, node I does not immediately creRREQ
SA at I predicts QoT, EED and BPD from S to
ate RREP and reply to S as the onD along the route S  I joins I  D
M
demand routing protocol. Instead, the
SA at I predict QoT and EED from S to Figure 3.8. Principle of process RREQ when
RC of node I has a route to the destination
D along the route S → I join with I →
D. If predicted QoT and EED satisfy the given constraints, RREP is created and
reply to source node. In contrast, node I proposes RREQ as case (i).
Algorithm 3.2: Predict QoT and BPD by SA when RC of I has a route to D.
(r)

(r)

(r)

(r)

(1)

Read information of QoT and EED from S to I (βsi and τsi ) in RREQ;

(2)

Read information of QoT and EED from I to D (βid and τid ) in RC of I;

(r)
(r)
(r)
(3) τsd ← τsi + τid ;
(4) if (Relay type of the nodes is
(r)
(r)
(r)
(5)
βsd ← min(βsi , βid );
(6) else
(r)

(r) −1

(r)

βsd ← 1/βsi + 1/βid

(7)

DF) then

;

(8)

end

(9)
(10)

if ((τs j ≤ τth ) and (βs j ≥ βreq )) then
(r)
Read information of BPD from S to I (Bsi ) tin RREQ;

(11)

Read information of BPD from I to D (Bid ) in RC of I;

(12)

Bsd = 1 − (1 − Bsi )(1 − Bid ); Create RREP, store Bsd into RREP;

(13)

(h)

(r)

(r)

(r)

(r)

(r)

else
Find set Qi according to Algorithm 3.1;

(14)
(15)

(h)

end

3.3.2.3. Improve the route selection mechanism at the source node
For the improved process of RREQ and RREP as Section 3.3.2.2, if a route is
found, this route always satisfies the constraints of QoT. The remaining problem
of the LBRQT algorithm is to choose a load balancing route. This is done at the
source node. According to the principle of the LBRQT algorithm, the criterion for
selecting a route is to minimize BPD according to the objective function (3.19).
Therefore, when the RREP packet is received, the source node selecting the route
with the minimum BPD value.
13

D


3.4. The operation principle of LBRQT algorithm
Intermediate node

Source node

S creates
RREQ

Start

I=S

Destination
node

For each J  Qi

Determine Qi
according to
Algorithm 3.1

I=J
No

Qi 

Yes
Node I broadcast
RREQ to all node
J  Qi

No
Discard No
RREQ

NRREP = 0;
Twait = 0;
No

Yes

Yes

No

S selects route with
minimum BPD

Qi 

End

No

RREP is
created?

Yes

NRREP = NRREP + 1

Yes

Node I broadcast
RREQ to all node

J  Qi

S receives RREP

Send
RREP to S

Predict QoT and
BPD according to
Algorithm 3.2

Determine Qi
according to
Algorithm 3.1

(NRREP = K) OR Sai
(Twait > Timeout)

Yes

RC of I has a
route to D?

Increase Twait

NRREP > 0

D create
RREP

I not yet received
this RREQ?
Yes

Reject request because the
route could not be found
No

Yes

I is destination (D)

Send
RREQ to S

Figure 3.9. Flowchart of LBRQT routing algorithm

3.5. Apply for AODV protocol
3.5.1. Introduction
The research results in Chapter 2 have shown that, for the discovery principle of
AODV, there are some cases where the route found does not satisfy the QoT constraint. To solve this problem, the author applied the LBRQT algorithm to improve
the route discovery mechanism of the AODV protocol [16], in order to find the load
balancing route, while satisfying the QoT constraints. The improved algorithm is
named LBRQT-AODV. This proposal of the author has been published in [B2]1 .
3.5.2. Modify the format of RREQ and RREP packets
32 bits
(1)

(2) (3) (4) (5) (6)

Type

32 bits
(8)

(9)

J R G D U Reversed CF Hop Count

(10)

(a)

(7)

RREQ ID

(11)

Destination IP Address

(12)

Destination Sequence Number

(b)

(1)

(2) (3)

(4)

(5)

(6)

Type

J R

Reversed

Prefix

Hop Count

(7)

Destination IP Address

(8)

Destination Sequence Number
Originator IP Address

(13)

Source IP Address

(9)

(14)

Source Sequence Number

(10)

Reversed

(15) BP

(16) QoT

(17) EED

Lifetime
Reversed

(11) BP

Figure 3.11. Format of (a) RREQ and (b) RREP packets in LBRQT-AODV
1 Journal

of Communications, Vol.13, No.7, 2018, pp. 338-349 (SCOPUS).
14


3.5.3. LBRQT-AODV algorithm
Algorithm 3.3: LBRQT-AODV algorithm at source node

(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)

S creates RREQ;
SA determines Qs according to 3.1;
if (Qi = 0)
/ then
Broadcast RREQ to all nodes in Qs ;
Wait until receives K of RREP packets or over timeout;
if (Number of received RREP packets > 0) then
Select the route with BPD value in RREP is the smallest RREP to update
into the RC of S;
else
Reject route discovery request;
end
else
Reject route discovery request;
end

Algorithm 3.4: LBRQT-AODV algorithm at intermediate or destination nodes
(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)

Node I receives RREQ;
if (I is intermediate node) then
if (I haven’t received this RREQ package before) then
Update the reverse route to S into the RC of I;
if ((RC of I don’t have a valid route to D) then
SA determines Qi according to Algorithm 3.1;
if (Qi = 0)
/ then
Broadcast RREQ to all nodes in Qs ;
else
Discard RREQ and End the processing RREQ;
end
else
if (DSN of route I → D is greater than DSN in RREQ) then
SA predicts QoT, EED and BPD along route S → I join I → D
according to 3.2;
if (RREP is created) then
Send RREP to S according to the reverse route;
else
Run the steps from 6 to 11;
end
else
Run the steps from 6 đến 11;
end
end
else
Discard RREQ and End the processing RREQ;
end

else
Update the reverse route to S into the RC of I;
Create RREP, send RREP to S according to the reverse route;
end

15


3.6. Apply for DSR protocol
3.6.1. Introduction
The research results in Chapter 2, for the discovery principle of DSR, there are
some cases where the route found does not satisfy the QoT constraint. To solve this
problem, the author applied the LBRQT algorithm to improve the route discovery
mechanism of DSR protocol. The improved algorithm is named LBRQT-DSR.
3.6.2. Modify the format of RREQ and RREP packets
The RREQ and RREP of the LBRQT-DSR are modified as shown in Figure 3.12.
3.6.3. LBRQT-DSR algorithm
Algorithm 3.5: LBRQT-DSR algorithm
(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)
(31)
(32)
(33)
(34)
(35)

S creates RREQ; I ← S; NRREP = 0;
repeat
Determine Qi according to Algorithm 3.1;
Broadcast RREQ to all node J in Qi ;
if (J has not received this RREQ before) then
Add a record to the RC of J containing the reverse route to S;

if (J is not destination (D)) then
if (RC of J don’t have a route to D) then
Update the reverse route to S into RC of J;
Update the route from S to J into RREQ;
I ← J;
else
SA at J predicts QoT, EED and BPD according to S → I join I → D
according to Algorithm 3.2;
if (RREP is created) then
Join route S → J to J → D;
NRREP ← NRREP + 1; Send RREP to S according to reverse route;
else
Update the reverse route to S into RC of J;
Update the route from S to J into RREQ;
I ← J;
end
end
else
Create RREP; Update the route S → D into RREP;
NRREP ← NRREP + 1; Send RREP to S according to reverse route;
end
else
Discard RREQ and End the proposing RREQ;
end
until (NRREP = K) or (over timeout);
if (NRREP > 0) then
S selects a route with BPD value in RREP is the smallest;
else
Reject the route discovery request from S to D;
end

16


Opt. type (*) Opt. Data Length (*) Identification (*)
Target Address (*)
Address [1] (*)
Address [2] (*)
(a)
… (*)
Address [n] (*)
Reserved
BP (**)
QoT (**)
E2E (**)

(b)

Opt. type (*) Opt. Data Len (*) Last Hop Ext. (*) Reserved (*)
Address [1] (*)
Address [2] (*)
Address [3] (*)
… (*)
Address [n] (*)
Reserved
BP (**)

Figure 3.12. Format of (a) RREQ and (b) RRREP in LBRQT-DSR algorithm

3.7. Simulate and analyze results

3.7.1. Simulation scenario
LBRQT-AODV and LBRQT-DSR algorithms are evaluated by simulation on OMNeT ++ [10], compared to AODV [16], DSR [22] and DSR-SNR algorithms in
[24]. The simulation scenario is set as Section 2.5.1, chapter 2.
3.7.2. Simulation results of LBRQT-AODV algorithm
Figure 3.13 compares SNR
of routes using AODV and
LBRQT-AODV in the case of
the 50 nodes topology, average mobility speed is 10 m/s.
We can observe that there
are many routes that do not
Figure 3.13. Compare SNR of (a) AODV and (b)
satisfy the QoT constraints.
LBRQT-AODV
For LBRQT-AODV, SNR has
been improved. Most of SNRs
are greater than required SNR (23.5 dB).
0.05

AODV

LBRQT-AODV

0.04

BPD

0.03

For throughput, LBRQT-AODV is also more efficient than the AODV algorithm. This is clearly
shown in Figure 3.18, corresponding to the case

where the number of nodes is 40, mobility speed
17

0.02

0.01

0.00
0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Traffic load (Erlang)

Figure 3.17. Compare BPD of
AODV and LBRQT-AODV

76E+6
74E+6
72E+6

Throughput (bit/s)

As the SNR of LBRQT-AODV algorithm improved, BPD decreased as shown in Figure 3.17.
This result is simulated on the 40 nodes topology, the average mobility speed of each node is
5 m/s. When the traffic load is 0.6 Erlang, the
BPD of the AODV is 0.0136. Meanwhile, this
value of LBRQT-AODV is only 0.0091. Thus,
BPD of LBRQT-AODV decreased by 33.21 %
compared to AODV.

70E+6
68E+6
66E+6
LBRQT-AODV

64E+6

AODV

62E+6
0

50

100

150

200

250

300

350

400

450

Simulation time (s)

Figure 3.18. Compare throughput
of AODV and LBRQT-AODV


5 m/s. The average throughput of the AODV and LBRQT-AODV algorithms are
69.85 and 71.55 Mbit/s, respectively. Thus, compared with the AODV algorithm,
the throughput of the LBRQT-AODV algorithm increases by 1.7 Mbit/s.
3.7.3. Simulation results of LBRQT-AODV algorithm
Figure 3.20 shows the minimum SNR of routes.
For DSR, SNR is greater than required SNR
when the number of nodes is less than 30. However, if the number of nodes is greater than
30, the SNR is smaller than required SNR. For
LBRQT-DSR, SNR has been improved, always
greater than required SNR despite the number

of nodes is large. For BPD, when using LBRQTDSR, BPD is also improved compared to DSR
(Figure 3.23). BPD of LBRQT-DSR decreased
on average 51.79 % compared to DSR.

3.8. Conclusion

Required SNR
DSR

Minimum SNR (dB)

25.00

LBRQT-DSR

24.00

23.00

22.00

21.00
20

25

30

35

40

45

50

Network size (nodes)

Figure 3.20. Minimum SNR of
LBRQT-DSR and DSR
72E+6

0.07
DSR

0.06

70E+6

LBRQT-DSR

Throughput (bit/s)

0.05

BPD

In terms of throughput, LBRQT-DSR always achieves a higher
throughput than the DSR
algorithm (Figure 3.26).

LBRQT-DSR algorithm
yields higher throughput than the average
DSR by 2.99 Mbit/s.

26.00

0.04
0.03
0.02
0.01

68E+6

66E+6

64E+6

62E+6

LBRQT-DSR
DSR

0.00
0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

Traffic load (Erlang)

Figure 3.23. Compare BPD of
LBRQT-DSR and DSR

1

60E+6
0

50

100

150

200

250

300

Simulation time (s)

Figure 3.26. Throughput of
LBRQT-DSR and DSR

Chapter 3 presented the load balancing routing algorithm ensuring quality of transmission (LBRQT), proposed for MANET. LBRQT algorithm finds the route that
satifies the QoT constraints, while balancing the traffic load across all connections. The LBRQT algorithm has been applied to improve the AODV routing protocols (LBRQT-AODV) and DSR (LBRQT-DSR). Simulation results on OMNeT
++ showed that the algorithms LBRQT-AODV and LBRQT-DSR have found the
routes that satify the constraints of QoT, so QoT of the data transmission routes is
always guaranteed. In addition, the routes are also selected according to the load
balancing criteria. Therefore, minimize local congestion in the network. Therefore,
the network performance is improved compared to DSR and AODV algorithms, especially in the case of network systems with large area area and high node density.
18


CHAPTER 4
LOAD BALANCING ROUTING ENSURING QUALITY OF TRANSMISSION
BASED ON THE ROUTE INFORMATION OF THE SOURCE NODE

This chapter presents the load balancing routing algorithm ensuring QoT, is proposed for MANET. The load balancing route is chosen based on the routing information stored in the route cache of the source node. The proposed algorithm is
named SLBQT-DSR (Source-based Load Balancing ensuring QoT based on DSR).
4.1. The idea of the proposed algorithm
4.1.1. Choose the load balancing route
The basic feature of DSR protocol is that the route cache of each node stores detailed information of each route from source to destination. Thus each node can
determine traffic load from it distributed to all connections in network based on
routing information in its route cache. Thence, when source node receives RREP
for route discovery results, based on routing information in its route cache, the
source node can select a route so that traffic load distributes to all connections is
most balanced. This is the idea of selecting load balancing route of SLBQT-DSR.

4.1.2. Determine the constraints of QoT
To ensure the QoT of routes
Predicting SNR, EED
from S to J  NI
found by SLBQT-DSR algoN
rithm, the conditions of QoT
constraint must be determined
S
L
RC of I doesn’t have a valid route to D
during the route discovery.
RREQ
The idea of determining the
...
M
A
I
constraint conditions of QoT
...
...
RC of I has a valid route to D
in SLBQT-DSR algorithm is
K
illustrated as in Figure 4.1.
Set Q
When an intermediate node
Set N
D
Predicting SNR, EED
(I) receives a RREQ of a refrom S to D

quest to discover a new route
from the source (S) to the des- Figure 4.1. Mô hình xác định điều kiện ràng buộc QoT
tination (D). In the case that
của thuật toán SLBQT-DSR
routing information in route
cache of I not have a route to D, SA at I predicts information about QoT from
S to all neighboring nodes of I (nodes in the set NI ). Then RREQ is broadcast
only to neighbors nodes of I that satisfy QoT constraint condition (nodes in set
I

I

19


QI ). In case the routing information in the route cache of node I has a route to
D, instead of sending RREP back to the source node as DSR, SA at I predict the
information about QoT from S to D along the route S → I joins I → D. If the given
QoT constraint is satisfied, node I then sends the RREP to the source. In contrast,
the route discovery process is continued as the case the routing information in the
route cache of node I has no route to D.
For this principle, the routes found always satisfy the constraints of QoT. The principle of predicting the QoT parameters at node I by SA is implemented based on
the cross-layer model as described in Section 3.3.2.1 of Chapter 3.
4.2. Analytical model for SLBQT-DSR
In order to formulate for SLBQT-DSR algorithm, the following symbols and notations are assumed:
(sx)

Defining Nsx = ni j

n×n

as a matrix denoting the links of the route from node S
(sx)

to node X (rsx ), where each element ni j is determined by
(sx)

ni j =

1 if rsx passes through connection ci j ,

(4.1)

0 otherwise.
(s)

Letting ρsx as the traffic load offers from node S to node X, Fs = fi j n×n as a
matrix denoting the traffic load from node S distributes to all connections in the
network. Thus Fs is determined by
(s)

Fs = fi j

m|x=s

=
n×n

∑

ρsx Nsx .

(4.2)

x=1

Consider the case of the node S want to discovery a new route to the node D. The
SLBQT-DSR algorithm will broadcast the RREQ packet to discovery the K routes
satisfying the constraints of QoT and end-to-end delay (EED). K found routes are
(k)
(sdk)
(sdk)
denoted by a matrix Nsd = ni j
, where each element ni j is determined
n×n
according to (4.1).
In order to denote the load balancing route which is selected in the K available
(k)
routes, we define the variable xsd as follows
(k)

xsd =

1 if the route kth is selected,
0 otherwise.

(4.3)

Thence the matrix that denotes the traffic load from the node S distributes to all
20

connections in the network being transformed into:
(s)

Fs = fi j

K

= Fs + ρsd
n×n

(k)

(k)

∑ xsd Nsd .

(4.4)

k=1
(s)

From (4.4) we have the element fi j of the matrix Fs is determined by
(s)

K

(s)

fi j = fi j + ρsd

(k) (k)

∑ xsd nsd .

(4.5)

k=1

After determining the Fs matrix, SLBQT-DSR algorithm is formulated as the following linear integer progeamming (ILP) problem
min

(s)

(4.6)

max fi j

(s)
∀ fi j ∈Fs

subject to the following constraints due to:
(k)

(k)

xsd (xsd − 1) = 0

K

(4.7)

(k)

∑ xsd

=1

(4.8)

k=1

where (4.7) is the binary and integer constraint according to define the variable
(k)
xsd as (4.3), (4.8) is the constraint of the route selection.
4.3. Implement SLBQT-DSR algorithm
4.3.1. Modify the format of RREQ

Opt. type

Opt. Data Length Identification
Target Address
Address [1]
Address [2]
…
Address [n]
Reserved
QoT
EED

In SLBQT-DSR algorithm, the author
uses RREQ to exchange information
about QoT and EED between nodes. The
structure of the RREQ is shown in Fig(a)
Figure 4.3. Format of RREQ in
ure 4.4. This RREQ was modified from
SLBQT-DSR algorithm
the RREQ of the DSR protocol by adding
QoT and EED fields to store the values of quality of transmission and delay, used
to identify constraints during the route discovery.
4.3.2. Flowchart of SLBQT-DSR algorithm
The principle of discovering the route of SLBQT-DSR routing algorithm is implemented according to the flowchart in Figure 4.4. The QoT constraints are defined
in steps (3) to (5) for the source node, steps (11) to (16) for the intermediate node,
in which, defining the set Qi is a set of neighboring nodes of node I satisfying the
constraint condition of QoT, implemented according to 3.1 of Chapter 3. When the
source node has received K of RREP packets of route discovery result, meaning
SLBQT-DSR algorithm has found K routes satisfying QoT constraint condition,
SLBQT-DSR algorithm will select one of the K routes available so that traffic
21

Opt. typ


load is evenly distributed to all connections in the network. This is done at step
(27) of the source node, according to Algorithm 4.1.
S creates
RREQ

Start

Update the route
into route cache
of node S

End

Reject request due to
do not find route

Determine Qs according
to Algorithm 3.1

S selects load
balancing according
to Algorithm 4.1
Yes
No
Nrrep > 0

No

Source node
Qi # 

No

Yes

Increase Twait

according to clock

(Nrrep = K) OR
(Twait > Timeout)

Nrrep = 0;
Twait = 0;

Nrrep = Nrrep+1

Node I broadcast RREQ to
all node J  Qs

Yes

S receives
RREP

Intermediate node

Node I receives RREQ

I is intermediate
node?
Yes

Yes

No

Update reverse
route to S into RC
of node I

I already received
this RREQ?
No

Create RREP and
send to S according
to reverse route

Update reverse route to S
into RC of node I

RC of I has a
route to D?

Yes

SA at I predict SNR and
EED from S to D along
route S → I join I → D

Destination
node

No
Determine Qi according
to Algorithm 3.1

Discard
RREQ

No

No

SNR and EED
satisfy constraints
of QoT?
Yes

Qi # 
Yes

Node I broadcast RREQ to
all nodes J  Qi

Create RREP and send to S
according to reverse route

End of processing
RREQ

Figure 4.4. Flowchart of SLBQT-DSR routing algorithm

22


Algorithm 4.1: Chọn một lộ trình cân bằng tải tại nút nguồn

(1)

Based on the information in the route cache of S, construct the traffic distribution

(2)

matrix Fs = fi j n×n according to (4.2);
Based on the information of K available routes, construct the traffic distribution

(s)

(s)

(3)

(4)
(5)
(6)

matrix Fs = fi j n×n according to (4.4);
Construct the ILP problem according to the objective function(4.6) subject to the
constraints of (4.7) and (4.8);
Solving the ILP problem;
Select the load balancing route based on the results of the ILP problem solving;
Update the information of the found route into the route cache of S;

4.4. Simulate and analyze results
4.4.1. Simulation scenario
The performance of SLBQT-DSR algorithm was assessed by simulation on OMNeT ++ [10]. The SLBQT-DSR algorithm is compared to the route discovery algorithm of DSR [22]. The simulation was performed in various scenarios with the
technique parameters set as Table 2.5 of Chapter 2.

26.00

4.4.2. Simulation results

Required SNR
DSR

The result in Figure 4.5 shows minimum
SNR of SLBQT-DSR and DSR. For DSR
algorithm, SNR is only greater than required SNR when number of nodes is less
than 30. For RLBQT-DSR, SNR is always
greater than required SNR even though
number of nodes is large. As SNR increases, BPD decreases as shown in Figure
4.8. For SLBQT-DSR, BPD decreased on
average by 46.77 % compared to DSR.

Minimum SNR (dB)

25.00

SLBQT-DSR

24.00

23.00

22.00

21.00
20

25

30

35

40

45

50

Network size (nodes)

Figure 4.5. Compare SNR of
SLBQT-DSR and DSR algorithms

0.07

74.0E+6

DSR

0.06

72.0E+6

SLBQT-DSR

70.0E+6

Throughput (bit/s)

BPD

0.05
0.04
0.03
0.02

68.0E+6
66.0E+6
64.0E+6
62.0E+6
60.0E+6
SLBQT-DSR

0.01

58.0E+6

0.00

56.0E+6

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

DSR

0

1

50

100

150

200

250

300

350

400

Simulation time (s)

Traffic load (Erlang)

Figure 4.8. Compare BPD of SLBQT-DSR
and DSR algorithms

23

Figure 4.11. Compare throughput of
SLBQT-DSR and DSR algorithms