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 Assoc Prof Dr Vo Thanh Tu Supervisor 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 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 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 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 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 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 presents the overview of MANET and the factors that affect on the network performance Chapter focuses on evaluating the quality of transmission of MANET network in the cases using the on-demand routing and load balancing routing protocols Chapter proposes a load balancing routing algorithm ensuring the quality of transmission based on the traffic load that offers to each route Chapter 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 ++ CHAPTER 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 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 Chapter 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 CHAPTER 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 = 4πd λ = 4π fc d c (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 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 =  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 Node Next Hops According to the principle of onA B B demand routing protocols, there are 24 D Node Next Hops H E 28 some cases that the route found 32 35 A Node Next Hops does not satisfy the constraints of A A 24 H C 29 QoT Considering an example as C 31 32 28 E shown in Figure 2.16 with AODV 29 Node Next Hops A E H protocol is used Assuming that A H H 29 F want to discover a route to H For 32 Node Next Hops Node Next Hops 32 A C I AODV, the found route is A → E A A 31 G Node Next Hops RREQ Node Next Hops → C → H Assuming the relay type A G A E 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 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 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 - 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 H 26.00 26.00 Required SNR DSR 25.00 SNR nhỏ (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 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 Figure 2.24 BPD versus traffic load in case of DSR protocol 0.08 BPD toàn phần BPD overall 0.07 BPD 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 10 15 20 Mobility speed (m/s) Figure 2.31 BPD versus mobility speed of AODV protocol 2.6 Conclusion of chapter Chapter 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 0.03 0.02 0.08 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 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   i∈N k∈N otherwise N N ∑∑ (sd) (h) x i j τi j ≤ τth (3.21) i=1 j=1  N N     ∑ ∑ (h)  i=1 j=1 βi j xi(sd) j ≤   (h)   β ≥ βreq  x(sd) =1 i j 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) (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 − 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 − 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 + 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 > 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 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 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 to 11; end else Run the steps from 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 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 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 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 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 50 100 150 200 250 300 350 400 450 Simulation time (s) Figure 3.18 Compare throughput of AODV and LBRQT-AODV 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 60E+6 50 100 150 200 250 300 Simulation time (s) Figure 3.26 Throughput of LBRQT-DSR and DSR Chapter 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 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 Mơ hình xác định điều kiện ràng buộc QoT tination (D) In the case that 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 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 = if rsx passes through connection ci j , (4.1) 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 = if the route kth is selected, 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 (s) (4.6) max fi j (s) ∀ fi j ∈Fs subject to the following constraints due to: (k) (k) xsd (xsd − 1) = 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 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 not find route Determine Qs according to Algorithm 3.1 S selects load balancing according to Algorithm 4.1 Yes No Nrrep > 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 lộ trình cân tải 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 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 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 In terms of throughput, because PBD of SLBQT-DSR decreases compared to DSR, the throughput increases This is shown in Figure 4.11 For DSR algorithm, the average throughput is 65,482 Mbit/s, while the average throughput of SLBQTDSR is 67,385 Mbit/s Thus, compared with DSR algorithm, the throughput of SLBQT-DSR algorithm increases an average of 1,903 Mbit/s 4.5 Conclusion of chapter In this chapter, the author presented the proposed load balancing routing algorithm QoT (SLBQT-DSR) for MANET network This algorithm finds the routes that satisfy the constraints of QoT and EED, and balances the traffic load of all links in the network The simulation results have proved that SLBQT-DSR algorithm achieves better network performance than DSR algorithm THE NEW CONTRIBUTIONS OF THE THESIS This thesis has three new contributions, including: The thesis has proposed a new method to determine the constraint of quality of transmission based on the cross-layer model This method is used for discovering the route of the on-demand routing protocols in mobile ad hoc networks Based on the traffic load offers each route, the thesis has proposed the Load Balancing Routing algorithm ensuring the Quality of Transmission (LBRQT) for mobile ad hoc networks Based on the characteristics of dynamic source routing protocol, the thesis has proposed the Source-based Load Balancing ensuring Quality of Transmission in DSR (SLBQT-DSR) for mobile ad hoc networks The contents of the future research Research integrated control plane between physical layer and network layer to implement QoT constraint routing protocols, using SDN technology Research routing protocols in multi-channels, multi-carriers MANET with consideration of the physical effects constraint Research to evaluate more measures of routing latency, routing costs and some other measures for algorithms proposed in the thesis Research and apply new traffic generation models, following the traffic trends of multimedia services, traffic in the IoT era to evaluate the efficiency of load balancing of proposed routing algorithms 24 Published papers of the author related to the thesis topic [B1] Le Huu Binh, Vo Thanh Tu and Nguyen Van Tam, “SLBQT-DSR: Sourcebased Load Balancing Routing Algorithm under Constraints of Quality of Transmision for MANET”, Journal of Computer Science and Cybernetics, Vol.34, No.3, 2018, pp 265-282 [B2] Le Huu Binh, Vo Thanh Tu, “QTA-AODV: An Improved Routing Algorithm to Guarantee Quality of Transmission for Mobile Ad Hoc Networks using Cross-Layer Model”, Journal of Communications, Vol 13, No 7, 2018, pp.338-349 (SCOPUS) [B3] Le Huu Binh, Vo Thanh Tu, Nguyen Van Tam, “Investigate the impact of physical effects and QoT routing techniques in MANET”, Proceedings of the XXI National Conference on Selected issues of Information Technology and Communications, Thanh Hoa, 27-28/7/2018, Science and Technology Publishing House, 2018, pp 162-169 [B4] Le Huu Binh, Vo Thanh Tu, Nguyen Van Tam, “Quality of Transmission Aware Routing in Adhoc networks based on Cross-Layer Model combined with the Static Agent”, Journal of Computer Science and Cybernetics, Vol.32, No.4, 2016, pp 351-366 [B5] Le Huu Binh, Vo Thanh Tu, Nguyen Van Tam, “A method of analyzing Adhoc network performance using analytical model”, Proceedings of the 10th National Science Conference on Fundamental Research and Applied Information Technology Research - FAIR’10, Đa Nang, 17-18/08/2017, Science and Technology Publishing House, 2017, pp 577-584 [B6] Le Huu Binh, Nguyen Đang Khoa, Nguyen Đinh Hoong Phuong, “Wireless Mesh Network Topology Design: A new method using the integer linear programming problem”, Journal of Research and Development on Information and Communication Technology, Vol V-3, No 18 (38), 2017, pp 58-66 [B7] Le Huu Binh, Vo Thanh Tu, Nguyen Van Tam, “A Cross-Layer Routing Algorithm Guaranteed QoT in MANET”, Proceedings of the 9th National Science Conference on Fundamental Research and Applied Information Technology Research - FAIR’9, Can Tho, 04-05/08/2016, Science and Technology Publishing House, 2016, pp 480-487 [B8] Le Huu Binh, Vo Thanh Tu, “Evaluating the impact of noise with respect to performance of MANET based on the on-demand protocols”, Proceedings of the 8th National Science Conference on Fundamental Research and Applied Information Technology Research - FAIR’8, Science and Technology Publishing House, 2015, pp 111-118 ... OMNeT ++ CHAPTER 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... 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... 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