Novel Design of Reliable Static Routing Algorithm For Multi-hop Linear Networks

Transcription

1 Novel Design of Reliable Static Routing Algorithm For Multi-hop Linear Networks A Thesis submitte in fulfilment of the requirements for the Degree of Doctor of Philosophy (Ph.D.) By Siva Kumar Subramaniam Department of Electronic an Computer Engineering College of Engineering, Design an Physical Sciences Brunel University, Lonon Unite Kingom August 2017

2 Abstract Abstract In the recent years, increasing eman on static multi-hop wireless sensor network (WSN) ha preominate the remote monitoring of oil an gas pipeline integrity. In a pipeline network, sensing points are connecte through wireless noes briging the remotely measure points to a centralise monitoring station. The eployment of a WSN on a pipeline network has crucial factors contributing to egraing of overall network performance that is proportional to the network ensity. Such geographically unique network architecture has a significant impact on estabilisation of network reliability, throughput unfairness, higher latency an energy consumption from the inaequate utilisation of network resources ue to competitive ata transmission that results in ata snowballing effect towars the estination noe. To unravel these factors, the first phase of this thesis has highlighte the Dual Interleaving Linear Static Routing (DI-LSR) for flat multi-hop topology an the Dual Cluster Hea Interleaving Linear Static Routing (DCHI-LSR) for a clusterbase multi-hop topology making it feasible on a scalable pipeline network with enhance network performance. Both DI-LSR an DCHI-LSR employs preefine interleaving routes with a beaconless network to achieve significant improvements in overall network performance with the available network resources on a pipeline network. The teste an analyse results in various simulation environment in accorance with IEEE stanar has enhance network reliability (elivery ratio) between 77-83% an capacity (throughput) in between Kbps without passive noes (active noes without transmitting opportunity) in a pipeline network. Noe starvation is another aspect that severely affects the overall network performance in a large scale multi-hop linear WSN. Inequality in network resources allocation among source noes contributes to noe starvation that is relatively an amplifie factor over generate ata packets an source noe istances from the estination noe. The secon phase of this thesis has highlighte a mathematical moel for calculating appropriate transmission control protocol (TCP) elaye acknowlegement timeout with reference to the I

3 Abstract istance between source to a estination noe (travel cost). The optimum throughput fairness on a flat topology network can be achieve by implementation of TCP elaye acknowlegement timeout moel for a fairness critical application (DATM-FFCA) an for a throughput critical application (DATM-FTCA). Whereas for a cluster-base topology network, the TCP elaye acknowlegement timeout moel for a fairness critical application (DATM-CFCA) an for a throughput critical application (DATM-CTCA) elivers optimum throughput fairness for all member noes. With the propose technique that overwrites the traitional TCP parameters ecreases packet collision an ensure optimum throughput fairness by eliminating passive noes in a multi-hop linear WSN. Results from simulation experiments have reveale that the propose moels are able to retain the network fairness rate at above 90% without compromising the performance characteristic on a scalable pipeline network. Energy optimisation in a multi-hop linear topology if often relate to the network lifetime an is critical on heterogeneous energy consumption noes in the network. The thir phase of this thesis has highlighte Active High Transmitterreceiver energy moel (AHiT) that is an aaptive sleep an wake-up interval for energy optimisation on noe level base on occurring events in a multi-hop WSN. The AHiT energy moel is in contrast to a traitional sleep an wake-up scheuling schemes that is solely epenent on ata traffic pattern of noes in a network. Noes aapt their sleep an wake-up interval base on ata traffic pattern that is generate by respective noes an also from its neighbouring noes. The noes locate in a multi-hop linear topology is connectivity critical to support ata transfer from the neighbouring noes. Simulations were carrie out to analyse the energy performance of the propose energy moel that has significantly reuce the energy consume above 40% at the noe level an prolonge the network lifetime between 4-100% subjecte to the ata traffic pattern. II

4 Acknowlegements Acknowlegements First an foremost, my most humble gratitue to Go. Only with His blessings an perseverance that was bestowe upon me in making this PhD journey possible. My special gratitue an appreciations to my PhD main supervisor, Dr Rajagopal Nilavalan who been a tremenous inspiration an mentor throughout my PhD journey. His eication, scholarly guiance an encouragement in my research ha refine my knowlege in the fiel of wireless communication. I am inee very prou an honoure for being a stuent uner your guiance. My eepest appreciation to my secon supervisor, Professor Dr Wamaeva Balachanran for his guiance an support in my acaemic as well as professional eneavours. My sincere gratitue to my family for their support, love an prayers throughout my PhD journey. Wors cannot express their sacrifices an inspiration that ha le me to what I am toay. Bawa, I starte this PhD journey with your blessing an etermination that kept me moving towars my goal till your last ays. I believe that wherever you are, you will be always prou an please that I ha finishe this with great success. My heartfelt gratitue to my loving wife Ms Kavitha Krishnan for her enless support in all my neee moments, being a motivator an showing her passion throughout my PhD journey. It was your ream to see me completing my PhD with flying colours. I know just how much my PhD means to you an to both our kis Divehssha SK an Jiethyaah SK. Thank you from the bottom of my heart for your unerstaning, scarifies an cooperation at times that I was unable to spare time for the family. I am grateful to Go for giving me such a wonerful family. I am gratefully inebte to Dr Shariq Mahmoo Khan an Late Professor Abul Hami Hamion for their valuable an generous support throughout my PhD journey. My sincere gratitue to all my friens an colleagues who ha contribute irectly or inirectly in completing my PhD thesis. I eicate this thesis to every single iniviual mentione above as well as for those who praye for my success. My profoun gratitue to my sponsors the Ministry of Eucation Malaysia (KPM) an Universiti Teknikal Malaysia Melaka (UTeM) who fune an facilitate me throughout my PhD egree. III

5 Table of contents Table of contents Abstract... I Acknowlegements... III Table of contents... 1 List of Figures... X List of Tables... XIII List of Abbreviations... XIV List of Publications... XVII Statement of Originality... XVIII Chapter 1. Introuction... 1 Introuction an Application of Multi-hop Linear Topology... 1 Remote Monitoring of Oil an Gas Pipelines... 1 Challenges of WSN in Oil an Gas Pipelines... 2 Motivation... 4 Objectives of the Thesis... 6 Research Methoology... 7 Contribution of Knowlege... 7 Thesis Organisation Conclusion Chapter 2. An Overview of Static Multi-hop Linear WSN Introuction to Static Multi-hop Linear Topology Flat One-tier Multi-hop Topology Hierarchical Multi-hop Multi-tier Topology Wireless Stanars The IEEE Stanar The IEEE Stanar Routing Protocol Dynamic Routing Protocol Hybri Routing Protocol Static Routing Protocol Commercial Routing Protocol Relate Work in Multi-hop Linear Topology Introuction to Transmission Control Protocol IV

6 Table of contents Chapter 3. Types of TCP TCP Tahoe TCP Reno TCP New-Reno Selective Acknowlegements (SACK) TCP Vegas TCP Agent at a Receiver Relate Work in TCP Fairness Introuction to Energy Moel Energy versus Network Architecture Energy State in a Network Relate Work in Energy Moel Conclusion Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Introuction Routing Stability Factor in Linear Static Topology Queuing Factor in WSN Time to Live in WSN Broacast Packets Propose Linear Static Routing Dual Interleaving Linear Static Routing Routing Table Generating Algorithm in DI-LSR Source Noe Routing Flows in DI-LSR Destination Noe Routing Flows in DI-LSR Performance Evaluation on Flat One-tier Linear Topology Performance Metrics on Flat One-tier Linear Topology Packet Delivery Ratio En-to-en Delay Throughput Throughput Fairness Inex Receive Data Packet Variation Passive Source Noes V

7 Table of contents Normalize Routing Loa Energy Consumption Simulation Results an Discussion for Flat One-tier Linear Topology Packet Delivery Ratio vs. Number of Source noes En-to-en Delay vs. Number of Source noes Throughput vs. Number of Source noes Throughput Fairness Inex vs. Number of Source noes Receive Data Packet Variation vs. Number of Source noes Passive Source Noes vs. Number of Source noes Normalise Routing Loa vs. Number of Source noes Energy Consumption vs. Number of Source noes Dual Cluster Hea Interleaving Linear Static Routing Routing Table Generating Algorithm in DCHI-LSR Source Noe Routing Flows in DCHI-LSR Destination Noe Routing Flows in DCHI-LSR Performance Evaluation on Linear Cluster Two-tier Topology Simulation Results an Discussion for Cluster Two-tier Linear Topology 78 Chapter 4. Packet Delivery Ratio vs. Number of Source noes En-to-en Delay vs. Number of Source noes Throughput vs. Number of Source noes Throughput Fairness Inex vs. Number of Source noes Receive Data Packet Variation vs. Number of Source noes Passive Source Noes vs. Number of Source noes Normalise Routing Loa vs. Number of Source noes Energy Consumption vs. Number of Source noes Conclusion Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Introuction Network Fairness Factor in Multi-hop Linear Topology Source Noe Distance in a Multi-hop Linear Topology Long-term Network Fairness an Data Frequency VI

8 Table of contents Degraing of Wireless Performance The Delaye Acknowlegement Timeout Moel for Flat One-tier an Cluster-base Two-tier Multi-hop Linear Topology Moel One; TCP Delaye Acknowlegement Timeout Moel for Fairness Critical Application Moel Two; TCP Delaye Acknowlegement Timeout Moel for Throughput Critical Application Network Fairness an Performance Evaluation on Flat One-tier Multihop Linear Topology Simulation Results an Discussion for Flat One-tier Multi-hop Linear Topology Moel Three; TCP Delaye Acknowlegement Timeout for Fairness Critical Application Moel Four; TCP Delaye Acknowlegement Timeout for Throughput Critical Application Network Fairness an Performance Evaluation on Cluster-base Twotier Multi-hop Linear Topology Simulation Results an Discussion for Linear Cluster Two-tier Topology Conclusion Chapter 5. AHiT Energy Moel for Multi-hop Linear Topology Introuction to Energy Moel Factors in Energy Moel Consume Energy an Network Lifetime Communication Range Energy Optimization Active High Transmitter-receiver Energy Moel (AHiT) Energy Efficiency Evaluation on Energy Moel Energy Performance Metrics Energy Consumption Variation of Energy Consume Network Lifetime Simulation Results an Discussion: DI-LSR an DCHI-LSR VII

9 Table of contents Flat Topology: Energy Consumption vs. Number of Source Noes Flat Topology: Consume Energy Variation vs. Number of Source Noes 145 Flat Topology: Network Lifetime vs. Number of Source Noes Cluster Topology: Energy Consumption vs. Number of Source Noes 147 Cluster Topology: Consume Energy Variation vs. Number of Source Noes 148 Cluster Topology: Network Lifetime vs. Number of Source Noes Simulation Results an Discussion: TCP Delaye Acknowlegement Timeout Moel for Flat One-tier Topology an Cluster-Base Two-tier Topology Flat Topology: Energy Consumption vs. Number of Source Noes Flat Topology: Consume Energy Variation vs. Number of Source Noes 151 Flat Topology: Network Lifetime vs. Number of Source Noes Cluster Topology: Energy Consumption vs. Number of Source Noes 153 Cluster Topology: Consume Energy Variation vs. Number of Source Noes 154 Cluster Topology: Network Lifetime vs. Number of Source Noes Conclusion Chapter 6. Conclusions an Future Works Conclusion Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization AHiT Energy Moel for Multi-hop Linear Topology Research Impact Future Research Directions VIII

10 Table of contents Appenix A A.1 Aitional Simulation Results an Discussion for Flat One-tier Multi-hop Linear Topology: A.1.1 En-to-en Delay vs. Number of Source Noes A.1.2 Passive Source Noes vs. Number of Source Noes A.1.3 Normalise Routing Loa vs. Number of Source Noes A.1.4 Energy Consumption vs. Number of Source Noes A.2 Aitional Simulation Results an Discussion for Linear Cluster Two-tier Topology: A.2.1 En-to-en Delay vs. Number of Source Noes A.2.2 Passive Source Noes vs. Number of Source Noes A.2.3 Normalise Routing Loa vs. Number of Source Noes A.2.4 Energy Consumption vs. Number of Source Noes References IX

11 List of Figures List of Figures Figure 1-1: Overview of an oil an gas process... 2 Figure 2-1: Multi-hop linear WSN with Nn an ND at opposite en Figure 2-2: Multi-hop linear WSN (single hop technique) Figure 2-3: ROLS hierarchical WSN Figure 2-4: Three-tier linear cluster WSN Figure 2-5: LEACH WSN Figure 2-6: Multi-Hop LEACH WSN Figure 3-1: Multi-hop linear WSN Figure 3-2: Route breakage in a multi-hop linear WSN Figure 3-3: Biirectional packets queue scheme in a linear WSN Figure 3-4: DI-LSR on a flat multi-hop linear topology Figure 3-5: Routing table generating process in DI-LSR Figure 3-6: Forwar path routing table in DI-LSR Figure 3-7: Reverse path routing table in DI-LSR Figure 3-8: Flow chart for ata packet cycle in DI-LSR Figure 3-9: Packet elivery ratio vs. number of source noes Figure 3-10: En-to-en elay vs. number of source noes Figure 3-11: Throughput vs. number of source noes Figure 3-12: Throughput fairness inex vs. number of source noes Figure 3-13: Receive packet variation vs. number of source noes Figure 3-14: Passive source noes vs. number of source noes Figure 3-15: Network routing loa vs. number of source noes Figure 3-16: Energy per-packet vs. number of source noes Figure 3-17: DCHI-LSR on a cluster-base multi-hop linear topology Figure 3-18: Routing table generating process in DCHI-LSR Figure 3-19: Forwar path with even sequence of cluster heas in DCHI-LSR Figure 3-20: Forwar path with o sequence of cluster heas in DCHI-LSR Figure 3-21: Reverse path in DCHI-LSR Figure 3-22: Flow chart for a ata packet cycle in DCHI-LSR Figure 3-23: Delivery ratio vs. number of source noes X

12 List of Figures Figure 3-24: En-to-en elay vs. number of source noes Figure 3-25: Throughput vs. number of source noes Figure 3-26: Throughput fairness inex vs. number of source noes Figure 3-27: Receive ata packet variation vs. number of source noes Figure 3-28: Passive source noes vs. number of source noes Figure 3-29: Network routing loa vs. number of source noes Figure 3-30: Energy per-packet vs. number of source noes Figure 4-1: Data travel rate in multi-hop linear topology Figure 4-2: Effect on ata travel rate in multi-hop linear topology Figure 4-3: The DATM-F with source noes (On/En) an estination noe (ND) Figure 4-4: The DATM-C with source noes (Nn) an estination noe (ND) Figure 4-5: The timing iagram of DATM-FFCA Figure 4-6: Delaye acknowlegement process in DATM-FFCA an DATM-FTCA Figure 4-7: Generating elaye acknowlegement timeout in DATM-FFCA Figure 4-8: The timing iagram of DATM-FTCA Figure 4-9: Generating elaye acknowlegement timeout in DATM-FTCA Figure 4-10: Throughput fairness inex vs. number of source noes Figure 4-11: Receive ata packet variation vs. number of source noes Figure 4-12: Packet elivery ratio vs. number of source noes Figure 4-13: Throughput vs. number of source noes Figure 4-14: The timing iagram of DATM-CFCA Figure 4-15: Delaye acknowlegement process in DATM-CFCA an DATM-CTCA Figure 4-16: Generating elaye acknowlegement timeout in DATM-CFCA Figure 4-17: The timing iagram of DATM-CTCA Figure 4-18: Generating elaye acknowlegement timeout in DATM-CFCA Figure 4-19: Throughput fairness inex vs. number of source noes Figure 4-20: Receive ata packet variation vs. number of source noes Figure 4-21: Packet elivery ratio vs. number of source noes Figure 4-22: Throughput vs. number of source noes XI

13 List of Figures Figure 5-1: Energy consumption in a multi-hop linear WSN Figure 5-2: Timing iagram in a flat one-tier multi-hop linear topology Figure 5-3: Timing iagram in a cluster two-tier multi-hop linear topology Figure 5-4: Flow chart of the AHiT energy moel Figure 5-5: Energy per-packet vs. number of source noes Figure 5-6: Variation of consume energy vs. number of source noes Figure 5-7: Network lifetime vs. number of source noes Figure 5-8: Energy per-packet vs. number of source noes Figure 5-9: Consume energy variation vs. number of source noes Figure 5-10: Network lifetime vs. number of source noes Figure 5-11: Energy per-packet vs. number of source noes Figure 5-12: Consume energy variation vs. number of source noes Figure 5-13: Network lifetime vs. number of source noes Figure 5-14: Energy per-packet vs. number of source noes Figure 5-15: Consume energy variation vs. number of source noes Figure 5-16: Network lifetime vs. number of source noes Figure A-1: En-to-en elay vs. number of source noes Figure A-2: Passive source noes vs. number of source noes Figure A-3: Network routing loa vs. number of source noes Figure A-4: Energy per-packet vs. number of source noes Figure A-5: En-to-en elay vs. number of source noes Figure A-6: Passive source noes vs. number of source noes Figure A-7: Network routing loa vs. number of source noes Figure A-8: Energy per-packet vs. number of source noes XII

14 List of Tables List of Tables Table 2-1: Comparison between IEEE an IEEE stanar Table 3-1: NS2 simulation parameters for flat topology Table 3-2: NS2 simulation parameters for cluster-base topology Table 3-3: Brief comparison of DI-LSR an DCHI-LSR Table 3-4: Overall network performance of DI-LSR an DCHI-LSR Table 4-1: Example of ata travel time for a multi-hop linear topology Table 4-2: NS2 simulation parameters for flat topology Table 4-3: NS2 simulation parameters for cluster-base topology Table 4-4: Comparison between DI-LSR, DATM-FFCA an DATM-FTCA Table 4-5: Comparison between DCHI-LSR, DATM-CFCA an DATM-CTCA Table 5-1: Energy efficiency with an without AHiT energy moel Table A-1: Comparison between DI-LSR, DATM-FFCA an DATM-FTCA Table A-2: Comparison between DCHI-LSR, DATM-CFCA an DATM-CTCA XIII

15 List of Abbreviations List of Abbreviations ADAM ADAM-sn A-LEACH AFVQ AHiT AODV AODV+ AODVC AODVC+ ATIM CATRA CTS CW DATM-CFCA DATM-CFCA+ DATM-CTCA DATM-CTCA+ DATM-FFCA DATM-FFCA+ DATM-FTCA DATM-FTCA+ Aaptive Delaye ACK Mechanism Aaptive Delaye ACK Mechanism with shorter noe istance Avance Low-Energy Aaptive Clustering Hierarchy ACKs-First Variable-size Queuing Active High Transmitter-receiver energy moel A hoc On-Deman Distance Vector A hoc On-Deman Distance Vector with Aaptive Delaye ACK Mechanism A hoc On-Deman Distance Vector in a two-tier cluster topology A hoc On-Deman Distance Vector in a two-tier cluster topology with Aaptive Delaye ACK Mechanism Dynamic A-hoc Traffic Inication Map Cooperation between channel access control with TCP Rate Aaptation Clear to sen Contention Winow Delaye acknowlegement timeout moel for a Fairness critical application for cluster-base topology Delaye acknowlegement timeout moel for a Fairness critical application for cluster-base topology with Active High Transmitter-receiver energy moel Delaye acknowlegement timeout moel for a Throughput critical application for cluster-base topology Delaye acknowlegement timeout moel for a Throughput critical application for cluster-base topology with Active High Transmitter-receiver energy moel Delaye acknowlegement timeout moel for a Fairness critical application for flat topology Delaye acknowlegement timeout moel for a Fairness critical application for flat topology with Active High Transmitter-receiver energy moel Delaye acknowlegement timeout moel for a Throughput critical application for flat topology Delaye acknowlegement timeout moel for a Throughput critical application for flat topology with Active High Transmitter-receiver energy moel XIV

16 List of Abbreviations DCCP DCHI-LSR DCHI-LSR+ DDEEC DEEC DI-LSR DI-LSR+ DSDV DSDV DSDVC DSDVC+ DSDV-sn DSR Dst EDDEEC EDEEC EDURo+ FFD FRP HC ifqlen LEACH LEACH-C LEACH-F LSR MH-DISCPLN MH-LEACH MH-PSM Datagram Congestion Control Protocol Dual Cluster Hea Interleaving Linear Static Routing Dual Cluster Hea Interleaving Linear Static Routing with Active High Transmitter-receiver energy moel Develope Distribute Energy-Efficient Clustering Distribute Energy-Efficient Clustering Dual Interleaving Linear Static Routing Dual Interleaving Linear Static Routing with Active High Transmitter-receiver energy moel Destination-sequence Distance-vector routing Destination-sequence Distance-vector routing with Aaptive Delaye ACK Mechanism Destination-sequence Distance-vector routing in a two-tier cluster topology Destination-sequence Distance-vector routing in a two-tier cluster topology with Aaptive Delaye ACK Mechanism Destination-sequence Distance-vector routing with shorter noe istance Dynamic Source Routing Designate estination point Enhance Develope Distribute Energy-Efficient Clustering Enhance Distribute Energy-Efficient Clustering Energy-Delay Unifie Routing protocol Full-Function Devices Fixe routing path Hop count Queue length Low-Energy Aaptive Clustering Hierarchy Low-Energy Aaptive Clustering Hierarchy-centralize Fixe Number of Cluster Low Energy Aaptive Clustering Hierarchy Link State Routing Multi-hop istance-base an low-energy aaptive clustering Multi-Hop Communication Multi-hop power saving moe XV

17 List of Abbreviations Multi-Hop LEACH ND NHS NOAH nw-mac Nxt OLSR RFC RFD RIX-MAC ROLS RTO RTS RTT SACK SCTP ssthresh TCP TCP-ADW TCP-IA TDMA TTL UDP Wi-Fi WLAN WSN Multi-Hop Low-Energy Aaptive Clustering Hierarchy Destination noe Network-Harmonize Scheuling No A-hoc Routing Agent Multiple wake-ups provisione uty cycle MAC protocol Neighbouring cluster heas or Neighbouring noes Optimise link state routing Request for Comments Reuce-Function Devices Receiver-Initiate X-MAC Routing protocol for Linear Structure Re-transmission time-out Request to sen Roun-trip time Selective Acknowlegements Stream Control Transmission Protocol Slow-Start Threshol Transmission Control Protocol Aaptive Delay Winow Intelligent Acknowlegement technique Time Division Multiple Access Time to live User Datagram Protocol Wireless fielity Wireless local area networks Wireless sensor network XVI

18 List of Publications Submitte Conference Paper List of Publications Siva Kumar Subramaniam, Shariq Mahmoo Khan, Rajagopal Nilavalan an Wamaeva Balachanran, Network Performance Optimization Using O an Even Routing Algorithm for pipeline network, in Proceeings of the 8th Computer Science an Electronic Engineering Conference, 2016, pp Siva Kumar Subramaniam, Shariq Mahmoo Khan, Rajagopal Nilavalan an Wamaeva Balachanran, Enhancing Pipeline Network Performance Using Dual Interleaving Cluster Hea Routing Protocol, in Proceeings of the 1st International Conference on Innovations in Computer Science & Software Engineering (ICONICS), Siva Kumar Subramaniam, Shariq Mahmoo Khan, Rajagopal Nilavalan an Wamaeva Balachanran, TCP Timeout Mechanism for Optimization of Network Fairness an Performance in Multi-Hop Pipeline Network, in Proceeings of the 1st International Conference on Innovations in Computer Science & Software Engineering (ICONICS), Poster Presentation Siva Kumar Subramaniam, Rajagopal Nilavalan an Wamaeva Balachanran, Talking pipe: Wireless sensor network for conition monitoring in oil an gas pipelines, Brunel University Lonon Research Stuent Conference, March Submitte Journal Paper Siva Kumar Subramaniam, Shariq Mahmoo Khan, Rajagopal Nilavalan an Wamaeva Balachanran, Static Network Performance Optimization Using Interleave Routing Algorithm, IIUM Engineering Journal. Siva Kumar Subramaniam, Shariq Mahmoo Khan, Rajagopal Nilavalan an Wamaeva Balachanran, Interleave Delaye Acknowlegement TCP Mechanism for Throughput Fairness Optimization in Pipeline Network, Wireless Personal Communications. XVII

19 Statement of Originality Statement of Originality While registering as a caniate for the above egree, I have not been registere for any other research awar. I certify that the results an conclusions presente in this thesis are my own work. This thesis has not previously been submitte for a egree nor has it been submitte as part of the requirements for a egree. Due acknowlegement has been mae in the text to all other material use. XVIII

20 Introuction Chapter 1 1. Introuction Introuction an Application of Multi-hop Linear Topology Over the last ecae, the propagation of new network topologies an traffic patterns emerge for a wie range of new applications with the implementation of wireless sensor network (WSN)[1]. In many new areas of remote monitoring, the WSN eployment imposes a multi-hop linear topology where ata from sensing noes are collecte over a chain of noes linke to a receiver noe locate at one en of a network. The characteristic of a static multi-hop linear network has a fixe architecture in which the noes are permanently locate with constant raio propagation at all-time [2, 3]. Over the years, the static multi-hop linear topology is a well-accepte esigne for special applications such as pipelines or linear structure that transmits the collecte ata from sensing points through a multi-hop wireless meium. In the recent years, such WSN technology an its application have gaine consierable importance in remote monitoring particularly on oil an gas pipelines [4-7] with rapi avancements in unerlying technologies on communication protocols, sensing capability, energy efficient wireless evices an implementation cost on vast applications [8, 9]. Remote Monitoring of Oil an Gas Pipelines The oil an gas pipelines are known as a cost-effective an a safer transportation meium that are still vulnerable towars angerous accients or physical amages [5, 6, 10]. Numerous stuies have highlighte that pipeline transportation ha a long history of failures however when compare with the traitional railroa transportation, the reporte accients are still significantly low. Oil leaks from accients in a pipeline contribute to irregularities in temperature reaings beneath the pipeline, whereas a rupture gas pipeline 1

21 Introuction rapily ecreases the temperature above the pipeline [7, 10, 11]. Hence, constant monitoring of temperature an pressure changes on sensors can promptly etect leaks or rupture pipeline proactively for faster response to an impening accient. The oil an gas inustry faces numerous challenges especially in complying with frequently changing environmental regulations in a pipeline monitoring an management system. The exceptional nee to monitor the unknown states on the oil an gas pipelines is essential for the inustry to wisely optimise the running resources [4, 10]. The unknown states or threats of an oil an gas pipeline coul affect pipeline integrity, safety, reliability an security whilst simultaneously meeting the emans on elivery scheules as well as optimising the cost factor at the same time. To oversee the entire process as shown in Figure 1-1, a reliable an scalable WSN plays a crucial role in sensing the abnormalities on pipelines an to communicate them to a centralise monitoring station in a istance away from where the future ecision is mae [12-14]. 1: Extraction an prouction 2: Pipeline an operation 3: Refining an processing Figure 1-1: Overview of an oil an gas process Therefore, the rapi evelopments of WSN are graually aapte for remote monitoring of oil an gas pipelines which empowers the monitoring station to oversee the entire stretch of pipelines irrespective of the unique geographical terrain no matter how intense the environmental characteristics are. Challenges of WSN in Oil an Gas Pipelines The WSN is broaly categorise as infrastructure or infrastructure-less network. A WSN eploye on oil an gas pipelines are consiere as an infrastructure base, comprises of static sensing noes with a more resilient network backbone. 2

22 Introuction Such a WSN architecture are usually stretch over an extene istance from one noe to another that is vulnerable to high probabilities of communication failure, intolerable latency, high error rate an limite network lifetime [6, 12]. The overall performance of a static multi-hop linear network primarily epens on the routing scheme whereby the traitional routing protocols are eeme inefficient in a pipeline network context [2, 15]. Furthermore, a multi-hop linear network also has limite network resources, especially with a single path for both sensing an ata stream, to accommoate the overwhelming ata flow towars a single receiver at the en of a network [12]. These features are unique that varies to the nature of application especially in a time constraint an 24/7 base system such as the oil an gas pipeline. Some of the critical challenges of WSN eployment on an oil an gas pipeline are: Limite communication range - The communication or the raio range of wireless noes are critical in a linear architecture since every noe is a communication backbone in ensuring the communication link between neighbouring noes in the network. This affects the cost of installation (epening on the network size) as well as the robustness of the network. Large-scale network Noes istribute in a linear architecture is critical ue to limite an share network resources such as banwith, transmission slot, time in-between transmission etc. Therefore, accomplishing real-time ata transmission in a static multi-hop linear topology is always challenging especially in a long range wireless network such as in a pipeline network. Huge ata stream The share network resources also restricts the ata stream that epens on the ata rate an ata size in a linear architecture such as pipeline network. Noes in the network often generate real-time ata packets an supporting packets (broacast an control packets) that affect the banwith utilization rate in a pipeline network. Power source Limite power supply an extene communication range is often a threat in a remote monitoring of oil an gas pipeline as this often results in energy starvation of wireless noes in the network. Thus, such a network consumes high energy as compare to other wireless network 3

23 Introuction topologies, since each noe is only able to establish a communication (long istances) with the closest neighbouring noe. In recent years, static routing protocols for a multi-hop linear network has been an extensive an emaning research area, especially the newly emerging areas with iniviual perspectives in terms of performance an many other relate wireless metrics [16-18]. However, the ever-changing eman an stringing requirement confronting a static multi-hop linear network have a significant implication towars the fully functional eployment on an oil an gas pipeline. Motivation The research aims to analyse the issues aresse to a traitional multi-hop linear topology in a pipeline network. The outcome of this thesis is to highlight the propose techniques by esigning a viable solution with etaile analysis to overcome the sai issues. The multi-hop linear topology is an important communication architecture on an extene range pipeline network. Due to the unique geographical terrain an ata accumulation factor in a single path communication contributes to the occurrence of noes without ata transmission opportunity (passive noes) between the sener an a receiver noe [19, 20]. Most of the traitional ynamic an hybri routing protocols have ifferent characteristics from network initialization to a route iscovery process [12, 21, 22]. One of the crucial factors in a multi-hop linear communication is queue overwhelming from both ata an control packets which lea to a bottleneck points in the network. Such factors in routing protocols are often relate to frequent communication link instability. Link instability will trigger route maintenance proceure at a certain point in the network, increases the consumption of network resources which will eventually contribute to unerperforming of network characteristics. This woul result in an increasing number of passive noes or noe failures (ue to limite power) which is a waste of network resources an allocation that subsequently egraes the overall network performance [17]. The issues with higher routing overhea, queue overwhelming, frequent upates on route an 4

24 Introuction passive noes, rives the research motivation to the evelopment of a static routing algorithm. This can eliminate broacast relate routing overhea, preefine routing path, reuce the effects of queue overwhelming an eventually enhance the overall network performance. Fairness inex is a key inicator of the equality rate of a share allocation particularly in a single path communication as in a multi-hop linear topology. Equal sharing of network resources in a single path communication is reflecte in a constraine performance riven for the entire network rather than iniviual noes [23, 24]. Generally, imbalance network allocation is a result of some greey noes usually locate within a shorter communication range from a receiver noe. A greey noe in a multi-hop linear network contributes towars noe starvation an passive noes that rastically egrae the overall network performance in the overtime [25, 26]. The network allocation is often relate to the ata rate an travel time known as roun-trip time (RTT) base on the location of noes in a linear arrangement [27, 28]. One of the most common reliable ata transport protocol is the Transmission Control Protocol (TCP) that employ a ata acknowlegement mechanism which plays a very important role in long range linear communication to ensure reliable ata transmission. With the TCP agent, the ata packet elivery ratio an fairness inex [24, 29] is not at a guarantee rate but can be manipulate for optimum results. Therefore, creating a backhan (at the receiver noe) scheuling mechanism woul ensure optimum network fairness an resources allocation among all source noes to achieve an optimum network performance. The network energy consumption in a static multi-hop linear topology is one of the utmost intricate an challenging issues [20, 30, 31]. This is ue to the snowballing ata accumulation factor towars the estination noe at one en of the network can cause loss of ata uring transmission. Even though a WSN in a structure network has a constant power source, yet the energy optimisation is always a esire goal in any wireless evices. The imbalance energy consumption of noes oes not only affect the respective noe but has a consequent effect towars the network lifespan that egraes the overall network performance [31-33]. The motivation to esign an energy efficient 5

25 Introuction moel with the ability to aapt to the ata transmission nature can reuce consierable energy usage hence exten the network lifetime without compromising the overall network performance. Objectives of the Thesis The presente research work in this thesis has consiere working on multiple challenges in a static multi-hop linear topology, hence the aim has chain like factors when to compare: Design an evelopment of a routing algorithm for the implementation on a static (flat one-tier an cluster-base two-tier) multi-hop linear topology. The aim of the propose routing algorithm is to enhance the packet elivery ratio an reucing the routing overhea by incorporating the ual interleaving techniques on a known routing path. The propose routing algorithm is also expecte to reuce network energy consumptions an achieve a reasonable throughput fairness among source noes in a multi-hop linear WSN. The research aims to esign an incorporate a fairness moel with the propose routing algorithm to achieve optimum throughput fairness inex among all source noes in the network with a TCP agent. The fairness moel shoul enhance the equality of network allocation among all noes in the network with a reasonable performance. The research aims to aress the energy constraint an imbalance energy usage throughout noes in a linear topology. The research aims to esign an energy efficient moel that reuces the energy consumption an prolongs the network lifetime. The research work presente in this thesis were implemente an conucte on Network Simulator 2 (NS2) a simulation tool to analyse the overall performance as well as the other wireless metrics on a static multi-hop linear topology [34]. The performance of the propose techniques has been evaluate with numerous wireless metrics such as packet elivery ratio, en-to-en elay, throughput, throughput fairness inex, receive packet variation, the ratio of passive noes, normalise routing overhea, energy consume per-packet, consume energy variation an network lifetime. The analysis base on the numerous wireless 6

26 Introuction metrics is an inicator to unerstan the characteristics an behaviour of the propose techniques in a network. The research work also explores factors base on the network ensity an transmission rate against the performance of the propose techniques. Research Methoology To achieve the esire goals an set objectives, a etail literature reviews on static multi-hop linear topology was conucte with a comprehensive analysis of numerous publishe works. The relate research articles from publishe journal an conference papers were reviewe in the initial phase of the research to unerstan the research irections as well as the research focus area. On the next stage, the routing process for static multi-hop routing protocols was stuie thoroughly particularly in the key area of application involve in various publishe articles relate to oil an gas pipelines. In the literature review stage, etail analysis from various publishe work relate to static multi-hop linear topology an challenges relate to the performance egraing factors in a multihop linear topology were ientifie. The propose algorithms an techniques were implemente in NS2 [34]. The propose algorithms an techniques were teste in numerous simulation environment using the simulator to replicate a pipeline network. Furthermore, the propose algorithms an techniques were compare with some existing methos relevant to the application. Finally, the essential wireless metrics (overall performance an fairness) were compare an analyse against some existing methos to valiate the quality of work prouce. Contribution of Knowlege The knowlege contribution of this thesis is the esign an evelopment of a reliable an performance riven static routing scheme for a flat one-tier multi-hop linear topology an a hierarchical two-tier linear topology. The research outcome has a number of contribution to knowlege in the respective area of application. The contribution of knowlege can be categorise into five components as such: 1. A new performance-riven routing algorithm name the Dual Interleaving Linear Static Routing (DI-LSR) is a routing algorithm establishe for a flat one- 7

27 Introuction tier multi-hop linear topology [12]. The esign concept of DI-LSR has eliminate broacast packets for route iscovery process which was eventually replace with a preefine ual interleaving route between source to a estination noe to improvise the ata rate an banwith allocation at a receiver noe. The new concept introuces a ual interleaving route among source noes for share ata sensing an the ata stream from a source to a estination noe splits the network traffic to elevate the problems with queue overwhelming as well as passive noes. The preefine route for DI-LSR is base upon the two routing tables (forwar: source to estination/reverse: estination to source) contains information of route navigation for respective noes e.g. ata travel irection (forwar an reverse), esignate estination noe an minimum travel cost which is relate to time to live. 2. A new performance-riven routing algorithm name the Dual Cluster Hea Interleaving Linear Static Routing (DCHI-LSR) is a routing algorithm establishe for a cluster-base multi-hop linear topology. The esigne DCHI- LSR has eliminate the traitional broacast packets for route iscovery an was replace with a preefine ual interleaving route between cluster hea to a estination noe improvises the ata rate an banwith allocation at a receiver noe. The new concept of ual interleaving route among cluster heas in a linear architecture employs a eicate ata sensing at source noes (first tier) an transfer the ata through a eicate ata stream between cluster heas (secon tier) to a estination noe. With the implementation of the ual interleaving cluster heas splits the network traffic to elevate the problems with queue overwhelming as well as passive noes. The preefine route for DCHI-LSR is base upon the two routing tables (forwar: source to estination/reverse: estination to source) contains information of route navigation for respective noes e.g. ata travel irection (forwar an reverse), estination noe an minimum travel cost. 3. The TCP elaye acknowlegement timeout moel for a Fairness critical application (DATM-FFCA) an the TCP elaye acknowlegement timeout moel for a Throughput critical application (DATM-FTCA) for a flat one-tier multi-hop linear architecture has been propose for improving the fairness in network resources especially on throughput. The propose formulation is 8

28 Introuction able to erive an ieal elaye acknowlegement timeout value base on the noe sequence in the network. The establishe acknowlegement packet scheuling mechanism in the propose DATM-FFCA or DATM-FTCA with DI- LSR woul respectively enhance the throughput fairness inex to an optimum level as well as the other wireless metrics with reuce packet collisions in a flat multi-hop linear topology. 4. The TCP elaye acknowlegement timeout moel for a Fairness critical application (DATM-CFCA) an the TCP elaye acknowlegement timeout moel for a Throughput critical application (DATM-CTCA) for a cluster-base multi-hop linear architecture has been propose for improving the fairness in network resources, especially on throughput. The propose formulation is able to erive an ieal elaye acknowlegement timeout value base on the noe sequence for a eicate cluster hea in the network. The establishe acknowlegement packet scheuling mechanism in the propose DATM- CFCA or DATM-CTCA with DCHI-LSR woul respectively enhance the throughput fairness to an optimum level as well as the other wireless metrics with reuce packet collisions in a cluster-base multi-hop linear topology. 5. To aress issues an challenges of energy consumption in a static multi-hop linear topology, an energy efficient moel with the ability to aapt to the varying nature of ata transmission has been propose. The propose Active High Transmitter-receiver energy moel (AHiT) emphasise on sleep an active switching scheme for respective noes in the network to optimise energy usage only in an active perio of time. The AHiT energy moel empowers the noes from sleep state to active state accoring to the ata transmission or receiving process in an active energy state only. Once the ata transmission or receiving process is accomplishe, the AHiT energy moel changes the energy moe of a respective noe from on state to sleep state to conserve energy from a traitional ile energy moe. The AHiT energy moel can be eploye on both a flat one-tier architecture an a cluster-base architecture isregars to the functionality of a noe in the network. 9

29 Introuction Thesis Organisation This thesis is organise into six chapters, where Chapter 1 is the introuctory chapter for the research work, Chapter 2 is the literature review, Chapter 3 to Chapter 5 is the contribution of the research work an Chapter 6 is the conclusion an future works. An overview of all the chapters are briefly escribe as below: Chapter 1 briefly introuces to the static multi-hop linear topology in an oil an gas pipelines. The subsequent subchapters aress the research challenges in the relevant application an propose some viable techniques as an alternative solution for the en users. The scope of research work is clearly escribe base on the propose techniques. Chapter 2 introuces to some backgroun information an elaborates the taxonomy of a static multi-hop routing protocol. This chapter is sectione into three parts, which outline the propose techniques between chapter three to chapter five base on the contextual factors an various existing eploye techniques in a wireless network. Chapter 3 briefly elaborates the factors in a multi-hop linear topology an provies a etail escription of the propose DI-LSR an the DCHI-LSR. This chapter is sectione into two parts which outline the comprehensive analysis of both DI-LSR an DCHI-LSR respectively with some existing routing protocols in numerous scenarios were presente. Chapter 4 briefly escribes the fairness issues in a multi-hop linear topology an provies a etail escription of the propose technique to improvise the issues aresse in a multi-hop linear topology. This chapter has two section with the first section outlines the DATM-FFCA an the DATM-FTCA for a flat one-tier multi-hop linear architecture. The secon section outlines the DATM- CFCA an the DATM-CTCA for a cluster-base multi-hop linear topology. A comprehensive analysis of all four techniques with some existing methos was presente. Chapter 5 briefly escribes the energy issues in a multi-hop linear topology an focuses on the propose AHiT energy moel. The AHiT energy moel concentrate on energy reuction for noes in a multi-hop linear topology. A 10

30 Introuction comprehensive analysis of the propose an the stanar energy moel was presente. Chapter 6 summarises the research work an followe by some suggestion for the future irection of the research work. Conclusion In a large scale WSN comprises of hunres to thousans of iniviual evices working to support each other in a unique esigne formation as require by the en user or a certain application. The characteristic of a WSN esign formation solely epens on the type of application that is tailore to one of the many types of network topologies. Generally, network topologies that are commonly use in WSN are tree, star, linear, mesh an hybri which presents a specific set of challenges (avantages an isavantages). Thus, the static multi-hop linear topology is a feasible architecture for remote monitoring of an oil an gas pipelines (fixe infrastructure). In the next chapter, further escribes the focal concepts, protocols an technologies relate to WSN eployment on an oil an gas pipelines. Basic unerstaning an concepts behin WSN in the area of remote monitoring as a prerequisite of oil an gas pipeline monitoring will be presente. 11

31 An Overview of Static Multi-hop Linear WSN Chapter 2 2. An Overview of Static Multi-hop Linear WSN Introuction to Static Multi-hop Linear Topology Oil an gas pipelines are fixe infrastructures which are stretche over a long istance from one point to the other which coul be hunres of miles away [14]. In general, the communication between two estination noes takes place through intermeiate noes that are arrange in series. In pipeline networks, all noes are statically locate to establish a chain of communicating through all available noes in the network. Pipeline network consists of a series of static noes where each noe is esigne to perform as a host or/an as a router [4, 6, 7, 35]. The arrangements of noes in a pipeline network to form a unique architecture that is known as the multi-hop linear topology epens on the network size an the complexity of the application [1, 17]. A network topology is often referre to the way or manner in which noes are linke to one another in a network. The physical an logical arrangements of noes in a network can be escribe as a schematic escription of a network topology. Noes are arrange an interconnecte to one another to establish ata transfer between a sener to a receiver in a unique form base on the network topology an the type of implemente routing protocol. The various network topologies can easily be implemente base on the nature an complexity of an application in a network. The consieration of a network topology can be outline base on the implementation scale, expecte traffic on the network an fault-tolerant communication links. Thus, the architecture of a multi-hop linear topology can be classifie in two major structure which is the flat one-tier topology an the hierarchical multi-tier topology [3, 16, 19]. 12

32 An Overview of Static Multi-hop Linear WSN Flat One-tier Multi-hop Topology Basically, a flat one-tier multi-hop topology is a simple architecture with a series of noes containing both sensing an ata transferring noes (in an intermeiate noe) to a esignate estination noe [12, 17]. The move towars a flat one-tier topology aims to reuce the number of wireless evices base on the wireless stanars in a network especially in managing higher routing overhea an to iminish latency in a multi-hop linear network [2, 16, 19]. A flat one-tier topology simplifies the traffic flow in a network making it feasible for reeployment in a long range network particularly with a way simpler in managing the general operation in a network. There are two major rawbacks of a flat one-tier topology; (1) is a simple noe failure an (2) a network with higher ata rate can interrupt ata transfer ue to snowballing effect from a certain point in a multi-hop linear topology [1, 12, 36]. Hence, energy utilisation an network resources shoul be at an optimum level uring the network active perio. Hierarchical Multi-hop Multi-tier Topology Clustering in a wireless network is often relate to a network with higher efficiency, scalability an fault tolerant especially in a multi-hop linear topology [37, 38]. The multi-tier cluster can be eploye from a two-tier hierarchical clustering architecture to a certain number of tier or level in a cluster-base network [39, 40]. The cluster-base two-tier topology is one of the most popular architecture in a WSN. Generally, this topology is ivie into two where on the lower level (first tier) is eploye with sensing noes (sener) at the level two will be a series of leaer noes or is also known as cluster heas [3, 16]. A noe in a specific region is only permitte to sen an receive packets to a eicate leaer noe or cluster hea whereas the cluster heas sen the collecte packets to a estination noe at the en of a network. The clustering technique splits the ata generate level an ata stream where the ata packet is sent to a estination noe [41]. The avantage of such architectural esign contributes to lower routing overhea, reuces the network energy consumption [40] as well as fault-tolerant (a network without noe failures). The types an selective of network topology epen on the wireless stanars implemente by the user. 13

33 An Overview of Static Multi-hop Linear WSN Wireless Stanars The wireless stanars have a great impact in terms of performance or outcome to the type of topology implemente for a certain application. This thesis has consiere only two major stanars that ensembles the application on a pipeline network from all the other available stanars. The two stanar evices that are potential for a real scale implementation on a multi-hop linear topology is IEEE an IEEE Thus, the given stanars an explanations are escribe in an overview instea of a etaile technical aspect. The IEEE Stanar The wireless local area networks (WLAN) also known as wireless fielity (Wi-Fi) falls uner IEEE stanars with many ifferent variations or new substanars in IEEE that has variation in technical characteristic were introuce over these years. The IEEE architecture comprises of several iniviual components that are use to interact for wireless ata communication that supports a set of static or mobile stations [42-44]. The IEEE stanar evices support higher ata rate with scalability for large scale application as in a pipeline network. In general, the popularity of WLAN has more visibility in the en user market where a wie range of application is supporte by one of the IEEE variations [45, 46]. The brief technical operation etail of IEEE stanars is as shown in Table 2-1 which is wiely available on various off the shelf evices. The IEEE Stanar The technical stanar of IEEE is efine as an operation of a low-rate wireless personal area network. The stanar has many ifferent substanars that have variation in the technical characteristic of the evice. The IEEE stanar is known as the Bluetooth wireless communication technology. Bluetooth is esigne for short-range communication evices that are small, low cost an has lower power consumption [44, 47]. The other popular wireless stanar that is comparable to Bluetooth is the IEEE stanar known as ZigBee. The ZigBee technology provies reliable communication on both multi-hop an mesh network with low power consumption [48, 49]. The basic 14

34 An Overview of Static Multi-hop Linear WSN operation feature of a ZigBee oes not support frequency hopping making, thus scan for a channel that contains the least amount of interference uring startup [44, 50]. Network evices in ZigBee can be categorise into two classes; Full- Function Devices (FFD) with route messages function in mesh networks an able to act as a network coorinator, whereas the Reuce-Function Devices (RFD) can only establish communication with one FFD [44, 50, 51]. ZigBee also features to operate in a beacone or a non-beacone moe as preferre by the user. The brief technical operation etail of IEEE an IEEE stanars is as shown in Table 2-1 [44, 47, 50-55]. Table 2-1: Comparison between IEEE an IEEE stanar Parameters WiFi ZigBee WirelessHART Z-Wave IEEE stanar Operational frequency a/b/g /n/ac 5GHz (a,ac) / 2.4GHz (b,g) / 2.4GHz-5GHz (n) GHz 2.4GHz MHz Noes per master 2007 More than Range 100 meter (a,b,g) / 250 meter (n) 1600 meter 250 meter 100 meter Data rate 11Mbps (b)/54mbps (a,g) / 450Mbps (ac) / 600Mbps (n) 250 kbps 250 kbps 100 kbps Battery Life/Cost Days-weeks / High Monthsyears / Low Months-years / Moerate Monthsyears / Low WirelessHART is another technology similar to ZigBee that share the same IEEE stanar an the basic working principles. WirelessHART operates 15

35 An Overview of Static Multi-hop Linear WSN base on Time Division Multiple Access (TDMA) where all evices in a network are time synchronise that communicates in pre-scheule time-slots [51]. TDMA can minimise packet collisions an also ecreases the power consumption. Unlike ZigBee, WirelessHART noe features hopping for every message that changes channels for every sent packet making it more reliable in a large network of noes [51, 52]. Another comparable an popular wireless protocol in the market is the Z-Wave that is wiely being use for home automation. The Z-Wave is esigne base on the IEEE physical raio stanar that employs mesh networking in orer to exten communication range between the communication points [55, 56]. The stanar operating frequency at sub-1ghz supports greater communication range as compare to similar proucts in the current market. Routing Protocol The network topology selection remains an important element in a network esign base on the nature of an application, whereas a routing protocol has the same counter effect when eploye on the network. Generally, the architecture of a multi-hop linear topology focuses primarily on optimum noe placements an the selection of a routing protocol in a network. A noe in a specific location which wants to transmit a ata packet must first iscover or set the route to the estination noe using route iscovering protocols [21, 22]. The traitional multihop linear topology has varying metrics an impening essential networking factors when implemente with a respective routing protocol. Generally, the routing protocol performance is measure in terms of links stability between noes, breaking an reconstruction of links, in which it is a crucial activity in a network where most ata packets might be lost [12, 21, 57]. In a wireless network, all noes will generate broacast messages in a timely interval to their neighbouring noes within the sensing range to ensure their presence is acknowlege an to retain the pre-establishe or new routes. In general, the routing protocols have ifferent characteristic on the route learning or ientifying process an how this route is maintaine uring the network-active perio [58, 59]. The three most common routing protocols are ynamic routing protocol, hybri routing protocol an static routing protocol. 16

36 An Overview of Static Multi-hop Linear WSN Dynamic Routing Protocol A ynamic routing protocol is wiely use in many applications especially in a wireless network environment. The ynamic routing protocol enables the noes to learn the neee routes from irect neighbouring noes in a network. Generally, the ynamic routing protocol shares the known number of noes in a network in which the respective noe is able to learn all the possible routes require for ata transmission process [21]. In a ynamic routing protocol, the route iscovery methos may iffer from one another, ue to the nature of routing protocols use in a network. The two most common methos in route iscovery are reactive routing protocol (on eman) an proactive routing protocol (tableriven) [57, 60]. The major avantage of ynamic routing protocol is the scalability an aaptability with the real-time changes in a network. The major rawback of a ynamic routing protocol is the routing overhea for constantly upate of routing table entries which utilise a higher proportion of the banwith an energy in a network [12]. These processes further a complexity to the routing protocol with a egrae network performance on a multi-hop linear network. Hybri Routing Protocol The hybri routing protocol is commonly referre as a routing combination of both Destination-sequence Distance-vector routing (DSDV) [12, 22] an Link State Routing (LSR) that is built for fast convergence with less energy an memory [12]. The hybri routing protocol generates accurate route information to ientify the best route to the estination noe with routing information upates as an when neee. Generally, DSDV shares the knowlege of the whole network with neighbouring noes an LSR establishes router to router-sharing of neighbouring or closest noes. The major avantage of hybri routing protocols is the best of both ynamic an static routing features combine into one protocol. In general, the application with a hybri routing protocol woul benefit from a lower routing overhea compare to a traitional ynamic routing protocol. The rawback of a hybri routing protocol is the reaction towars the traffic eman on a network that epens on certain graient of incoming traffic. 17

37 An Overview of Static Multi-hop Linear WSN Static Routing Protocol A static routing protocol is commonly generate by a network aministrator using known network facts or architecture for a specific application. A static routing protocol is typically implemente on a complex network where the routing table entries are generate manually base the user's eman [12, 37, 61]. Another metho in static routing is known as a efault routing where a efault route is preefine between noes in the network with a route to the estination noe. The avantage of using a static routing protocol is the reuce routing overhea or perioic signalling packets for routing table entries [12]. The major rawback of a static routing are the unexpecte changes in a network that are not inicate in a routing table cannot be etecte for rerouting thus, making it ifficult to be implemente in a complex network. Commercial Routing Protocol A hoc On-Deman Distance Vector (AODV) [21, 22, 62] is a well-known commercial reactive routing protocol which operates on a eman-basis for route searching. The sequence number of the estination is use to ientify the newest route to the estination. In a multi-hop linear topology, if two noes are within the communication range, the source noe will route to its estination noe with an option to bypass its intermeiate noe base on real-time changes. AODV is use in many topologies such as unstructure or flat, cluster-base, chain an hybri. Figure 2-1 illustrates many possible routes using AODV which is generate between the source (N1 - Nn) an the estination noe (ND) in a linear multi-hop WSN with two noes locate within the communication range. 18

38 An Overview of Static Multi-hop Linear WSN Nn n hops N5 5 hops N4 4 hops N3 3 hops N2 2 hops N1 ND 1 hop Figure 2-1: Multi-hop linear WSN with Nn an ND at opposite en Dynamic Source Routing (DSR) [21, 62] is another well-known on eman routing protocol similar to AODV which navigate route between a source noe to the estination noe using its ata packet. All noes hol the accumulate route information s which is then use to route all the ata packets to the esignate estination as require. DSR is not viable for long range linear communication ue to high routing overhea. The most known commercial, the proactive routing protocol is Destination-sequence Distance-vector routing (DSDV) [21, 22]. DSDV ientify the available route to the estination noe in the network which shows less elay for route set up the process. The obvious limitations of DSDV is the frequent upates on the routing table s entries are require base on real-time changes on the network. The HELLO packets generation at every secon among neighbouring noes in the communication range is require to help the noes to ientify the possible route towars the esignate estination noe. Such process consumes high energy an a portion of the banwith even at an ile state. DSDV is use in many topologies such as unstructure or flat, cluster-base, chain an hybri. Figure 2-1 illustrates many possible routes using DSDV which is generate between the source (N1 - Nn) an the estination noe (ND) in a linear multi-hop WSN with two noes locate within the communication range. 19

39 An Overview of Static Multi-hop Linear WSN The Optimise link state routing (OLSR) [21] is another commercial table-riven routing protocol similar to DSDV but ientify routes to the estination noe which is known an retaine before ata packets are sent. Having the routes available to the estination noe, the route iscovery elay for fining a new route is almost at zero. The biggest concerns on using OLSR is with the high value of routing overhea generate which coul be greater than a stanar reactive protocol. Relate Work in Multi-hop Linear Topology In pipeline networks, all noes are statically locate to establish a chain of communication through all available noes in the network. Pipeline network consists of a series of static noes where each noe is esigne to perform as a host or/an as a router to establish a communication between two estination noes. A noe in a specific location that intens to transmit a ata packet first nees to iscover or set the route to the estination noe using route iscovering protocols. The two most common methos in route iscovery are reactive routing protocol (on eman) an proactive routing protocol (table-riven) [10, 11]. The selection of a routing protocol takes effect on to the type of topology being implemente. The authors in [2, 63] propose Chain-Base 1, a routing protocol for constructing multiple linear chains with a single sink noe (estination noe). The sener (source noe) generates the ata packet an forwars it to it neighbouring noe where this process will continue on all intermeiate noes until all ata packet arrives at the last noe on the chain. The last noe on each chain will aggregate the ata packet before transmitting it to the sink noe. Later, the authors in [2] propose Chain-base 2 where a routing protocol which consumes less energy an prolongs the network lifetime by incorporating another chain network among the last noe of each chain. The last noe of each chain is also known as a leaer noe [2, 63]. All the leaer noe from multiple chains from another main chain have the higher initial energy to support the aggregation an ata transmission to the sink noe from each respective noe in its own chain. The research work in [64, 65] has implemente a manually configure routing table by ignoring special routing perioic signalling packets with a preselecte 20

40 An Overview of Static Multi-hop Linear WSN path between the static source noe/noes to the static estination noe/noes in a gri topology. Fixe routing path (FRP) is an efficient static routing protocol with suppresse routing messages an broacast messages occurs only locally where a route is manually pre-calculate to an optimal path or the shortest path [64, 65] as preefine by the user similar to Figure 2-2. Due to these changes, the routing table entry of FRP requires no upates or refreshing uring its operation. Since FRP esign has vali estination route between the source an the estination noe, there are no outgoing packet queues for ata packet generate from all the source noes in the network since the IP layer queues were remove. Nn N5 N4 N3 N2 N1 ND n hops 5 hops 4 hops 3 hops 2 hops 1 hop Figure 2-2: Multi-hop linear WSN (single hop technique) The research work in [66] has implemente a manually efine communication routing path between the source an the estination noe for ata packet transfer. A static routing protocol is known as No A-hoc Routing Agent (NOAH) supports single-hop irect communication only between preefine wireless noes without generating any routing messages. The routing table is manually create base on the esire source to a single-hop estination noe which coul be a mobile noe or a base station. NOAH is known as a routing protocol with low routing overhea (biirectional packets) where the routes remain unchange once it has been configure. The use of NOAH routing protocol is not feasible in multi-hop wireless scenarios, especially in a very large WSN. 21

41 An Overview of Static Multi-hop Linear WSN 1 hops 4 th tier ND 1 hop 2 n tier 3 r tier M2 3 CH4 CH3 CH2 CH M1 Communication range of M1 Communication range of CH1 1 st tier Nn N11 N10 N9 N8 N7 N6 N5 N4 N3 N2 N1 Communication range of N1 Figure 2-3: ROLS hierarchical WSN The research work in [15, 17, 67] has implemente a hierarchical an concept of sectioning noes into three types of noes similar to Figure 2-3 which is efine as basic sensor noes, ata relay noes an ata ischarge noes. In Routing protocol for Linear Structure (ROLS), all noes are place in a linear arrangement with the respective purpose for each noe [15, 17, 67] base on multi-layer noe aressing scheme. The step by step ata transfer from etection point to the esignate noe is accomplishe in three levels of ata transfer using two routing algorithm. The lack of connectivity in an immeiate neighbour noe enables the ata relay noes to ouble the communication range by increasing the transmission power using the jump always algorithm to successfully sen a ata packet to the estination noe. The reirect always algorithm ientifies the reirecte flag an sen a negative acknowlegement to back into the source noe for the regeneration of the ata packet in the opposite irection. The reirect always algorithm will also upate its atabase for non-functional noes to reach the ata ischarge noes in the network. The authors in [68] propose an efficient oil an gas pipeline information collection algorithm using three-tiere network architecture similar to the illustration in Figure 2-4 which is efine as pipeline sensor noes, intermeiate stations sink noes an pipeline monitoring control centre. Wireless noes with flat ata collection algorithm response to impromptu ata an forwars it to the 22

42 An Overview of Static Multi-hop Linear WSN neighbouring noes with minimum buffere waiting time [68]. With the flat ata collection algorithm, multiple transmission routes are generate between all source noes to the esignate estination noe. The flat ata collection algorithm is incorporate with hierarchical or cluster for multi-hop scalability in wireless sensor network. With the ata collection algorithm, a communication is establishe with a specific cluster that enables ata fusion in the effort to reuce the transmitte ata rate to the pipeline monitoring control centre. 2 hops 1 hop 2 n tier 3 r tier M2 3 CH4 CH3 CH2 CH M1 4 ND Communication range of CH1 1 st tier Nn N11 N10 N9 N8 N7 N6 N5 N4 N3 N2 N1 Communication range of N1 Figure 2-4: Three-tier linear cluster WSN One of the first an well-known hierarchical routing protocols use in wireless sensor network is the Low-Energy Aaptive Clustering Hierarchy (LEACH) [3, 69]. The LEACH algorithm enables a group of sensors to self-organization an reclustering its cluster hea base on a rotation cycle. The cluster hea rotation is create to retain the noe with high energy to coorinate the member noes since the energy consumption of a cluster hea is very high. The role of a cluster hea is to be in irect contact with all its member noes as well as the estination noe which is at a single-hop from the cluster hea in both irections as shown in Figure 2-5. The cluster hea collects an aggregate ata packets from the source noes an relays it to the estination noe as require. 23

43 An Overview of Static Multi-hop Linear WSN 1 hop 1 hop 3 r tier ND 2 n tier CH2 CH1 Communication range of CH1 1 st tier Nn N5 N4 N3 N2 N1 Communication range of N1 Figure 2-5: LEACH WSN The research work in [70] Distribute Energy-Efficient Clustering (DEEC) is an algorithm esigne for cluster hea selection base on energy parameter in a heterogeneous WSN. DEEC algorithm selects the cluster hea base on the ratio of resiual energy on each noe as well as the network average energy. The noe in a cluster with highest initial an resiual energy will have more probability to be the cluster hea which can prolong the network lifetime. The continue rotation of cluster hea is require for fairness usage of energy an at the same time to retain as many noes as possible in the clustering group. The Avance Low-Energy Aaptive Clustering Hierarchy (A-LEACH) [61, 71] is a static cluster wireless sensor network for heterogeneous routing protocol which incorporate from LEACH [3] an DEEC protocol [70]. The network characteristic of A-LEACH is the ability to reuce its broacast packets by limiting the sensor noes in many small clusters groups with limite communication range only for each clustering groups. In each group, A-LEACH assigns a cluster hea which will be the point of communication to the estination noe. The concept of hierarchical or cluster encourages wireless multi-hop communication between a cluster to another in a very large wireless network. The Multi-Hop Low-Energy Aaptive Clustering Hierarchy (Multi-Hop LEACH) [41, 71]was introuce to enable multi-hop among cluster heas when the esignate 24

44 An Overview of Static Multi-hop Linear WSN estination noe is a istance away. All the sensor noes which are at a single-hop to the cluster hea sens a ata packet to a specific cluster hea where the cluster hea will aggregate an relay the ata packets through its neighbouring cluster hea towars the estination noe as shown in Figure 2-6. Communication range of CH1 n hops n 2 tier st 1 tier Nn 3 hops CHn N11 CH3 1 N10 2 hops N9 N8 CH2 1 N7 1 hop N6 N5 CH1 1 N4 N3 N2 1 ND N1 Communication range of N1 Figure 2-6: Multi-Hop LEACH WSN The authors in [72] propose a Distribute Algorithm for Multi-Hop Communication (MH-LEACH), an algorithm to establish multi-hop communication links between sensor noes with optimum energy saving. The MH-LEACH enables the cluster heas to establish a ata transfer to the nearest cluster hea in a multihop path to the base station [69]. The energy characteristic of MH-LEACH is to exten the network lifetime with short istance ata packet transmission which consumes lower energy. The authors in [41, 73] propose the Fixe Number of Cluster Low Energy Aaptive Clustering Hierarchy (LEACH-F) where a cluster hea is selecte at the network initializing state an remains fixe. The cluster hea rotation among noes is similar to the conventional LEACH. LEACH-F utilizes the centralise cluster formation algorithm from LEACH-C an has zero overhea at the network initializing state. Introuction to Transmission Control Protocol A full eployment of WSN on oil an gas pipelines has a severe network performance egraing issues especially on equal sharing of network resources among noes in a network. Sharing or managing of network resources is a crucial task particularly in a highly ensest multi-hop linear network with a limite 25

45 An Overview of Static Multi-hop Linear WSN resource [24, 74]. It is quite a common scenario in which a long range multi-hop linear WSN is prone towars inefficient fairness which leas to source noe starvation [23, 27] in a network with a Transmission Control Protocol (TCP) agent. In general, there are several types of agents that support a network such as TCP, User Datagram Protocol (UDP), Stream Control Transmission Protocol (SCTP), Datagram Congestion Control Protocol (DCCP) an etc. TCP is one of the most popular agents in many IEEE base application where various mechanism with high performance an congestion collapse avoiance is achieve. TCP is a reliable agent with a close loop process from a ata generation at a source noe until a receipt of an acknowlegement packet in return from the receiver to the respective source noe. The key function of this TCP mechanism is to coorinate the ata sening rate in a WSN at an optimum state. Avertise winow at the estination noe an congestion winow at the source noes oversee the banwith utilisation for the purposes of ata packet transmission. The newer versions of TCP mechanism are built with four intertwine algorithms such as slow start, congestion avoiance, fast retransmit an fast recovery as state in (RFC 5681) [75]. The slow start algorithm increases the congestion winow with every acknowlegement packets receive. This enables the congestion winow to be effectively increase base on every Roun Trip Time (RTT) value [76-78]. In a state where the congestion winow excees the set threshol of ssthresh, the congestion avoiance algorithm reorganises the congestion winow base on the type of TCP agent use in a network. In this perio, the congestion winow is increase by one for each RTT value until a lost packet is etecte in the network. The congestion avoiance algorithm has a unique proceure in hanling the realtime ata traffic in a ata stream base on the characteristic of a TCP agent [78, 79]. In general, a TCP agent retains a timer starting from the moment a ata packet is sent which will be use as a mechanism to etect a lost packet without acknowlegement packets receive after the timer has expire. It is quite common for the TCP agent to realise the lost packet cause by the lengthy waiting perio before an action is taken. Thus, a fast re-transmit algorithm was propose which uses a uplicate acknowlegement in etecting a lost packet (usually, is set 26

46 An Overview of Static Multi-hop Linear WSN to three). In a scenario where an acknowlegement packet is receive with the same sequence number at a source noe, the fast re-transmit algorithm consiers it as a lost ata packet an regenerates the ata packet for the secon time. The fast recovery algorithm reuces the size of a congestion winow by half an expans the congestion winow base on the usable winow with minimum (awn, cwn+nup) where awn is the avertise winow, cwn is the congestion winow an nup is the number of uplicate acknowlegement receive [23, 79]. The acknowlegement of new ata also known as recovery ACK is receive to return to congestion avoiance phase. One of a key metrics in TCP is RTT, whereby an RTT value is one cycle time from the source noe to sen a ata packet to the estination noe an to receive an acknowlegement packet in return. The TCP mechanism at the source noes employs a retransmission time-out (RTO) mechanism for an estimate RTT [23, 27, 79]. The etaile characteristic of RTT is as state in RFC 2988 [80]. Many types of research are conucte in enhancing TCP performance in hanling congestion, reucing errors an hanling the loss. Thus, there are many types of TCP algorithms available for users. Types of TCP A TCP agent is also known as a connection-oriente protocol in a transport layer. In general, the TCP agent implements a ata flow control, error control an congestion control mechanisms to establish a reliable en-to-en ata an acknowlegement packet transmission at the both en (sener an receiver) of a network [78, 79]. Thus, the TCP agents can be classifie in two agents, one at the source noe (sener) an another at the estination noe (receiver). All the TCP agents have ifference characteristic only when there is a lost packet etecte in a network. In a state where all ata packets have reache the esignate estination noe successfully, all the TCP agents operate at a similar proceure. TCP Tahoe TCP Tahoe refers to the congestion control algorithm as propose by Van Jacobson [81, 82] base on a concept of conservation of packets. The concept is unable to 27

47 An Overview of Static Multi-hop Linear WSN permit a new entry of ata packet if the network is operating at full capacity until a ata packet exits the ata stream. The network with TCP Tahoe acknowleges the sener for every ata packets sent in a sequential manner. The congestion winow in TCP Tahoe is maintaine to reflect the real-time network capacity uring the active perio. TCP Tahoe is base on a slow start proceure where at initialization or re-initialization ue to packet loss, the congestion winow is set to one an exponentially increase until another packet is lost. Thus, the congestion will be reuce an the network will be emptie when an agent encounters packet loss by timeout. When the agents encounter congestion in a network, TCP Tahoe saves the value of winow by half as a threshol for the next cycle with linear increments till the next packet loss. The timeout process cost lengthy elays until a new packet is generate at a source noe, therefore, it is not practical in a highly ensest network [79]. TCP Reno The TCP Reno maintains the funamental features of a TCP Tahoe such as slow start proceure however with an early etection of loss packets, the network is not emptie when to encounter a packet loss [83]. The TCP Reno acknowlege the ata packets receive at a esignate estination noe instantly hence, a sequence of an acknowlegement packet is generate base on a preetermine time (sufficient time to travel between a source an a estination noe) for the retransmission of a lost ata packet [76, 77, 84]. The fast re-transmit in TCP Reno is a proceure when three uplicate acknowlegement packets are receive at a source noe that inicates the probability of a packet loss in the network. This proceure will also reuce the Slow-Start Threshol (ssthresh) value to half of the winow size an set the same value for congestion winow [79]. The congestion winow will be increase by one for each acknowlegement packet receive an is limite to the amount of ata packets in the ata stream. TCP Reno is known as a reliable agent in a network with small losses of ata packets but will unerperform in a network with multiple ata packet loss in a single winow [23, 24, 27]. 28

48 An Overview of Static Multi-hop Linear WSN TCP New-Reno The TCP New-Reno is an enhance version of TCP Reno with fast re-transmission in a scenario with multiple ata packet loss in a single winow [83, 85]. Unlike in the fast re-transmit state in TCP-Reno, TCP New-Reno remains in the state of fast recovery until all outstaning ata packets are acknowlege at a perio of fast recovery [76, 79]. Hence, the reuction of congestion winow is not require as frequently as in TCP-Reno. The cost of one RTT is require to etect a packet loss from the sequence of issue acknowlegement packet from the estination noe. Selective Acknowlegements (SACK) The TCP SACK is an extension of both TCP-Reno an TCP New-Reno which works aroun the limitation mainly on etection of multiple packet loss per RTT value an retransmission of more than one packet loss [83]. TCP SACK acknowleges a ata packet selectively with a block inicating the packets being acknowlege. This helps the sener to trace any outstaning ata packets in the ata stream. In the event of fast recovery, a variable ata stream is estimate on the outstaning of ata packets an set the congestion winow by half. The ata stream is reuce by one for every acknowlegement packet an increase by one for every transmitte ata packet. In a state where congestion winow is larger than the ata stream, the unelivere packets are checke an retransmitte accoringly [76, 79]. Thus, more than one lost packet can be retransmitte in one RTT value. TCP Vegas The TCP Vegas is an extension of TCP Reno with a proactive mechanism to encounter congestion in a network before the packet loss occurs. TCP Vegas can also be consiere as a congestion avoiance mechanism that gives emphasis to packet elay rather than a lost packet [76, 86]. The proactive mechanism is utilise in controlling the congestion winow base on the real-time state of the network. Thus, it etects the congestion at an early state base on the value of RTT unlike congestion is only etecte as it occurs in a network using TCP Reno an TCP New- Reno. The issues with timeout are efficiently verifie in a scheule which also overcomes the uplicate acknowlegements in orer to etect a lost packet. The congestion in a network is not etecte by the means of a lost packet but it is 29

49 An Overview of Static Multi-hop Linear WSN etecte way before a packet is lost [77, 79]. The traitional mechanism for etecting a lost packet as in TCP Tahoe an TCP Reno base is still retaine in TCP Vegas. TCP Agent at a Receiver The TCP sink is the simplest TCP agent at a estination noe in a network with one acknowlegement packet is generate for each successfully receive ata packet. Whereas the TCP sink with elaye acknowlegement is a configurable elay (time) before acknowleging a receive ata packet at a estination noe [79]. The TCP sink with elaye acknowlegement can be configure base on the user s requirements. A TCP sink with selective acknowlegement is base on selective repeat retransmission policy. The receiver generates an acknowlegement packet base on the selection mae by the sener on the receive ata packets. The missing or lost ata packets can be retransmitte from the sener to the receiver for the secon time. A TCP sink with selective an elaye acknowlegement has a combination features of both selective an elaye acknowlegements [79, 83]. Base on the selective acknowlegement, the configurable elaye acknowlegement time can be ae at the estination noe. Relate Work in TCP Fairness The propose ACKs-First Variable-size Queuing (AFVQ) [87] mechanism incorporates the ACK-s first scheuling an a variable-size queuing metho to reuce the effects of network unfairness. In general, the acknowlegement packets are separate from ata packets that are transmitte prior to a ata packets transmission. Hence, the acknowlegement packets queuing elay is reuce. The variable-size queuing is able to reuce the number of acknowlegement packets in the queue by ropping the acknowlegement packets overflow from the output queue, therefore, creating a better allocation for ata packet transmission. The acknowlegement packet overflow is roppe an a new acknowlegement packet is generate to the TCP sener inicating that all ata packets are successfully transmitte. 30

50 An Overview of Static Multi-hop Linear WSN Aaptive Delaye ACK Mechanism (ADAM) [23] works by properly ajusting the TCP parameter which affects the throughput fairness an further reucing frame collisions. The ADAM is propose for a multi-hop implementation with unparallele ata transmission among all sener noes in a certain network. Thus, the elaye acknowlegement time for all sener noes is calculate base on the RTT value between a sener an a receiver noe to achieve optimum throughput fairness. The cycle time for a ata cycle is formulate base on the accumulate network RTT value in a multi-hop network. ADAM+ [27] is another metho introuce by the author to improve throughput fairness further by eliminating the request to sen (RTS) an clear to sen (CTS). With the RTS an CTS function being isable, it further reuces the routing overhea an creates higher ata flow towars the receiver noe in a network. The author also introuce ADAM in shorter noe istance (ADAM-sn) where two sener noes are locate in a communication range that enables a bypass communication with a ecrease in the accumulate network RTT value. Both ADAM+ an ADAM-sn are esigne base on the configuration of ADAM. TCP with Aaptive Delay Winow (TCP-ADW) [88] is esigne to reuce the acknowlegement packets appropriately by reaching an optimal ynamic elay winow. Base on the real-time changes in a network, the goal of TCP-ADW is to regulate the number of acknowlegement packets an to ynamically ajust (increase or ecrease) the elay winow. The number of acknowlegement packets is appropriately reuce by the propose moel by manipulating the elaye winow (own) referring to the size of congestion winow (transmission rate at a sener), the packet inter-arrival time (inicates congestion level an path istance) an lost packet event. The elay avois the transmission time-out at the source noes unless the retransmission timer expires, the elay winow will be increase by the estination noe base on the ata rate. The authors [89] have purpose a technique to ictate the buffering of ata packets in a network to achieve throughput fairness. The Propose Packet Reverse Function enhances the network performance by improving both the fairness inex as well as the average en-to-en elay from multiple sources in a network. A special characteristic for the propose scheuler enables better 31

51 An Overview of Static Multi-hop Linear WSN management for ata packet transmission with a reverse function which eliminates ata packet loss in a network. A ual buffer mechanism in each noe manages the overflow of ata packets generate by assigning them to a primary buffer in a neighbouring noe (moving towars the receiver noe). In a state where the primary buffer is at full capacity, the ata packet will be move to the seconary buffer in the same neighbouring noe. If the buffer state is at full then the ata packet will be reirecte to the seconary buffer of the sener noe unless. If the seconary buffer state is at full at the sener noe, then the ata packet is move to another neighbouring noe an will be roppe if the seconary buffer state is still at full. The propose technique creates more room to accommoate the generate ata packets in multiple locations (noes) until the ata packet is sent to the esignate receiver. The Intelligent Acknowlegement technique (TCP-IA) [90] incorporates the value of projecte banwith an elay to the MAC heaer to enhance the TCP performance particularly in a network with banwith unfairness. The ata packet arriving at a neighbouring noe, the link layer calculates the banwith an elay in a cycle repetition until the ata packet reaches the receiver noe. Any ata packets that are not in orer or with a low banwith an a higher elay is an inicator to reuce the elaying winow size at a receiver noe. In this state, an acknowlegement packet is generate to the sener. Otherwise, the receiver noe generates a single acknowlegement packet to the sener noe inicating the receive ata packet. After receiving the acknowlegement packet from the receiver noe, the sener noe regulates the size of congestion winow referring to the banwith an elay. Cooperation between channel access control is a new cross-layer scheme along with TCP Rate Aaptation (CATRA) [91, 92] will etermine an optimal Contention Winow (CW) size to achieve fair share on each TCP flow. The propose CATRA gathers useful information from both MAC an physical layers in orer to ynamically ajust the CW size to manipulate the contention between stations. The moifie CW size enables the network to achieve fair channel access perstation an an efficient spatial channel utilisation. The fair banwith allocation value is calculate for each TCP flow is irecte to the transport layer. Referring 32

52 An Overview of Static Multi-hop Linear WSN to the value obtaine, the TCP sening rate is ajuste on each flow to achieve fairness between flows hence results in throughput fairness. The authors in [74] propose a transmission scheuler algorithm to achieve improve network fairness an maximise the allocate banwith. The propose transmission scheuler algorithm is esigne to assign each noe with a weight referring to its location in the network an the aggregate traffic loa. The respective noe transmission time is erive base on the weight factor which is the number of hops from a gateway, the aggregate traffic loa inclues own an other ata packets an the number of interfering noes in the transmission range. The propose transmission scheuler algorithm improves the throughput of each noe as well as the overall network allocation fairness among all sensing noes. The authors in [93] propose a practical ual queue approach with a combination of Optimal Queue Selection known as the TCP-Optimal Queue Selection. The two ifferent queues are esigne to accommoate ata packets in one queue an the acknowlegement packets on the other. Base on the priority or probability, the optimal queue size is selecte to achieve a great amount of fairness in a network. In the priority an probability scheuling, there are two ifferent methos propose to assist the queues. In the priority scheuling, an acknowlegement packet has a higher priority whereas in the probability scheuling the queue selection is referre from the optimal probability. With the optimal probability, the ieal queue size, as well as the queuing elay for an acknowlegement packet queue for probability scheuling, is erive. Introuction to Energy Moel The sustainability of an implementation WSN on an oil an gas pipelines solely epens on a constant energy supply in a network. A multi-hop linear architecture has limite network resources an a set of iverse challenges in terms of energy optimisation in both infrastructure an non-infrastructure base system [12, 32]. Thus, an energy efficient routing protocol an energy optimisation technique are essentially neee to rive towars a feasible implementation on an oil an gas pipelines. Energy wastage in a multi-hop linear architecture has numerous impening factors against overall network performance since a single noe failure 33

53 An Overview of Static Multi-hop Linear WSN coul be a threat to cannibalise the entire network at a point [1, 2, 12]. Some of the primary reason of energy starvation an eclining the network lifetime in a WSN is ile listening, waiting perio in-between transmission, retransmission (ue to packet rop), higher routing overhea an overhearing [20, 23, 30, 94]. Therefore, an energy optimisation technique an managing of network resources are essential, particularly in a highly ensest multi-hop linear network. Generally, the unique geographical architecture of a long range multi-hop linear WSN has an imbalance or inequality use of network energy which applies to both infrastructure an a non-infrastructure base system [4, 6, 35]. The energy usage is often relate to the ata accumulating factor moving towars a centralise monitoring station (estination noe) in an oil an gas pipelines. Therefore, a step increase in energy consumption can be visualise on noes in a path moving towars the estination noe. It is quite a common scenario in a multi-hop linear WSN that is prone to inefficient energy usage that can result in energy starvation in a network with a stanar or inefficient energy moel. The five basic factors that influence the energy eman in a network are the network architecture, ensity of noes, istance between noes, routing management an ata rate [3, 17, 18, 95]. All these factors are epenent on the network architecture of a WSN. The two most common network architecture in a pipeline network is a flat (onetier) topology an a cluster-base (multi-tier) topology [15, 16, 96]. Energy versus Network Architecture A WSN is typically compose of a large number of noes being eploye in a specifie region of measurement. In a typical oil an gas pipeline scenario, the specifie region is often relate to an outoor harsh environment where noes are expose to the climatic effect [5, 8]. The noes in a pipeline network are statically eploye between a sensing point an a receiver point. In a multi-hop linear network inepenent to the network architecture, ata packets are transmitte through a series of noes to a centralise monitoring station (receiver noe) [6, 35, 96]. A sensing point can communicate irectly to a receiver point in a single-hop or a multi-hop communication through its neighbour noes in a network. Either way, the communication links in a typical multi-hop linear 34

54 An Overview of Static Multi-hop Linear WSN architecture relay ata packets to the estination noe that is in a istance away, resulting in high-energy consumption in the network [20, 31]. In typical flat one-tier multi-hop architecture, each of the source noes has a ual role where it acts as a sensing point as well as an intermeiate noe for the other neighbouring noes [97]. The noes in such an architectural esign are energy riven since both sensing as well as transmitting ata packets to a esignate noe has a high eman on energy consumption. The network nature of a multihop flat architecture may rain higher energy ue to the combine operation of a sensing an ata stream in a single path. Noe starvation is not a favourable factor in a network ue to higher eman for energy that coul lea towars noe failure which can consequently cause termination of the communication links in a multihop linear network [28]. In multi-hop hierarchical architectures, each source noes are organise into clusters with a eicate leaer noe known as the cluster hea. In each cluster, source noes are only permitte to communicate irectly to a eicate cluster hea at all-time whereas, one or more leaer noes are responsible for establishing a multi-hop communication path with neighbouring cluster heas to a estination noe [37, 97, 98]. In a static network environment, a leaer noe is ynamically selecte base on some set network criteria which inclue remaining energy, the istance between source noes with other cluster heas an noe homogeneity [61, 70]. The multi-hop hierarchical architectures are a complex structure of network type with noes at each level are esigne to perform a specific task which may vary from one level to another base on the routing protocol. Source noes locate at the lowest level is eicate for sensing changes on a sensor an transfers it to the leaer noe in a cluster. Base on the hierarchical architectures, the leaer noe may transfer it to a level above or to another neighbouring leaer noe in the network [72, 98]. The complexity epens on the number of levels create as well as the types of evices eploye in the network. Thus, the energy requirements in multi-hop hierarchical architecture may vary upon factors that contribute to the network complexity. 35

55 An Overview of Static Multi-hop Linear WSN Energy State in a Network The energy moel response base on the characteristic of a routing protocol an the network energy eman uring the network-active perio. The energy moel is responsible for the network power management in all network energy state [34]. The energy usage in a network enotes the various network activity that can be classifie as energy for transmitting, receiving, ile, sleep an transition [99-101]. In a practical implementation, the consume energy may vary upon the characteristic of a selecte harware which inicates the rate energy for each attribute on a network activity. Another factor that influences the increasing eman for network energy is the characteristic of a routing protocol especially the ones with a higher routing overhea [22, 57, 102]. The five most common energy moe is on, off, ile, sleep an transition are implemente to optimise the network energy [34, 103, 104]. Generally, the ile, transmit an receive states in a WSN can be constitute as a network-active state or as on state. During the on state, the source noes are permitte to transmit an receive packets at the cost of transmission an reception power respectively [34, 104]. Both the transmission an reception power inclues all the supporting packets (broacast an control) neee in a network. In any wireless evice, the power ratings in an active state are comparable with the types of network activity attempte by a noe. The highest energy attributes in a WSN is from the packet transmission state where maximum energy is consume, while noes in a receive state consume slightly lower energy than the transmission state but higher energy than in an ile state [20, 105]. In a network environment with no network activity for a uration of time, this state can be consiere as an ile state. However, a network with no ata packet transmission still cost the network with ile power uring the ile an listening phase of the neighbouring noes. The energy rating for each network activity epens on the type an the characteristic of the eploye wireless evice [106]. In the network-active state, all noes consume full energy in respect to the network activity. The sleep an off state in a WSN can be constitute as a network-inactive state or as off state. A network in a sleep state consumes energy at a lower rate than that 36

56 An Overview of Static Multi-hop Linear WSN at an ile state whereas, in the off state there is no consume energy in a WSN [34, 103]. In the sleep state, the raio is set at off which isables the noes from transmitting an receiving ata packets but on the backgroun, energy is consume for etecting signals base on the characteristic of a routing protocol. However, the consume energy at sleep state epens on the eploye routing protocol an network size. In a network environment where noes are set to off state, the energy consume is equivalent to zero [104]. In the off state, no network activity is permitte since the raio is at off state an unable to etect any incoming packets or even establishing an outgoing packet. This is a nonprouctivity state in a network which oes not consume any network resources. The transition in a WSN can constitute as a transition between the networkinactive state to network-active state. Transition in a WSN is a common activity in between ifferent state of energy moe with a time factor. The three most common transitions in a WSN is the state from ile to transmit, ile to receive an sleep to ile [34, 103]. During the specific transition perio, the transition state is attaine with an aitional energy inepenent to the other network activity. Generally, the transition perio takes into account of the raio activation perio in a wireless evice. In orer to ensure sustainability as well as to prolong the lifetime of a WSN, it is essential to properly outline the workflow of source noes, network architecture an an efficient energy moel. Over the years, an increasing number of research was conucte in optimising the energy moel especially in a remote monitoring an battery operate WSN [94, 101, ]. Numerous research works were relate to the implementation of an efficient, optimum an energy longevity in a pipeline network as a factor in a performance-riven WSN. Hence there are many types of energy moel available for users. Relate Work in Energy Moel The propose nw-mac [111] is an asynchronously scheule with multiple wakeups provisione uty cycle MAC protocol in a WSN. The nw-mac is an asynchronous scheule selection technique to setup the maximum number of n wake-up uring the receiver operational cycle. The propose nw-mac is best 37

57 An Overview of Static Multi-hop Linear WSN eploye in both low an high traffic application by utilising the reception winow-base meium access an limiting the reception opportunity at each wakeup. With the implementation of nw-mac, the optimum energy consumption in a network is achieve using the multiple wake-up techniques in each allocate wake-up cycle for noes. The authors in [101]ha moifie the stanar IEEE power moel to improve energy efficiency an en-to-en elay in a WSN. Accoring to the propose metho, a ata packet will be able to travel many hops between a source to a estination noe within a beacon interval. During the Dynamic A-hoc Traffic Inication Map (ATIM) winow ajustment is slice into the smaller initial slot (contention slot). An empty ata packet will be sent uring the contention slot which retains at on state for the sener an receiver noes enabling the sener noe to avertise with the ATIM packet. The receiver noe generates an ATIM- ACK packet to establish ata packets between the respective noes uring the beacon interval. All other noes in the network are excepte to be at sleep state instea of being at on state uring the ATIM winow to achieving better energy efficiency. The author ha propose [112] a technique to group the ata transmissions by efining the harmonising perio that is use in aligning the ata transmission from multiple application at interval bounaries. A protocol is esigne by referring to the harmonising perio which coorinates the ata transmissions across noes by ensuring real-time guarantee in a multi-hop WSN. The propose Network-Harmonize Scheuling (NHS) is a protocol which takes avantage from the perioicity introuce to allocate offsets across noes base on ifferent hop levels. The concept of ifferent hop levels ensures that collisions among ata packets will always be avoie with an enforce eterministic behaviour in a network. The NHS is esigne for a light-weight an scattere protocol without any global state-keeping mechanism in a network. The propose Receiver-Initiate X-MAC (RIX-MAC) [113] is an energy-efficient MAC protocol establishe with an asynchronous uty cycle. The RIX-MAC can improve the energy efficiency by utilising short preambles an aopting a 38

58 An Overview of Static Multi-hop Linear WSN receiver-initiate approach. The RIX-MAC is capable of minimising the energy consumption of sener noes by facilitating transmitters base on the preicte wake-up times of receiver noes. With the receiver-initiate ouble wake-up, scheule reuces the sener's control frames in a wakeup perio between a sener an receiver noe. This metho in RIX-MAC is able to reuce the energy consumption of receiver noes through reuce control frames. The Multi-hop power saving moe (MH-PSM) [114] is esigne to improve both en-to-en elay an sleep perio compare to the conventional power saving moe. The propose MH-PSM is extene from the stanar IEEE power saving moe that allows a WSN to switch states from a high power state to a low power sleep state in orer to save energy. The energy state switching is establishe through a traffic announcement scheme which is use in facilitating a multi-hop communication. The scheme broacasts traffic announcements throughout a multi-hop path that ensures all intermeiate noes are remaine at on state to receive an forwar any pening ata packets with minimum latency in a network. The Develope DEEC (DDEEC) [70] ynamically changes the cluster hea selection conitions for noes in a network base on their resiual energy. The propose [115, 116] Enhance DEEC (EDEEC) where the cluster hea selection probability is base on the energy level inicate as super an avance, that is higher than the normal noes. The EDEEC continues to penalise noes at super an avance energy level even at the same normal noes energy level. Therefore, noes with super an avance energy level will be terminate quickly compare to the normal noes in a network with EDEEC. The authors in [116] propose an enhance evelope istribute energy-efficient clustering (EDDEEC) scheme to be implemente on a heterogeneous WSN. The EDDEEC primarily incorporates three elements which are the heterogeneous network moel, an energy consumption moel an a clustering-base routing mechanism. The propose heterogeneous network moel is base upon three energy level of noes in a network. The energy consumption moel in EDDEEC consiers the impact of raio environment. The clustering mechanism in EDDEEC enables the probability of the selecte cluster hea in an efficient an ynamic metho. 39

59 An Overview of Static Multi-hop Linear WSN The authors in [108] propose a metho on energy saving moel for improving network lifetime in both homogeneous an heterogeneous environments. The propose metho introuces the noe pairing concept where the pairing noes are place close to one another in a WSN. The propose energy moel optimises the remaining energy level of all sensor noes (etection noes) in a network by ON an OFF scheuling technique for ata transmission. The two-phase switching metho among sensor noes enables the ata transfer in phase one for a series of noes in ON state is able to transmit ata packets to the base station while the other noes will be set to OFF state. Whereas in phase two, initial noes in ON state is set to OFF state an initial OFF state noes is set to ON state for ata packet transmission to the base station. The cluster hea is selecte base on the noe with highest remaining energy in a specific region. The propose power switching between ON an OFF state allows optimisation of energy usage in a network with longer network lifetime. The authors in [117] propose the Energy-Delay Unifie Routing protocol [EDURo+] which incorporates Prim s- Dual algorithm that consiers five funamental parameters to exten the lifetime of a sensor noe in the network. The consiere parameters are resiual energy in noes, the rate of packet loss, transmission power, receiving power an interference to etermine the next hop. The propose metho is critical to event monitoring; thus, the sleep an wake scheule can be amene from a low power state to a high power state. All noes in the network are set at low power state unless the noes in a ata transmission process are switche to high power state. The multi-hop istance-base an low-energy aaptive clustering (MH-DISCPLN) [118] is an energy efficient protocol in multi-hop cluster-base WSN. The MH- DISCPLN protocol introuces the concept of multi-hop clustering with an equal number of selecte cluster heas in each roun. The selecte cluster heas are from the mile an outer rectangle regions. In the mile region, the cluster hea is selecte base on highest remaining energy to aggregate an forwar ata packets from cluster members as well as the outer region cluster hea to a esignate base station. The outer region cluster hea can only transfer the receive ata packets to the nearest mile rectangle cluster hea. A selecte 40

60 An Overview of Static Multi-hop Linear WSN cluster hea will not be a cluster hea again until all the member noes in that region ha the opportunity to be a cluster hea. The MH-DISCPLN protocol is capable of increasing the network lifetime as well as achieving higher throughput with lower elay an packet loss. Conclusion The overview an ifferent aspect in the application of a static multi-hop linear topology were presente in this chapter. The technical characteristic an features of a static multi-hop linear network were briefly escribe in terms of network topology, resources unfairness an energy constraints. The applications of a static multi-hop linear network were highlighte with ifferent types of relate works that contribute to the overall network performance. In the next Chapter, a routing algorithm for both flat (one-tier) an a cluster-base (two-tier) topology for enhancement of overall network performance will be presente. A ual interleaving scheme for multi-hop linear topology can further improve the network performance in a pipeline network. 41

61 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Chapter 3 3. Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Introuction Wireless sensor network (WSN) is wiely use in the monitoring of oil an gas pipeline integrity. In a large scale linear WSN eployment benefits from many prefixe sensing locations through the entire strength of an oil an gas pipeline [4, 8]. A long range multipoint linear WSN communicates valuable information to a centralise remote monitoring station which is usually locate miles away from the sensing locations. A multipoint linear WSN has no fixe central authority to gather the collecte ata but every single sensor noe on a certain network carries not only its own ata but also the traffic from every other noe in the pipeline network towars the estination noe [12, 96]. Figure 3-1 illustrates these characteristics of an establishe route between Nn to ND on a multipoint linear WSN at time t. (DP+CP)Fn (DP+CP)F5 (DP+CP)F3 (DP+CP)F2 Communication range of Nn Nn n hops N5 5 hops N4 4 hops N3 3 hops N2 2 hops N1 ND 1 hop Figure 3-1: Multi-hop linear WSN Data packets are inicate as DP an control packets are inicate as CP from respective noes inicate inflows, Fn where n represents the maximum number 42

62 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network of flows containing bi-irectional packets moving towars a estination noe. In such a WSN setup enables communication with every single noe in its communication range irectly to move the collecte ata forwar to a esignate noe [12]. The benefit of a multi-hop linear WSN layout is the ability to establish a communication route among the neighbouring noes within its communication range to transfer ata packets towars a esignate noe (ND) [15, 119]. With an implementation of a reliable routing algorithm creates the ability to automatically generate a route, etect real-time route breakage an also correct them as neee. The other benefit with a multi-hop linear WSN is the capabilities to outsprea a larger coverage area by aing more noes to the existing network. The most important features of WSN in monitoring oil an gas pipeline integrity will be scalability, reliability an robustness [4, 13]. The overall wireless network performance is critical for the sustainability of the network in the long run. However, ue to its unique geographical linear WSN structure on pipelines, the three foremost rawbacks of such a topology are; (1) amplifie complexity which leas towars (2) noe failures hence (3) eclining overall network performance [35, 96, 120]. A multi-hop linear WSN tens to be challenging to regain from noe failures ue to a large number of network changing variables to maintain the network. A multi-hop linear WSN typically has limite an reuce capacity compare to other known wireless topologies, ue to the amplifie overhea in managing the network routing as well as increase contention in traffic towars the estination noe. The accumulation of ata from each inepenent noes as shown in equation 3-1 has to share the same path to transfer the ata towars the estination noes as shown in Figure 3-1. = ( + ) Where TP is the total packets for n number of noes, DPi is the total ata packets an CPi is the total control packets at noe i with 1 i n. While DPj is the total ata packets at neighbouring noe j, CPj is the total control packets at neighbouring noe j an IfQleni is the efault queue size set at 50 packets in a network [12]. 43

63 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Figure 3-2 illustrates the route breakage at N4 on a multi-hop linear WSN at time t+1 that creates a link breakage that results in communication interruptions beyon N4. Such a scenario creates an unfavourable state for the active noes locate beyon N4. In any multi-hop linear WSN, this is a crucial issue that leas towars loss of important ata from the oil an gas pipeline. (DP+CP)Fn (DP+CP)F5 (DP+CP)F3 (DP+CP)F2 Communication range of Nn Nn N5 N4 N3 N2 N1 ND n hops 5 hops 4 hops 3 hops 2 hops 1 hop Figure 3-2: Route breakage in a multi-hop linear WSN In a real life set-up, all noes are consiere as a sensing point which is likely to sen ata packets to a estination noe simultaneously. In such a scenario, there are higher chances of ata packet accumulation on a certain noe in the network as shown in equation 3-1 which will buil up an result in a bottleneck state [23, 28]. In accorance with IEEE stanar, there are a number of crucial factors which make the linear topology least popular compare to the other known topologies [18, 121]. Some of the crucial factors in WSN is the transmission range, carrier sensing range, queue size, transmission power, battery lifetime an banwith. Nevertheless, a multi-hop linear WSN is a useful topology in many circumstances where the geographical factor of the application matters the most. These factors coul be manipulate by overwriting or with an improve routing algorithm to enhance its existing performance [95]. This chapter mainly focuses on two types of routing algorithm applie on a flat (one-tier) an a cluster-base (two-tier) topology. The Dual Interleaving Linear Static Routing (DI-LSR) [12] is a routing algorithm establishe for a flat multi-hop linear topology while the Dual Cluster Hea Interleaving Linear Static Routing (DCHI-LSR) [96] is a routing algorithm establishe for a multi-hop linear cluster- 44

64 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network base topology. Both routing algorithms are esigne to accommoate an increasing amount of traffic in a multi-hop linear WSN with the possibilities to minimise passive noes. Routing Stability Factor in Linear Static Topology The successfulness in sening a ata packet from the originator (source noe) to a receiver (estination noe) in WSN epens on the optimum path generate by the routing protocol. Every routing protocol has its own characteristic an limitations which are known as routing factors against the measure performance metrics base on the type of topology it s been implemente [3, 57]. Routing factors in a WSN have a great impact towars the overall network performance an in most cases results in poor network life cycle in the long run making it unfeasible solution for a pipeline network. Unerstaning the root cause of this factors will help to ientify methos to eliminate or minimise the effects of routing factors which is then incorporate in the propose routing algorithm. In the following section, these factors an relate parameters are efine with a feasible solution for the pipeline network. Queuing Factor in WSN In a large WSN, networks of queues among noes are a common system where noes will have a specific queue length to accommoate bi-irectional packets passing in an out on a respective noe at a specific time t. In a fraction of time t+1, the packets on a queue change it state base on the real-time status of a certain network an the type of routing protocols being implemente. Base on the queue length (ifqlen) in [34, 122], bi-irectional packets on the queue will be controlle by a scheuling system known as first in first out [24, 78]. The biirectional packets arriving at a respective noe will be place in the queue an will leave the noe on a controlle sequence which is base on the implemente scheuling metho. There are many research an methos of a queue is being eploye to accommoate an manage these bi-irectional packets efficiently with minimum buffer time. In a WSN with high occupancy rate for example, in a linear topology with bi-irectional packets arriving at maximum queue limit, can result in queue overflow an the respective packets will be roppe [23, 27]. Each 45

65 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network roppe packets will then be regenerate base on the routing protocols implemente in the network. In general, a network with an efficient queue management system will reuce the unwante roppe packets as well as reuce the queuing buffer time [28]. The queue factor is very crucial in a linear topology ue to its amplifie bi-irectional packets travelling in a single multi-hop linear path from the source to the esignate estination noe especially in one sink point network (monitoring station for oil an gas pipelines) as shown in Figure 3-3. The amplifie volume of ata moving towars a single sink point (estination noe ND) from respective source noes will buil up the queue length at maximum limit (ifqlen) especially for the noes nearest to the estination noe [26, 28]. This state can be also consiere as a bottle neck points in a linear topology. As a result from this scenario, the ata packets will be roppe which is consiere a waste of available resources such as energy, time an the ata packet which is the most important component in a certain application. Biirectional packets ifqnn Communication range of Nn Nn Biirectional packets ifqn2 n hops N2 Biirectional packets ifqn1 2 hops N1 Biirectional packets ifqnd ND Communication range of ND 1 hop Figure 3-3: Biirectional packets queue scheme in a linear WSN Time to Live in WSN The mechanism which ictates an limits the lifetime of a generate ata packet from a certain noe is calle time to live (TTL). The TTL of a ata packet which can also be escribe as hop limits iscar a generate ata packet once the hop limit or time span has elapse [23, 60]. The core function of a TTL mechanism is an inicator which ientifies the source noe on the generate packet have been too long in the network an shoul be iscare. The TTL is initially set by the source 46

66 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network noe (sener) base on the characteristic of a routing protocol with a precalculate hop limit between a source an a estination noe. The number of hops to the estination is usually inicate in the routing table where each noe that receives the ata packet will subtract the hop count by one to inicate the remaining hop limit to the next noe in its path towars the estination noe [12, 102]. The receiving noe will forwar the ata packets to the next noe if the remaining hop count is greater than zero, otherwise, the ata packet will be iscare at that point. Once a ata packet is iscare, a control message will be generate base on the transporting agent to a source noe (sener) which will regenerate the ata packet for the secon time. The core function of a TTL or hop limit mechanism is to keep track of any ata packets which are uneliverable to a estination noe from circulating or clogging the network without a proper path. There are also possibilities for a route generation for creating an incorrect routing table which contributes to issues with the unknown path. This is a rare scenario in a static routing scheme as there no changes in noe locations. In a routing table generation process, an optimum path will be ientifie at the source noe or at the intermeiate noes base on the routing protocol use. A goo routing protocol will generate an optimum path with exact hop limit between the source an a estination noe to ensure an ieal time for the ata packets to reach its estination noe before it gets iscare. The efault hop limit or hop limit with successively higher TTL value coul result in a waste of network resources, especially on overall elay, elivery ratio an consume energy. Broacast Packets A broacast packet is a mechanism that enables a WSN to efficiently share an upate their ata path among noes in its communication range. Broacast packets have a significant role in initializing a network by generating broacast packets for network iscovery process, ientify a possible path to a estination noe an as a query for a respective ata packet in the network [21, 62]. In a large WSN, a broacast packet is use as an efficient metho for exchanging their local measurements with all the noes in a certain network. The broacast packets are generate at ifferent time intervals base on the characteristic of routing protocols to maintain the network stability an connectivity among noes. One of 47

67 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network an establishe metho of broacasting is known as flooing where each an every noe in the network will rebroacast when the first broacast packet is receive for the first time. In a practical application, flooing is known as broacast storm problem that creates serious broacast reunancy, waste of banwith an packet collisions. Since broacast packets are bi-irectional an the methos of generating a broacast packet are epenent on the characteristic of a routing protocol, the effects in a certain network will iffer from one routing protocol to another [22, 38]. Broacast packets in a highly ense WSN increase the routing overhea an network energy consumption especially in a linear topology which has to accommoate bi-irectional packets in a single path. A routing protocol with an efficient technique for reucing broacast packet reunancy increases banwith capacity an reuces energy wastage in a WSN. Propose Linear Static Routing In a multi-hop linear topology, a routing protocol with one-tier or multi-tier approach can be implemente base on the scale an complexity of an application. There are two types of routing algorithm in the propose multi-hop linear static routing that can be applie on a flat (one-tier) or a cluster-base (multi-tier) topology. The Dual Interleaving Linear Static Routing (DI-LSR) [12] is the propose routing algorithm one with a ual interleaving technique in route generation between source noes (multiple seners) to a single sink point (receiver) best implemente on a flat (one-tier) multi-hop linear topology. The Dual Cluster Hea Interleaving Linear Static Routing (DCHI-LSR) [96] is the propose routing algorithm two with a ual interleaving technique in route generation between cluster heas (leaer noe an intermeiate noes) to a estination noe (receiver) best implemente on a cluster-base two-tier multihop linear topology. Dual Interleaving Linear Static Routing In the propose Dual Interleaving Linear Static Routing (DI-LSR), a ata packet from a source noe (sener) is forware through preetermine intermeiate noes to a estination noe (receiver). The ual interleaving routes are generate base on a preefine routing conition referring to the o an even sequence of 48

68 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network source noes in the network. The DI-LSR is esigne to complement the routing stability factor with high link reliability an to enhance overall network efficiency in a pipeline network. The DI-LSR is a passive routing algorithm where changes in the network conitions are not esire through its operation. An infrastructure base WSN requires only minimum or no changes in the route selection since most of the variable parameters which influence a routing protocol is in a static (fixe) state. The noes in such a set-up are permanently locate at a uniform interval subjecte to the communication range of a wireless evice with a constant power source. Base on these factors, the DI-LSR esign has two major moifications mae; (1) to accommoate the preefine routing path function an (2) to reuce network routing overhea. The most funamental structure in DI-LSR is as shown in Figure 3-4, where two inepenent interleaving paths between each source noes (On/En) are generate to a single estination noe (ND). n hops 2 hops 1 hop En On E2 O2 E1 O1 ND 2 hops 1 hop n hops Figure 3-4: DI-LSR on a flat multi-hop linear topology In a multi-hop linear topology, the overall network performance an link stability are often relate to noe placement particularly in a highly scattere network of noes [123]. The major rawback on a highly scattere linear network of noes is the effects of passive noes. Passive noes are escribe as source noes with no opportunity to transmit ata or with a elivery ratio of zero in a network [23, 28]. A network stability factor which relates to link connectivity an passive noes between source an a estination noe relies on the number active noes in a specific communication range. Optimum placement of noes an the propose 49

69 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network interleaving technique in a multi-hop linear topology woul minimise the number of passive noes in a WSN. Referring to Figure 3-4, the noe placement in DI-LSR is uniformly istribute at istance in meter within a maximum transmission range of two for a pair of o an even sequence of source noes. In general, a routing protocol generates or upates its routing table base on available noes in a transmission range at the initializing perio before the first ata packet is sent [3, 124]. A routing table is generate at a first active perio of time t, will then be constantly upate at time t+1 as traffic eman increases at a specific noe in the network. A common process for generating a routing table epens on a chain sequence of route upates between a source an a estination noe which will be fully or partially store in respective noes. Routing Table Generating Algorithm in DI-LSR Unlike a stanar practice of generating a routing table, at time t, DI-LSR generates two preefine routing table from a sequence of o an even source noes eploye in a network as shown in Figure 3-4. With the elimination of both broacast an hello packets, the routing table for DI-LSR is generate base on an ieal network environment by assuming noes that are always in a reay state. With the elimination of physical search for neighbouring noes an route search, hence reuces buffer time in-between start to the first packet generate by a source noe. The routing conition in DI-LSR permits o an even sequence of source noes to sen an receive ata packets along with the control packets only in a preefine path in the network. There is three essential information neee in generating a forwar an a reverse path routing table in DI-LSR which is the number of noes in the network, path irection for ata communication an travel cost in hop count between a source to a estination noe [12]. The preefine forwar an reverse path routing table generating process in DI-LSR is as briefly escribe in the pseuocoe of Figure

70 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Algorithm 1: Forwar an reverse path routing table Suppose number of noes is nn, o sequence of noes is On, even sequence of noes is En an estination noe is ND = 0 At time t: 1: IF nn%2 == 0 then: Forwar path routing table 2 (for even sequence of noes) else: Forwar path routing table 1 (for o sequence of noes) 2: For nn then: Reverse path routing table (for o an even sequence of noes) 3: For On/En-2 0 then: Neighbour noe/noes at forwar path else IF On-1=0 then: Neighbour noe at forwar path 4: For On/En+2 nn then: Neighbour noe/noes at reverse path 5: For (On+1)/2 then: Require hop/hops to estination noe for forwar/reverse path (o) 6: For En/2 then: Require hop/hops to estination noe for forwar/reverse path (even) 7: At time t+1: Source noes On/En start sening ata packets to ND 8: At time t+n: ND start sening TCP acknowlegement packets to On/En Figure 3-5: Routing table generating process in DI-LSR The step by step process involves o an even sequence of source noes segregation, ata flow irection an hop count between respective source noes to a estination noe. The segregation of o an even sequence of noes in respective forwar routing table entries is create base on the total number of noes in a network. Once the segregation process is establishe, noes are only allowe to communicate with the selecte noes in the routing table. Once the routing table entries for o an even sequence of source noes are create, forwar (to a estination noe) an reverse (to all source noes) path is 51

71 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network preefine with available noes in the respective routing table. In a forwar path routing table, a respective noe hols the neee entries of other source noes base on the segregation process to a esignate estination noe. The forwar an the reverse path routing table in DI-LSR is as briefly escribe in Figure 3-6 an Figure 3-7 respectively. n hops 2 hops 1 hop Preefine path for even sequence of noes Nxt Dst HC Nxt Dst HC Nxt Dst HC E2 ND n E1 ND 2 ND ND 1 E2 E1 2 E1 E1 1 E2 E2 1 En On E2 O2 E1 O1 ND Nxt Dst HC Nxt Dst HC Nxt Dst HC O2 ND n O1 ND 2 ND ND 1 O2 O1 2 O1 O1 1 O2 O2 1 Preefine path for o sequence of noes n hops 2 hops 1 hop Figure 3-6: Forwar path routing table in DI-LSR n hops En 2 hops 1 hop Preefine path for even sequence of noes Nxt Dst HC Nxt Dst HC En En n-2 E2 E2 1 E2 En n-1 On E2 O2 E1 O1 ND Nxt Dst HC Nxt Dst HC On On n-2 O2 O2 1 O2 On n-1 Preefine path for o sequence of noes n hops 2 hops 1 hop Nxt Dst HC O1 O1 1 E1 E1 1 O1 O2 2 E1 E2 2 O1 On n E1 En n Figure 3-7: Reverse path routing table in DI-LSR 52

72 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network In a reverse path routing table, the estination noe hols the entries of all source noes an a respective noe hols the neee entries of other source noes in the network with preefine sequence from the segregation process. The istance from a respective source noe location to a estination noe is calculate in terms of packet travel cost in hop count. In DI-LSR, the hop count is always maintaine at minimum travel cost between a source an a estination noe in orer to optimise the network resources wisely. Source Noe Routing Flows in DI-LSR In a large wireless network, queue restriction on noes is a stanar mechanism which is use in controlling an managing bi-irectional flowing packets. A queue restriction has a crucial factor in reucing passive noes an bottleneck points in a single path wireless network. With the queue restriction [122], there will be increasing rate on packet roppe especially when the route is not ientifie to a esignate noe as in a conventional routing protocol. With the implementation of DI-LSR, this factors can be reuce an further increases the ata packet transfer rate with the propose ual path metho. A efault queue limit [34, 103] at 50 packets is set at each intermeiate noes in both paths as escribe in the flow chart of Figure 3-8. Any ata packets arriving at an intermeiate (neighbouring) noe with queue length > IfQlenOn or IfQlenEn will be roppe from moving forwar towars the estination noe. This state is known as queue overflow where the roppe packets will result in an unwante waste of network resources especially the energy consume from packet generation till packet roppe. Generally, a queue overflow is ue to overwhelming of control packets which inclue broacast an hello packets epening on the characteristic of a routing protocol. The elimination of both broacast an hello packets in DI-LSR creates less traffic, thus packet rops can be reuce compare to a routing protocol which relies on this packets to generate an maintain a routing table. Referring to the flowchart in Figure 3-8, a generate ata packet from a sensing point will be forware from a source noe to the esignate estination noe through a preefine sequence of intermeiate noes base on the routing table entries. 53

73 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Start: Data packet Start Simulation Routing table an ata flow process at the source noes O numbere source noe routing table *Generating Preefine source noe routing table Even numbere source noe routing table Data packets sen request from the O numbere source noe Data packets sen request from the Even numbere source noe Data flow process at intermeiate noes Repetitive process for n number of intermeiate noes Yes Is the biirectional packets Queue size? No No Is the biirectional packets Queue size? Yes Data packets in noe On queue Data packets roppe ue to queue overflow Data packets in noe En queue Data flow process at estination noe No Is the biirectional packets Queue size? Yes Data packets roppe at estination noe ue to queue overflow Data packets in queue at estination noe En: Data packets Data packets receive at estination noe * Refer to pseuucoe algorithm 1 A Figure 3-8: Flow chart for ata packet cycle in DI-LSR Distributing traffic in two paths further reuces the routing overhea by half, therefore allocates a better proportion of ata packets compare to any conventional single-path routing protocol. Data transfer accumulation factor for 54

74 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network both paths can be escribe respectively in equation 3-2 an 3-3 between a source an intermeiate noes towars a estination noe. = ( + ) Where TPO is the total packets for nn number of noes (o), DPOi is the total ata packets an CPOi is the total control packets at noe i with 1 i nn. While DPOj is the total ata packets at neighbouring noe j, CPOj is the total control packets at neighbouring noe j an IfQlenOn as the queue length in the network. = ( + ) Where TPE is the total packets for nn number of noes (even), DPEi is the total ata packets an CPEi is the total control packets at noe i with 1 i nn. While DPEj is the total ata packets at neighbouring noe j, CPEj is the total control packets at neighbouring noe j an IfQlenEn as the queue length in the network. A estination noe is place at en of each interleaving route with an integration point as inicate in the flow chart of Figure 3-8. The successful ata packets accumulation factor at the estination noe (receiver) from both routes is as escribe in equation 3-4. Any incoming ata packets at the estination noe with queue length > IfQlen will be roppe [103]. = Where NTP is the network total packets at the estination, the value of TPO is from equation 3-2 an the value of TPE is from equation 3-3. Destination Noe Routing Flows in DI-LSR A estination noe routing table hols entries of all source noes eploye in a network base on two preefine reverse paths. At time t, DI-LSR generates a reverse path routing table in which will be use by control packets to reach a esignate source noe. Once a ata packet has successfully reache the estination noe, a TCP acknowlegement packet will be generate in return to acknowlege the respective source noe (sener) [79]. Unlike the forwar route as escribe in Figure 3-8, the reverse path routing table has the entire list of both o an even sequence of source noes. The preefine reverse path routing table 55

75 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network generating process in DI-LSR is as briefly escribe in the pseuocoe of Figure 3-5. The hop count is always retaine at minimum travel cost between a estination to a respective source noe in orer to optimise the network resource wisely. The reverse path hop count is base on the same hop count in a forwar path between two respective noes in the network [12]. At time t+2, a generate TCP acknowlegement packet is forware on a preefine return path to a esignate source noe with a efault queue limit [122] set at each intermeiate noes in both paths as escribe in the flow chart in Figure 3-8. Since the routing path is preefine with no changes is expecte uring the noe active perio, no aggressive broacast packets are require to maintain neither the routing table nor the routing path in DI-LSR. Performance Evaluation on Flat One-tier Linear Topology The overall network performance was simulate in Network Simulator 2 version 2.35 (NS2) [34, 125] to visualise the behaviour an impening factors when full eployment of DI-LSR in a pipeline network. The DI-LSR was implemente on NS2 where etail analysis on various wireless performance metrics was compare with other routing protocols such as AODV, DSDV an FRP. The propose DI-LSR were compare with AODV a reactive routing protocol, DSDV a proactive routing protocol [22, 62] an FRP a manual routing algorithm [64, 65] for performance comparison purposes. All the simulate results are from an average value of five runs per cycle on the manipulating variable (nn number of source noes) with ifferent see values in between 1 to 10 over a simulation uration of 500 secons. In a flat one-tier topology, the number of source noes in each cycle was uniformly increase by 12 noes starting from 12 to 120 source noes. A ata rate of one packet/sec with a packet size given in Table 3-1 were assigne with a ranom start function to all source noes in the network. To replicate a real-time application, the start time an start sequence of a source noe are generate using a custom ranom function in the simulator. The start time for a source noe is generate ranomly between 0 2 secons with non-sequential starting noes. All noes inicate in the results were stationary for the entire simulation uration with one estination noe (ND). The configuration an key simulation parameters 56

76 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network employe in NS2 in simulating the effects on varying source noes ensity is as tabulate in Table 3-1. Table 3-1: NS2 simulation parameters for flat topology Channel type Key parameters Raio propagation moel Wireless channel Two Ray Groun Set value MAC type Interface queue type (ifq) Drop Tail/PriQueue Source noes (SN) 12, 24, 36, 48, 60, 72, 84, 96, 108, 120 Destination noe (ND) 1 Queue length (ifqlen) Agent type Traffic type Transmission range (RX Thresh) Sensing range (CS Thresh) Packet size 50 (packets) Transmission control protocol (TCP) Constant bit rate (CBR) 100 meter 125 meter 512 bytes Performance Metrics on Flat One-tier Linear Topology To test an evaluate the propose algorithm, a number of varying scenarios were create in the simulation tool. The performance characteristic of the propose algorithm can be visualise on the following wireless performance metrics: Packet Delivery Ratio In any wireless network, the elivery ratio is a crucial parameter measure which inicates a percentage of successfully receive ata packets over sen ata packets in the network [22, 34, 62] as escribe in equation

77 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network = % 3-5 Where RPi is the total receive ata packets for nn number of source noes an SPi is the total sen ata packets for nn number of source noes. En-to-en Delay En-to-en elay is an average value of the total time taken to transmit ata packets from a source noe (sener) to a estination noe (receiver) in the network [22, 34] as escribe in equation 3-6 (10). Data packets which are roppe or lost in route becomes invali uring calculation. ( ) 3-6 Where En ti Start ti is the total of Duration ti for nn number of source noes an RPi is the total receive packets for nn number of source noes. Throughput Throughput is the total receive ata (ata packets) at a estination noe in per unit time usually measure in bits/sec. Throughput is an important parameter where the goal of a network is to achieve higher throughput with the available network resources. The average throughput over all flows in the network is calculate as in [34, 126] equation 3-7. = ( ) ( ) 3-7 Where Pkt size is as efine in the simulation parameter, RPi is the total receive packets for nn number of source noes an En ti Start ti is the total of Duration ti for the entire simulation uration. Throughput Fairness Inex In a linear topology, fairness or equality within a network is a crucial factor in terms of network stability. The Jain s fairness inex as escribe in equation

78 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network where an optimum inex value of one is the esire value for a non-zero an fairest ata allocation among the source noes in the network [23, 29, 127]. = 3-8 Where ni is the throughput for n number of flows an nn is the number of source noes in the network. Receive Data Packet Variation The variation of receive ata packets is the measurement ifference between a maximum an a minimum number of receive ata packets from source noes in a network as escribe in equation 3-9 [12]. ( ) 100% 3-9 Where Max pkt is the maximum number of receive ata packets recore an Min pkt is the minimum number of receive ata packets recore for a certain number of source noes in the network. Passive Source Noes Passive source noes or noes without successful opportunity to transmit ata packets to the estination noe are unesirable in a multi-hop linear topology. The passive source noes can be calculate as in equation 3-10 [12]. 100% 3-10 Where PSN is the number of passive noes for nn number of source noes in the network. Normalize Routing Loa The normalise routing loa can be efine as the total number of routing packets over a receive ata packet at a estination noe as escribe in equation The high value of normalise routing loa correspons to higher network usage or cost of network resources that is not a esirable aspect in a WSN [128, 129]. 59

79 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network = 3-11 Where Total routing pkt is the number of supporting packets use to sen an receive ata packets an RP is the total receive ata packet in a network. Energy Consumption Energy consumption in a network can be efine as the total use energy over total receive ata packets in the network as escribe in equation Where Total consume energy is the total use energy by all noes an RP is the total receive a ata packet in the network. Simulation Results an Discussion for Flat One-tier Linear Topology The overall network performance of DI-LSR is compare with the traitional AODV, DSDV an FRP for flat one-tier topology in terms of the performance metrics in section 3.5. Packet Delivery Ratio vs. Number of Source noes Delivery ratio (%) AODV DSDV DI-LSR FRP Number of source noes in network Figure 3-9: Packet elivery ratio vs. number of source noes 60

80 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network The effect of varying number of source noes on the elivery ratio is as shown in Figure 3-9. Generally, the elivery ratio is inverse proportional to the increasing number of source noes for all compare routing protocols. It is observe that at low network ensity with 12 source noes, the elivery ratio is almost at the same rate among all compare routing protocols. The elivery ratio of DI-LSR outperforms all the other routing protocol between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. There is a significant ifference in elivery ratio between 5% - 14% better than FRP, 3% - 17% better than AODV an 2% - 30% better than DSDV at varying simulation environment compare to DI-LSR. The effect of reuce control packets particularly with broacast packets enables better allocation of ata packets transfer rate hence improves the elivery ratio in DI-LSR in all varying stages. Another possible factor in achieving higher elivery ratio is base on a preefine ual routing path where the generate ata packets are transferre to the estination noe remains constant at all time. En-to-en Delay vs. Number of Source noes AODV DSDV DI-LSR FRP 8.0 En-to-en elay (ms) Number of source noes in network Figure 3-10: En-to-en elay vs. number of source noes The results comparison on average en-to-en elay for varying number of source noes is as shown in Figure 3-10, which inicates DI-LSR has a consistent increase 61

81 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network in elay proportional to the increasing network ensity with 12 source noes to a larger network ensity with 120 source noes. Generally, with the higher ata rate at the estination noe increases the overall elay in any network particularly in a multi-hop linear topology since the position of each source noe is in a respective istance from a estination noe. The elay factor in DI-LSR is fairly low compare to the moerate fairness an higher ata rate achieve compare to the other routing protocols. In a multi-hop linear topology, the elay factor varies from the istance between a source an a estination noe. Using a preefine ual interleaving path, DI-LSR is capable of forwaring higher ata in a more efficient manner compare to a single path multi-hop metho. Throughput vs. Number of Source noes Throughput (Kbps) AODV DSDV DI-LSR FRP Number of source noes in network Figure 3-11: Throughput vs. number of source noes The effect of varying number of source noes increases the throughput in the network for all compare routing protocols as shown in Figure The throughput achieve by a network with the implementation of DI-LSR outperforms all other routing protocols between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. The DI-LSR has emonstrate a significant ifference in throughput between 20 Kbps 61 Kbps particularly after 24 source noes an above as compare with the other routing 62

82 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network protocols. Higher throughput in DI-LSR is a result of the better proportion of ata packets transfer rate in all varying stages using the preefine ual interleaving path to a estination noe. Throughput Fairness Inex vs. Number of Source noes DI-LSR AODV DSDV FRP Throughput fairness inex Number of source noes in network Figure 3-12: Throughput fairness inex vs. number of source noes Achieving a fair throughput among source noes in a multi-hop linear wireless network is a challenging task with any routing algorithm. Referring to Figure 3-12, the throughput fairness inex of DI-LSR outperforms all the other routing protocol with a significant ifference in fairness inex that is clearly visible in a range of 0.13 to 0.25 between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. In a low network ensity with 12 noes, fairness is harly visible since the performance of all routing algorithms is fairly similar. The higher rate of throughput fairness is comparatively achievable with throughput equality in a multi-hop linear network the propose DI-LSR. Receive Data Packet Variation vs. Number of Source noes Variation of receive ata packets is another parameter in visualising fairness between source noes in a network. Referring to Figure 3-13, the percentage of receive ata packet variation of DI-LSR outperforms all the other routing 63

83 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network protocol with a significant ifference between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. The ifference becomes greater with the varying number of source noes in a network for all compare routing protocols. 100 Receive packet variation (%) AODV DSDV DI-LSR FRP Number of source noes in network Figure 3-13: Receive packet variation vs. number of source noes Passive Source Noes vs. Number of Source noes AODV DSDV DI-LSR FRP Passive source noes (%) Number of source noes in network Figure 3-14: Passive source noes vs. number of source noes 64

84 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network The effect of a passive noe in network reflects on the fairness inex as well as the network performance in a multi-hop linear wireless network. The effect of varying number of source noes on the passive source noes is as shown in Figure With the implementation of DI-LSR, zero passive noes were achieve between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. This is only achievable in DI-LSR since the better proportion of ata packets transfer rate in a preefine ual interleaving path with low control packets. Where else the other routing protocols contribute towars passive noes with a varying number of source noes in a network. Normalise Routing Loa vs. Number of Source noes Network routing loa (packets) AODV DSDV DI-LSR FRP Number of source noes in network Figure 3-15: Network routing loa vs. number of source noes The effect of varying number of source noes increases the routing overhea in the network for all compare routing protocols as shown in Figure The routing overhea in a network with the implementation of DI-LSR is relatively lower when compare in terms of receive ata packets among the other routing protocols between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. With reuce control packets particularly with broacast packets reuce the traffic of supporting packets in a network especially with increasing number of source noes in a network. 65

85 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Energy Consumption vs. Number of Source noes AODV DSDV DI-LSR FRP Energy per-packet (Joules) Number of source noes in network Figure 3-16: Energy per-packet vs. number of source noes The energy consumption per-packet effect on varying number of source noes is as shown in Figure 3-16, which inicates DI-LSR has a consistent increase in energy to the network size. The energy consumption per-packet increasing factor in DI-LSR is ue to the amplification of ata packet at each varying simulation environment. Base on the number of ata packets receive, DI-LSR has a fair use of energy compare to all the other routing protocol from a low network ensity with 12 source noes to a larger network ensity with 120 source noes. Where else the other routing protocols consume higher energy per-packet with relatively low ata rate an throughput unfairness as compare to DI-LSR in the teste environment ue to higher routing loa. Energy efficient routing protocol ensures a sustainability factor in a typical linear wireless network, particularly when the noes are battery power epenent. Dual Cluster Hea Interleaving Linear Static Routing As a complementing routing technique to the DI-LSR an a flat one-tier topology, the Dual Cluster Hea Interleaving Linear Static Routing (DCHI-LSR) is esigne 66

86 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network to be implemente on a cluster-base architecture with high communication link reliability to enhance the overall network efficiency in a pipeline network [96]. A ata packet in DCHI-LSR is transferre from a source noe (sener) to a single estination noe (receiver) through preetermine intermeiate cluster heas in a network. The ual interleaving path is generate base on a preefine routing conition referring to o an even sequence of cluster heas in a multi-hop linear network. Wireless noes are statically positione in a fixe infrastructure of an oil an gas pipeline to form a communication chain between a sensing point an a monitoring station [4, 15]. Unlike a flat one-tier topology, the DCHI-LSR is a reliable an efficient routing algorithm, where a route to a estination noe is preefine over a cluster-base two-tier topology at simulation initialization perio. The routes between all source noes (sener) to a single estination noe (centralise monitoring station) are preefine base on o an even sequence of cluster hea with two interleaving paths [96]. Cluster heas in DCHI-LSR are uniformly istribute at a 1 istance within a maximum transmission range of two 1 in a multi-hop linear architecture with n number of sensing noes uniformly istribute at istance uner respective cluster hea. The funamental structure of DCHI-LSR is as shown in Figure 3-17, where sensing noes are place uner respective cluster hea which communicates ata on two inepenent interleaving paths between cluster heas (On/En) to a estination noe (ND). n hops n hops Preefine path for o sequence of cluster hea 1 hop 1 hop En N12 On 1 N11 N10 N9 N8 E1 1 N7 N6 N5 O1 1 N4 N3 N2 1 ND N1 Figure 3-17: DCHI-LSR on a cluster-base multi-hop linear topology The eploye sensors on a pipeline network, communicate ata through a eicate cluster hea an over multiple-hop intermeiate cluster heas in a 67

87 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network uniform interval (istance in meter). The DCHI-LSR can be escribe as a passive routing algorithm where changes in the network conitions are not esire throughout its operation. An infrastructure or static (fixe) base WSN requires only minimum or no changes in the route selection since most of the variable parameters which influence a routing protocol is at low active state [37, 96, 130]. The sensing noes an cluster heas in such a set-up are permanently locate at a uniform interval subjecte to the range of a wireless communication evice with a constant power source. Base on these factors, the esign of DCHI-LSR has two major moifications mae as in DI-LSR to accommoate the preefine routing path function an to reuce network routing overhea particularly with increasing number of noes in a cluster architecture [96] as compare to any conventional routing protocols [e.g. LEACH] [41, 69]. Noe placement in a multi-hop linear cluster-base topology often influences the behaviour of a WSN particularly on link stability issues between a source an a estination noe [119, 131]. Link instability results in increasing passive noes (noes with zero ata transfer) therefore ata transfer from that point onwars will be terminate. As a result, the overall network performance is compromise particularly in a highly scattere multi-hop linear network of noes. Passive noes are escribe as source noes with no opportunity to transmit ata or with a zero elivery ratio at a specific noe in a network. A network stability factor which relates to link connectivity an passive noes between source to a estination noe relies on the number active noes in a specific communication range. Thus, optimum placement of source noes uner respective cluster hea further improves the sensing capacity or coverage area without influencing the main ata stream towars a estination noe. In DCHI-LSR, a cluster hea is the main ata stream to a esignate estination noe unlike in a flat one-tier topology sensing an ata stream takes place on the same noe. With the propose interleaving technique among cluster heas in a multi-hop linear topology woul minimize the number of passive noes in a large WSN [96]. Generally, a routing protocol generates or upates its routing table base on the availability of noes in a specific transmission range at start time t or t + 1 when there are route changes among noes. Whereby for a multi-hop linear topology, a routing table is fully or 68

88 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network partially store in respective noes between a sener an a estination noe through a single routing chain buil with a sequence of noes in a network. Routing Table Generating Algorithm in DCHI-LSR The broacast an hello packets function is eliminate in DCHI-LSR, therefore the nee in physical search for neighbouring noes for optimum route search is not require. The routing table for DCHI-LSR is generate base on an ieal network environment with noes are always in a reay state hence reuces buffer time inbetween start to the first packet generate in a network. The routing table in DCHI-LSR is generate at time t with two preefine routing table from a sequence of o an even cluster heas with nn number of source noes eploye in a network as shown in Figure DCHI-LSR permits source noes to sen an receive ata packets along with the control packets only through a preefine sequence of o an even cluster hea in a network. There is four essential information neee in generating a forwar an reverse path routing table in DCHI-LSR which is the number of cluster heas an source noes in the network, path irection for ata communication an travel cost in hop count between source to a estination noe. The etaile routing process involves o an even sequence of cluster hea segregation, allocating nn number of source noes for a respective cluster hea, the irection of ata travel path an require hop count between respective source noes to the estination noe. The preefine forwar an reverse the path routing table generating process in DCHI-LSR is as briefly escribe in the pseuocoe of Figure Algorithm 2: Forwar an reverse path routing table Suppose number of source noes is nn, number of cluster hea is ch, o sequence of cluster hea is CHOn, even sequence of cluster hea is CHEn an estination noe is ND = 0 At time t: 1: IF ch%2 == 0 then: Forwar path routing table 2 (for even sequence of cluster hea) 69

89 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network else: Forwar path routing table 1 (for o sequence of cluster hea) 2: For nn then: Respective source noes are assigne to forwar path routing table 1 an table 2 3: For ch an nn then: Reverse path routing table (to respective source noes base on o an even sequence of cluster hea) 4: For CHOn/CHEn-2 0 then: Neighbour cluster hea at forwar path else IF CHOn -1=0 then: Neighbour cluster hea at the forwar path 5: For CHOn/CHEn +2 nn then: Neighbour cluster hea at reverse path 6: For ((CHOn +1)/2) +1 then: Require hop/hops to estination noe for forwar/reverse path (o) 7: For (CHEn/2) +1 then: Require hop/hops to estination noe for forwar/reverse path (even) 8: At time t+1: Source noes nn start sening ata packets to ND 9: At time t+n: ND start sening TCP acknowlegement packets to source noes nn Figure 3-18: Routing table generating process in DCHI-LSR The segregation of o an even sequence of cluster heas in a respective forwar path routing table entries is create base on the ch number of cluster hea in a network. Once the segregation process is establishe, nn number of source noes are assigne to a respective cluster hea (e.g. three source noes are place uner a cluster hea as in Figure 3-17). Source noes in DCHI-LSR are only permitte to communicate with a series of other source noes an cluster heas base on the generate routing table. Once the routing table entries for o an even sequence of cluster heas are create, forwar (to a estination noe) an reverse (to all source noes) path is preefine with available cluster heas in a respective routing table. The forwar path routing table in DCHI-LSR is as briefly escribe with an even sequence of cluster heas in Figure 3-19 an o sequence of cluster 70

90 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network heas in Figure In a forwar path routing table, a respective source noe hols the neee entries of cluster heas base on the segregation process to a esignate estination noe. A source noe is only permitte to communicate to its irect cluster hea or leaer noe as in any cluster base routing at all time in a network [41, 61, 71]. n hops 1 hop Preefine path for even sequence of cluster hea Nxt Dst HC E1 ND n E1 E1 n-1 E1 N4 (n-1)+1 E1 N5 (n-1)+1 E1 N6 (n-1)+1 Nxt Dst HC ND ND 1 N4 N4 1 N5 N5 1 N6 N6 1 En On E1 O1 ND N12 N11 N10 N9 N8 N7 N6 N5 N4 N3 N2 N1 Nxt Dst HC En ND 1+n En En 1 En E1 1+(n-1) En N4 1+(n-1)+1 En N5 1+(n-1)+1 En N6 1+(n-1)+1 Nxt Dst HC En ND 1+n En En 1 En E1 1+(n-1) En N4 1+(n-1)+1 En N5 1+(n-1)+1 En N6 1+(n-1)+1 Nxt Dst HC Nxt Dst HC E1 ND 1+1 E1 ND 1+1 E1 E1 1 E1 E1 1 Nxt Dst HC E1 ND 1+1 E1 E1 1 Nxt Dst HC En ND 1+n En En 1 En E1 1+(n-1) En N4 1+(n-1)+1 En N5 1+(n-1)+1 En N6 1+(n-1)+1 Figure 3-19: Forwar path with even sequence of cluster heas in DCHI-LSR Base on Figure 3-19 an Figure 3-20, the ata flow is inicate at respective source noes with the path irection referring to a cluster hea an neighbouring cluster heas (Nxt), the esignate estination point (Dst) which is ieally the estination noe (ND) or any other cluster heas an source noes in the network. A point to point travel cost in hop count (HC) is inicate for a respective source noe base on a minimum hops require to a esignate estination. This mechanism ensures that there will be zero ata packets floating without irection at a certain point in the network [96]. 71

91 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network 1 hop n hops Preefine path for o sequence of cluster hea Nxt Dst HC Nxt Dst HC O1 ND n ND ND 1 O1 O1 n-1 N1 N1 1 O1 N1 (n-1)+1 N2 N2 1 O1 N2 (n-1)+1 N3 N3 1 O1 N3 (n-1)+1 En N12 N11 On 1 N10 N9 N8 E1 1 N7 N6 N5 O1 1 N4 N3 1 N2 ND N1 Nxt Dst HC Nxt Dst HC Nxt Dst HC Nxt Dst HC On ND 1+n On ND 1+n O1 ND 1+1 O1 ND 1+1 On On 1 On On 1 O1 O1 O1 O1 On O1 1+(n-1) On O1 1+(n-1) On N1 1+(n-1)+1 On N1 1+(n-1)+1 On N2 1+(n-1)+1 On N2 1+(n-1)+1 On N3 1+(n-1)+1 On N3 1+(n-1)+1 Nxt Dst HC On ND 1+n On On 1 On O1 1+(n-1) 1 1 Nxt Dst HC O1 ND 1+1 O1 O1 1 On N1 1+(n-1)+1 On N2 1+(n-1)+1 On N3 1+(n-1)+1 Figure 3-20: Forwar path with o sequence of cluster heas in DCHI-LSR In a reverse path routing table, the estination noe hols the entries for o an even sequence of cluster heas an source noes in the network. Where else a respective cluster hea hols the neee entries of other cluster heas an its source noes (member noes uner a respective cluster hea) in the network with preefine sequence from the segregation process. The istance between a source noe to a estination noe is calculate in terms of packet travel cost in hop count. In DCHI-LSR, the hop count is always maintaine at minimum travel cost between a source an a estination noe in orer to optimise the network resources wisely. The reverse path routing table in DCHI-LSR is as briefly escribe in Figure

92 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network n hops 1 hop n hops Preefine path for o sequence of cluster hea Nxt Dst HC Nxt Dst HC Nxt Dst HC Nxt Dst HC N10 N10 n+1 N7 N7 n+1 En N11 N11 n+1 N8 N8 n+1 En N10 (n-1)+1 On N7 (n-1)+1 N12 N12 n+1 N9 N9 n+1 En N11 (n-1)+1 On N8 (n-1)+1 En N12 (n-1)+1 On N9 (n-1)+1 En N12 N11 On 1 N10 N9 Nxt Dst HC 1 hop N8 En n-1 E1 1 N7 N6 Nxt Dst Nxt Dst On On n-1 O1 1 N5 N4 N3 N2 HC Nxt Dst HC E1 E1 1 O1 O1 1 E1 En 1+(n-1) O1 On 1+(n-1) E1 N10 1+(n-1)+1 O1 N7 1+(n-1)+1 E1 N11 1+(n-1)+1 O1 N8 1+(n-1)+1 E1 N12 1+(n-1)+1 O1 N9 1+(n-1)+1 HC Nxt Dst ND 1 N1 HC E1 E1 1 E1 E1 1 E1 En 1+(n-1) E1 En 1+(n-1) E1 N10 1+(n-1)+1 E1 N10 1+(n-1)+1 E1 N11 1+(n-1)+1 E1 N11 1+(n-1)+1 E1 N12 1+(n-1)+1 E1 N12 1+(n-1)+1 O1 O1 O1 N O1 N2 1+1 O1 N3 1+1 E1 E1 E1 N E1 N5 1+1 E1 N6 1+1 O1 On O1 N7 n+1 O1 N8 n+1 O1 N9 n+1 E1 En n n E1 N10 n+1 E1 N11 n+1 E1 N12 n+1 Nxt Dst HC O1 O1 1 O1 On 1+(n-1) O1 N7 1+(n-1)+1 O1 N8 1+(n-1)+1 O1 N9 1+(n-1)+1 Nxt Dst HC O1 O1 1 O1 On 1+(n-1) O1 N7 1+(n-1)+1 O1 N8 1+(n-1)+1 O1 N9 1+(n-1)+1 Figure 3-21: Reverse path in DCHI-LSR Source Noe Routing Flows in DCHI-LSR In a large WSN, queue limitation is a stanar mechanism for controlling an managing bi-irectional packets among noes in a network [103, 122, 125]. Aequate queue limitation can eliminate passive noes an bottleneck points in a single path wireless network. In general, a queue limitation increases the rate of packet roppe especially in a network with the uniscovere route as in any stanar routing protocol [28]. With the implementation of DCHI-LSR, this factor coul be minimise thus further increase the rate of ata packet transfer with the propose ual interleaving path metho. Splitting a single path as in a conventional routing protocol into two paths, further, reuces the routing overhea by half hence allocates better proportion for ata packets in a respective path as escribe in Figure A efault queue limit [122, 125] is set at each intermeiate cluster heas in both paths as escribe in the flow chart in Figure Referring to the flowchart in Figure 3-22, a generate ata packet from a sensing point will be forware from a source noe at level one to a cluster hea 73

93 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network at level two to a respective cluster hea in the network. The ata packet than will be forware to the esignate estination noe through a preefine sequence of cluster heas base on the routing table entries. Start: Data packet Start Simulation Routing table an ata flow process at the source noes O sequence of cluster hea *Generating Preefine source noe routing table Even sequence of cluster hea Data packets sen request from source noe Data packets sen request from source noe Data flow process at intermeiate cluster heas Repetitive process for ch number of cluster heas Yes Is the biirectional packets Queue size? No No Is the biirectional packets Queue size? Yes Data packets in cluster hea CHOn queue Data packets roppe at intermeiate cluster hea ue to queue overflow Data packets in cluster hea CHEn queue Data flow process at estination noe No Is the biirectional packets Queue size? Yes Data packets roppe at estination noe ue to queue overflow Data packets in queue at estination noe En: Data packets Data packets receive at estination noe * Refer to pseuucoe algorithm 2 A Figure 3-22: Flow chart for a ata packet cycle in DCHI-LSR 74

94 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network With the restricte queue limit, any ata packets arriving at an intermeiate (neighbouring) cluster hea with queue length > IfQlenCHOn or IfQlenCHEn will be roppe from moving forwar towars the estination noe. This is commonly known as queue overflow where the roppe packets will be regenerate after a specific uration [24, 77, 78]. In a WSN, roppe packets are not favourable ue to unwante loss of network resources particularly on the energy consume from packet generation till packet roppe. In a WSN, queue overflow is ue to overwhelming of control packets which inclue broacast an hello packets epening on the characteristic of a routing protocol [3, 132]. The queue factor remains same at each intermeiate cluster heas till the ata packet reaches its estination noe. The reuce routing overhea in DCHI-LSR creates less traffic, thus packet rops can be reuce compare to a routing protocol which relies on this broacast an signalling packets to generate or maintain a routing table. The concept of istributing traffic in two paths further reuces the routing overhea by half, therefore, allocates a better proportion of ata packets compare to any conventional single-path routing protocol. The concept of a ata accumulation factor for o an even sequence of cluster heas can be escribe respectively in equation 3-13 an 3-14 between a source an intermeiate noes towars a estination noe. = Where TPCHO is the total packets in an o sequence of cluster hea, CPOi is the total control packets for a ch number of cluster heas (o), DPj is the total ata packets an CPj is the total control packets for nn number of source noes uner a cluster hea with IfQlenCHOn as the queue length in o routing path. = Where TPCHE is the total packets in an even sequence of cluster hea, CPEi is the total control packets for a ch number of cluster heas (even), DPj is the total ata packets an CPj is the total control packets for nn number of source noes uner a cluster hea with IfQlenCHEn as the queue length even routing path. A estination noe is place at en of each interleaving route with an integration point as inicate in the flow chart of 75

95 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Figure A ata packet forware from a source noe through a series of cluster heas will be irecte to this point at all time. The successful ata packets accumulation factor at the estination noe (receiver) from both routes is as escribe in equation Any incoming ata packets at the estination noe with queue length > IfQlen will be roppe. = Where NTP is the network total packets at the estination for a ch number of cluster heas an the nn number of source noes with TPCHO or TPCHE is from equation 3-13 an Destination Noe Routing Flows in DCHI-LSR A estination noe routing table hols entries of all source noes eploye in a network base on two preefine reverse paths. At time t, DCHI-LSR generates a reverse path routing table in which will be use by control packets to reach a esignate source noe. Once a ata packet has successfully reache the estination noe, a TCP acknowlegement packet will be generate in return to acknowlege the respective source noe (sener). A generate TCP acknowlegement packet will be forware from a estination noe at level one to a series of respective cluster heas base on o an even path at level two. The TCP acknowlegement packet is forware to a respective source noe through a preefine sequence of cluster heas base on the routing table entries. Unlike a forwar route as escribe in Figure 3-18, the reverse path routing table at the estination noe has the entire list of both o an even sequence of cluster heas an source noes. The preefine reverse path routing table generating process in DCHI-LSR is as briefly escribe in the pseuocoe of Figure The hop count is always retaine at minimum travel cost between a estination to respective source noes in orer to optimise the network resource wisely. The reverse path hop count is base on the same hop count in a forwar path between two respective noes in the network. At time t+2, a generate TCP acknowlegement packet is forware on a preefine return path to a esignate source noe with a efault queue limit [122, 125] is set at each intermeiate cluster heas in both routes as escribe in the flow chart in 76

96 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Figure Since the routing path is preefine with no changes is expecte uring the noe active perio, no aggressive broacast packets are require to maintain neither the routing table nor the routing path in DCHI-LSR. Performance Evaluation on Linear Cluster Two-tier Topology The overall network performance is teste in various simulation environment create in NS2 [34, 103] to visualise the behaviour an impening factors when full eployment of DCHI-LSR in a pipeline network. In all simulate environment, the propose DCHI-LSR is compare with DI-LSR in a flat one-tier topology as a reference, AODV a reactive an DSDV a proactive routing protocol in a clusterbase two-tier topology for performance comparison in NS2. The basic configuring an preefine setting for the simulation environment is as tabulate in Table 3-2. Table 3-2: NS2 simulation parameters for cluster-base topology Channel type Key parameters Raio propagation moel Wireless channel Two Ray Groun Set value MAC type Interface queue type (ifq) Drop Tail/PriQueue Source noes (SN) 12, 24, 36, 48, 60, 72, 84, 96, 108, 120 Destination noe (ND) 1 Cluster hea (CH) 4, 8, 12, 16, 20, 24, 28, 32, 36, 40 Queue length (ifqlen) Agent type Traffic type Transmission range (RX Thresh) Sensing range (CS Thresh) Packet size 50 (packets) Transmission control protocol (TCP) Constant bit rate (CBR) 350 meter 550 meter 512 bytes 77

97 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network In each simulation environment, an average value of five runs per cycle for the manipulating variable (ch number of cluster heas an the nn number of source noes) with ifferent see values in between 1 to 10 over a simulation uration of 500 secons. The ata rate of one packet/sec were assigne to all source noes with other corresponing parameters as given in Table 3-2. In a cluster-base two-tier topology, the number of cluster hea is uniformly increase by 4 cluster heas an the number of source noes is uniformly increase by 12 source noes in each cycle. In a hierarchical topology, the number of cluster heas in each cycle is uniformly increase by 4 cluster hea starting from 4 cluster heas to 40 cluster heas with an increasing number of member noes from 12 source noes to 120 source noes. The simulation environment was create using a custom ranom to test the efficiency an robustness of the propose routing algorithm in a near realtime application where the selection of non-chronological sening orer an start time for source noes were ranomly chosen for each simulation cycle. The ata packet transmission is assigne ranomly between 0 2 secons at the beginning of the simulation. All noes mentione in the simulation are statically place uring the simulation uration with one estination noe (ND). Simulation Results an Discussion for Cluster Two-tier Linear Topology The overall network performance of DCHI-LSR is compare with AODV an DSDV in a cluster-base two-tier topology that is presente as AODVC an DSDVC with a reference guie from DI-LSR for flat one-tier topology in terms of the performance metrics in section 3.5. Packet Delivery Ratio vs. Number of Source noes The effect of varying number of source noes on the elivery ratio is as shown in Figure Generally, the elivery ratio ecreases with the increasing number of source noes for all compare routing protocols. It is observe that at low network ensity with 12 source noes, the elivery ratio is almost at the same rate among all compare routing protocols. The elivery ratio of DCHI-LSR is at a moerate rate when compare with the other cluster-base routing protocol between a low network ensity with 12 source noes to a larger network ensity 78

98 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network with 120 source noes. There is a significant ifference in elivery ratio, particularly between DI-LSR an DCHI-LSR. The unappealing rate of the elivery ratio in DCHI-LSR is ue to a higher ata transfer rate, competitive ata transfer an queuing factor at each cluster heas is increase by 3 member noes (source noes) that can be further improve with a network fairness mechanism AODVC DSDVC DCHI-LSR DI-LSR Delivery ratio (%) Number of source noes in network Figure 3-23: Delivery ratio vs. number of source noes En-to-en Delay vs. Number of Source noes The result comparison on average en-to-en elay for varying number of source noes is as shown in Figure 3-24, which inicates DCHI-LSR has a graual increase in elay proportional from a low network ensity with 12 source noes to a larger network ensity with 120 source noes. Higher ata rate has irect implication in terms of overall elay typically ue to the total ata packets receive an the fairness factor among source noes in the network. The elay in DCHI-LSR is fairly low compare to the moerate rate of fairness an higher ata rate achieve when compare to the other routing protocols. In a multi-hop linear topology, the elay factor varies from the istance between a source an a estination noe. Using a preefine ual interleaving path, the DCHI-LSR is capable of forwaring ata in a more efficient manner using a eicate ata stream as compare to a single one-tier multi-hop metho. 79

99 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network AODV-C DSDV-C DCHI-LSR DI-LSR En-to-en elay (ms) Number of source noes in network Figure 3-24: En-to-en elay vs. number of source noes Throughput vs. Number of Source noes Throughput (Kbps) AODV-C DSDV-C DCHI-LSR DI-LSR Number of source noes in network Figure 3-25: Throughput vs. number of source noes The effect of varying number of source noes increases the throughput in the network for all compare routing protocols as shown in Figure The throughput in a network with the implementation of DCHI-LSR outperforms all the other routing protocols from a low network ensity with 12 source noes to a 80

100 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network larger network ensity with 120 source noes. With the implementation of ual interleaving cluster hea splits the traffic into two paths, therefore a significant increase in the throughput value between 23 Kbps 42 Kbps can be achieve at varying stages compare with the other routing protocols. With the implementation of ual interleaving cluster hea in a eicate ata stream path in DCHI-LSR has the capacity to achieve higher throughput within the available network resources. Throughput Fairness Inex vs. Number of Source noes AODV-C DCHI-LSR DSDV-C DI-LSR Throughput fairness inex Number of source noes in network Figure 3-26: Throughput fairness inex vs. number of source noes Network imbalance is a crucial factor in any multi-hop linear wireless network. Referring to Figure 3-26, the throughput fairness inex of DCHI-LSR outperforms all the other cluster-base routing protocols with a significant ifference in fairness inex from a low network ensity with 12 source noes to a larger network ensity with 120 source noes. In a low network ensity with 12 noes, fairness is harly visible since the performance of all other routing protocol is fairly similar. When compare with DI-LSR, there is a merely a small ifference in fairness inex ue to a greater network capacity an competitive ata transfer in DCHI-LSR. A higher ata rate an moerate throughput fairness are comparatively 81

101 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network achievable in a multi-hop linear cluster-base wireless network with the propose DCHI-LSR. Receive Data Packet Variation vs. Number of Source noes Variation of receive ata packets is another parameter in visualising fairness between source noes in a network. Referring to Figure 3-27, the percentage of receive ata packet variation of DCHI-LSR is relatively similar to all the other routing protocol from a meium network ensity with 36 source noes to a larger network ensity with 120 source noes. The ifference becomes greater with the varying number of source noes an comparatively with increasing receive ata packets at a estination noe for all compare routing protocols. 100 Receive packet variation (%) AODV-C DSDV-C DCHI-LSR DI-LSR Number of source noes in network Figure 3-27: Receive ata packet variation vs. number of source noes Passive Source Noes vs. Number of Source noes The effect of passive noes in network reflects on the fairness inex as well as the network performance in a multi-hop linear cluster-base wireless network. The effect of varying number of source noes on the passive source noes is as shown in Figure 3-28 where zero passive noes were achieve between a low network ensity with 12 source noes to a larger network ensity with 120 source noes with DCHI-LSR. With a better proportion of ata packets transfer rate in a 82

102 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network preefine ual interleaving path with low routing overheas as in DI-LSR eliminates the passive noes. While the other routing protocols contribute towars passive noes with a varying number of source noes in a network. 60 Passive source noes (%) AODV-C DSDV-C DCHI-LSR DCI-LSR Number of source noes in network Figure 3-28: Passive source noes vs. number of source noes Normalise Routing Loa vs. Number of Source noes The effect of varying number of source noes increases the network routing overhea in the network for all compare routing protocols as shown in Figure The network routing overhea in a network with the implementation of DCHI- LSR is relatively lower when compare in terms of receive ata packets among the other routing protocols between a low network ensity with 12 source noes to a larger network ensity with 120 source noes. When compare against DI- LSR, the routing overhea is much lower after 60 source noes with the implementation of DCHI-LSR. With reuce control packets particularly with broacast packets reuce the traffic of supporting packets in a network especially with increasing number of source noes. 83

103 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network Network routing loa (packets) AODV-C DSDV-C DCHI-LSR DI-LSR Number of source noes in network Figure 3-29: Network routing loa vs. number of source noes Energy Consumption vs. Number of Source noes AODV-C DSDV-C DCHI-LSR DI-LSR Energy per-packet (Joules) Number of source noes in network Figure 3-30: Energy per-packet vs. number of source noes The energy consumption per-packet effect on varying number of source noes is as shown in Figure 3-30, which inicates that DCHI-LSR has a graual increase in energy to the network size. The energy consumption per-packet increasing factor in DCHI-LSR is ue to the amplification of ata packet at each varying simulation 84

104 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network environment. Base on the number of ata packets receive, DCHI-LSR has relatively lower energy consumption compare to all the other routing protocol from a low network ensity with 12 source noes to a larger network ensity with 120 source noes. The other routing protocols consume higher energy perpacket with relatively low ata rate compare to DCHI-LSR in the teste environment ue to higher routing loa. An energy efficient routing protocol ensures a sustainability factor in a typical multi-hop linear wireless network, particularly when the noes are battery power epenent. Conclusion This chapter mainly presente two routing algorithms which were implemente on a flat (one-tier) an a cluster-base (two-tier) multi-hop linear topology. The outcome from implementing DI-LSR an DCHI-LSR in respective wireless topologies has highlighte potential optimisation factors towars the overall network performance on a pipeline network. Both the propose routing algorithm features reuce network routing loa with a preefine ual interleaving routing path to further enhance the overall network performance as well as to eliminate issues with passive noes. The basic comparison between DI-LSR an DCHI-LSR is as shown in Table 3-3. Table 3-3: Brief comparison of DI-LSR an DCHI-LSR Features DI-LSR DCHI-LSR Architecture One-tier linear (sensing noes acts as intermeiate noes) Two-tier linear cluster (sensing noes, leaer noe acts as intermeiate cluster heas) Sensing point an ata stream Network complexity Noe failures Dual interleaving sensing an ata stream (share) Low (one-tier)- fewer noes require Sensing noes effects ata stream path (either o or even) Dual interleaving cluster heas as ata stream an nn number of sensing points uner each leaer noe High (two-tier)- more noes require Sensing noes oes not affect ata stream path but cluster heas effects ata stream path (either o or even) 85

105 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network The simulate results have emonstrate that both DI-LSR an DCHI-LSR has achieve a significant level of improvements in reliability (elivery ratio), network capacity (throughput), latency (en-to-en elay) an responsiveness (ealing with noe failures) which ha crucial implication towars the overall network performance on a pipeline network. The overall network performance of DI-LSR an DCHI-LSR in average from varying simulation environment is as tabulate in Table 3-4. Table 3-4: Overall network performance of DI-LSR an DCHI-LSR Measure parameters DI-LSR DCHI-LSR Average elivery ratio (%) Average en-to-en elay (ms) Average throughput (Kbps) Average throughput fairness inex Average receive packet variation (%) Average passive noes (%) 0 0 Average network routing loa (packets) Average energy per-packet (mili joules) A multi-hop linear topology has a unique geographical architecture that results in limite network resources (in a single path) an high risk of communication failure (epens on the number of noes within a communication range), especially in a large WSN. Similarly, in a pipeline network with the propose technique, noes are arrange to establish communication with multiple noes in the network that incurs aitional cost as compare to other wireless topologies to accommoate minimum three noes within the communication range to reuce the risk of noe failures. The results analysis outcome has inicate that the impening network factors on reliability, fairness issues in network resources an energy consumption nee attention in a multi-hop linear topology. In the next Chapter, a scheuling scheme for TCP elaye acknowlegement to be incorporate with both DI-LSR an DCHI-LSR for better network equality in a 86

106 Reliable an Efficient Interleaving Linear Static Routing for Pipeline Network multi-hop linear network will be presente. A scheuling scheme for TCP elaye acknowlegement can further optimise the fairness in network resources sharing in a multi-hop linear network. 87

107 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Chapter 4 4. Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Introuction Over the years, remote monitoring is establishe with various wireless stanars an communication technologies in relaying real-time ata to a centralise monitoring station from sensors attache to critical pipeline points [4, 8, 14]. In a static geographical infrastructure of an oil an gas pipeline usually stretche over hunres of miles between the sensing points (source noes) to a centralise monitoring station (ata collection for future analysis) [6, 9]. The network of sensors which are linke in series to form a communication chain to sense changes on sensing points an forwar this information to a estination noe with preefine route etermine by a routing protocol. Figure 4-1 illustrates the ata packets travel time factor at time t an ata rate for a conventional multi-hop linear topology with a Nn number of source noes an a single estination noe (ND). Flow Nn Flow N5 Flow N4 Flow N3 Flow N2 Flow N1 nrtt/t 5RTT/t 3RTT/t 2RTT/t Communication range of Nn Nn n hops N5 5 hops N4 4 hops N3 3 hops N2 2 hops N1 ND 1 hop Figure 4-1: Data travel rate in multi-hop linear topology Respective noes are inicate inflows, Nn where n is the source noe number or sequence in a network which represents bi-irectional packets moving towars a estination noe. The ata packet travel time factor is escribe with Roun 88

108 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Trip Time (RTT) for respective source noes in a wireless network. The RTT value is efine as one cycle time for a ata packet to travel from a source to a estination noe an receives an acknowlegement packet to confirm its recipient of the sent ata packet [23, 28, 77]. Such a pipeline network has unique unerlying issues affecting full eployment of WSN on oil an gas pipelines in terms of egraing factor on overall network performance as well as network fairness among source noes [95, 133]. In a typical multi-hop linear WSN, the long range ata transmission [134] has severe performance egraation issues that lea towars inefficient fairness in the wireless network particularly using the Transmission Control Protocol (TCP) [23, 28, 135]. Figure 4-2 illustrates the ata packet travel time factor with fairness implication on a conventional multi-hop linear topology at time t+1. The typical characteristic of a multi-hop linear topology in such setup results in unbalance ata transfer rate between source noes in a certain network [24, 27]. When the effect is amplifie at time t+2, the network will suffer from severe source noe starvation particularly noes that are locate in a istance away from the estination noe. In any multi-hop linear WSN, this is a crucial issue that leas towars ata packet loss an a waste of network resources. nrtt/t Moerate ata rate 5RTT/t 4RTT/t 3RTT/t Flow Nn Flow N5 Flow N4 Flow N3 Communication range of Nn Nn Flow N2 Flow N1 N5 N4 N3 N2 N1 ND n hops 5 hops 4 hops 3 hops 2 hops 1 hop Figure 4-2: Effect on ata travel rate in multi-hop linear topology In the recent years, continuous research on new protocols, applications, techniques an evices have been stuie in TCP for achieving optimum fairness outcomes. A network with TCP agent experience performance issues where the 89

109 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization root cause is from noe starvation an unbalance ata packet flow, particularly in a long range linear topology [27, 135, 136]. In terms of network fairness or equality among source noes are rather ifficult to be efine in a single fairness metric since fairness is a subjective factor when reflecte against performance metrics. In practical terms an rational consieration, each implementation of fairness moel has a unique characteristic in actual eployment varies between the nature of its application, the routing protocol use an in the implemente topology [24, 28, 135, 137]. Therefore sharing of network resources can be evaluate by the optimisation rate of available network resources among all source noes with a constant rate in longterm network fairness. The actual efinition of network fairness in a bigger picture is base on the attributes neee in a specific eployment either by the network user or by a respective inustry. Thus, in most research on network fairness attributes towars a esire rate on the share network resource. In a practical worl, a network fairness moel comes with special attributes towars the requirements an cost factor incorporate with the esire outcome. This chapter mainly focuses on four mathematical formulations of TCP elaye acknowlegement timeout moel for a flat (one-tier) an cluster-base (two-tier) multi-hop linear topology. The TCP elaye acknowlegement timeout moel for a Fairness critical application (DATM-FFCA) an with a Throughput critical application (DATM-FTCA) [28] is propose for a flat topology. While the TCP elaye acknowlegement timeout moel for a Fairness critical application (DATM-CFCA) an with a Throughput critical application (DATM-CTCA) is propose for a cluster-base topology. Both moels are formulate for TCP acknowlegement timeout to accommoate the increasing ata traffic with optimum throughput fairness in a multi-hop linear WSN with zero passive noes. Network Fairness Factor in Multi-hop Linear Topology Network fairness or equality among source noes (seners) in a TCP agent is a highly esirable metric in a multi-hop linear topology. The effect of network fairness over varying network ensity results in severely egraing throughput an increasing passive noes [RW.ERROR - Unable to fin reference:214]. 90

110 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Network unfairness is a conition in which affects the overall throughput an network resources equality factor in a multiple-to-one inline traffic pattern that is relatively common in a multi-hop linear WSN. This conition is quite common when multiple source noes in a varying istance sen ata packets to the same estination noe (receiver) as shown in Figure 4-1. The communication link between the source an the estination noe becomes a bottleneck point, thus resulting in heavy loss of ata packets resulting in low network throughput [78, 138]. Generally, there is three foremost network fairness factor which significantly contributes towars; (1) the network throughput, (2) en-to-en elay an (3) energy wastage which is not practical for eployment in a pipeline network [24, 26, 139]. By clearly unerstaning this factors woul help to recognise methos to overcome effects of network fairness with a TCP elaye acknowlegement timeout moel. In the following section, these factors an relate parameters are efine with a feasible solution for optimum fairness rate in a pipeline network. Source Noe Distance in a Multi-hop Linear Topology The unconitional sharing of network resources among source noes eploye in a linear architecture may not mean that there is an equal allocation of the available network resources. It is quite common that each source noe is assigne with an equal level of the available network capacity. However, in a multi-hop linear topology with varying istance contributes towars unfair access to network resources [24, 27, 136]. On the other han, the concept of sharing in a multiple-toone inline traffic pattern in a single path with overwhelming ata packets contributes to ata accumulation factor towars the estination noe [28, 140]. Referring to Figure 4-2, in an ieal scenario where ata packet travel time factor varies from each sensing points between N1 to Nn in a conventional multihop linear topology contributes towars resultant fairness at time t+2. The conception of istance between source to a estination noe is proportional to the ata packet travel time cost which is a clear inication of unfairness in a multi-hop linear topology. The further a source noe is locate, a greater travel time is require for a ata packet to reach its estination 91

111 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization noe. An example of ata travel time factor base on Figure 4-2 is tabulate in Table 4-1 with a simulation time, t is 1 secon. Table 4-1: Example of ata travel time for a multi-hop linear topology Source noe Data travel time Travel time Packets per secon N1 RTT 14.1 ms 70 N2 2RTT 28.2 ms 35 N3 3RTT 42.3 ms 23 N4 4RTT 56.4 ms 17 N5 5RTT 70.5 ms 14 Nn nrtt n ms n Furthermore, the consieration of unfairness network allocation at some point strive towars source noe starvation that mostly affects those noes locate a istance away from the estination noe [26, 118]. This conition is also known as silent or passive noes where the source noe will constantly be generating ata packets without allocation towars the estination noe is an extremely waste of network resources in a WSN. In an aequate WSN, unwante waste of network resources both energy an banwith is not a goo sign in terms of fairness as well as the network performance on a pipeline network. Long-term Network Fairness an Data Frequency Consiering on a perio of time, network fairness measures the available network resources allocation at the en of a cycle time or over a longer usage perio. Longterm fairness has a crucial factor when a network resource is infrequent or in an uncontrolle environment [23, 24]. On the other han, ata frequency is an influencing element for overwhelming ata packet factor which is challenging for a guarantee network resources allocation among all varying source noes in a multi-hop linear network [26, 28, 92]. In a typical environment where equal network resources allocation is given at time t+1 to all source noes in a network 92

112 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization oesn t significantly affect network fairness in terms of network capacity (throughput). But at time t+2, where the amplification of ata packets moving towars a estination noe implying that unfairness allocation among source noes creates variation between iniviual capacity in terms of fairness rather than the network capacity as shown in Figure 4-2 [28]. The challenging network environment where aaptation of network changes relies on many varying parameters especially on the characteristic of an implemente routing protocol an how efficient is the rate of long-term network fairness. The concept of iniviual fairness or the overall network fairness is always relate towars longterm network fairness in a multi-hop linear topology [141]. In terms of iniviual fairness, a respective source noe is measure whether was it treate fairly in the network rather than the overall network fairness that is impossible to achieve in a network with varying istance. Therefore, when a certain network is able to achieve optimum network fairness an a certain level of performance is consiere a network with fair allocate resources. Base on this fairness characteristic, network fairness can be emphasise as a subjective metric where ifferent applications an users esire a unique resources allocation rate [28]. Degraing of Wireless Performance Network fairness is often referre as a wireless metric that is quite common in a static architecture with a fixe set of source an estination noes sharing network resources mainly for ata packet transfer [92, 142]. In a practical implementation, the number of ata flows or connections are highly ynamic. The unplanne varying conition of increasing an ecreasing ata arrival occurs between source to a estination noe till the entire process is complete. In the sharing process where fairness is the prime concern in a network, the network performance will be a traeoff metric or will be compromise in the long-run. Network fairness consiers equal resources allocation isregars to the location of source noes from a estination noe [27]. Thus this scenario creates an unfair state for noes in a closer istance compare to a noe further away from a estination noe. In a typical linear arrangement of source noes, as shown in Figure 4-1 where the performance of iniviual noes is normalise with an optimum fairness rather than network capacity which results in egraing of 93

113 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization throughput, increasing en-to-en elay an higher consume network energy [1, 95]. The performance egraing factor in a multi-hop linear network is proportional to the ensity of noes with network resources that are equally allocate causing increase waiting time between ata transmission [27]. On the other han, performance egraing factor is often relate to the characteristic of a routing protocol implemente in a WSN. Unfortunately, relying on to routing protocols are inaequate in achieving both fairness an performance in a multihop linear topology. Without a proper ata flow controlling mechanism in routing protocols woul result in severe egraing of performance, unfairness an source noe starvation [26, 92]. The Delaye Acknowlegement Timeout Moel for Flat Onetier an Cluster-base Two-tier Multi-hop Linear Topology In a multi-hop linear topology, network fairness can be achieve base on the scale an complexity of an application establishe on a one-tier or multi-tier approach. A fair allocation of network resources can be escribe as an outcome of a process in a WSN where respective source noes have equal opportunity in the long run. In a non-ynamic WSN such as in a pipeline network, temporal changes in resources allocation are uncommon as the system is in a control state especially with a static routing algorithm as escribe in Chapter 3 [15, 96]. Therefore, targete an resultant network fairness coul be retaine at a set level for the entire network active uration. On the other han, network fairness can also be consiere from a perspective of both overall network an iniviual source noes base on the en user s implementation [28]. Moreover, in a multi-hop linear topology where a respective source noe is esigne to perform a specific sensing task an also as an intermeiate router between another source noe to the estination noe in a network. Therefore, from the perspective of a pipeline network, the overall network fairness can be efine from the network activity in whole rather than an iniviual task. Referring to these network fairness characteristic an nature of the application, a mathematical formulation to calculate optimum TCP elaye acknowlegement timeout is propose with the best implementation of the propose routing algorithms from Chapter 3 over a pipeline network scenario 94

114 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization shown in Figure 4-1. In this section, there are four TCP elaye acknowlegement time out moels are propose with two moels to be implemente on a flat (onetier) an a cluster-base (multi-tier) multi-hop linear topology respectively. The first propose moel; TCP elaye acknowlegement timeout moel for a Fairness critical application (DATM-FFCA) an the secon propose moel; TCP elaye acknowlegement timeout moel for a Throughput critical application (DATM-FTCA) is establishe for a flat one-tier multi-hop linear architecture [28]. The propose DATM-FFCA an DATM-FTCA is a mathematical formulation for optimum fairness with variation in the overall network performance. The thir propose moel; TCP elaye acknowlegment timeout moel for a Fairness critical application (DATM-CFCA) an the fourth propose moel; TCP elaye acknowlegment timeout moel for a Throughput critical application (DATM-CTCA) is establishe for a multi-hop linear cluster-base topology. The propose moel is best to be implemente with the ual interleaving technique among cluster heas (leaer noe an intermeiate noes) to a estination noe [96]. The propose DATM-CFCA an DATM-CTCA is a mathematical formulation for optimum fairness with variation in the overall network performance. All four propose moels woul be a practical an in complex solution in an application where the esire normalize throughput fairness goal is set at a ratio of one among all source noes in the network. The practical implementation of the propose moels requires some minor moification in the efault settings of TCP parameters to ensure optimum network fairness outcome. To achieve the network fairness ratio of one among all source noes, the source noe avertise winow (awn) an the congestion winow (cwn) is set to the lowest value to restrict packet flows at a specific time interval. In general, the effect of controlling the avertise winow at a estination noe can avoi or further reuces the rate of ata packet overflow over the available network resources at the estination noe [23, 88]. Whereas the effect of controlling the congestion winow can avoi an further reuces the sening rate of ata packets than the permitte allocations from a respective source noe in a certain network. Thus, a constant flow of ata packets can be achieve whereas the packet 95

115 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization collisions an ata roppe rate is reuce. To achieve an optimum network fairness among all source noes, the avertise winow is set to one (the lowest value) an the congestion winow is ecie by the type of TCP agent use which has a unique characteristic of congestion control algorithm base on the congestion state in the network. A efault TCP elaye acknowlegement timeout is controlle by setting the respective source noes acknowlegement timeout base on their location from the estination noe unlike the efault expiry of acknowlegement timeout which is unable to achieve a network fairness inex value close to one [23, 143]. Generally, in a TCP agent the elaye acknowlegement is set by a factor of two (RFC1122) [78, 144, 145] where an acknowlegement packet will be generate when the estination noe receives one or two ata packets within the efault acknowlegement expiry time. Base on this factor wherein a particular source noe with a capability to sen two ata packets within the efault acknowlegement expiry time reuces the number of acknowlegement packets to one. If only one ata packet is receive within the efault acknowlegement expiry time, then the acknowlegement packet is sent to the eicate noe for the respective packet receive. This technique can further reuce the routing overhea an energy consumption in a network. One of the main component in elaye acknowlegement timeout is the value of RTT. The RTT is known as one cycle time taken for a source noe to sen a ata packet to the estination noe an receives an acknowlegement packet in return to confirm its recipient of the sent ata packet [79, 138]. In a flat or a cluster-base multi-hop linear topology with varying number of source noes have ifferent travel time (t) implication base on the istance factor [119, 123]. Theoretically, in a scenario with linear increasing istance is proportional to the increase in travel time (t) in WSN. Thus, the efault acknowlegement expiry time is unable to justify network fairness over performance especially in a hierarchal arrangement. On the first section, the two propose moels were implemente on a similar flat one-tier multi-hop linear topology setup with two varying throughput factor that reflects on other wireless performance metrics. The first moel uses a higher cycle time in a transmission block with a practical 96

116 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization implementation on a network with the highly ynamic rate of ata transmission. Typically, a transmission block represents 1 Sum H cycle (t) or fair cycle time (t) value escribes the time perio neee for fair ata transmission [28]. While the secon moel uses a lower cycle time in a transmission block with a practical implementation on a network with the low non-ynamic rate of ata transmission. The mathematical formulation in DATM-FFCA an DATM-FTCA is use to set an optimum elaye acknowlegement timeout for respective source noes in a flat one-tier multi-hop linear topology as in Figure 4-3. n hops 2 hops 1 hop Data-E2 Sequence no.1 Sequence no.2 Sequence no.3 Sequence no.4 Sequence no.5 Sequence no.6 En On E2 O2 E1 O1 ND n hops Data-O2 Ack-O2 Ack-O1 2 hops 1 hop Sequence no.1 Sequence no.2 Sequence no.3 Sequence no.4 Sequence no.5 Sequence no.6 Figure 4-3: The DATM-F with source noes (On/En) an estination noe (ND) On the secon section, the two propose moels were implemente on a similar multi-hop linear cluster-base two-tier topology setup with two varying throughput factor that reflects on other wireless performance metrics. The first moel uses a higher cycle time in a transmission block with a practical implementation on a network with the highly ynamic rate of ata transmission. While the secon moel uses a lower cycle time in a transmission block with a practical implementation on a network with the low non-ynamic rate of ata transmission. The mathematical formulation in DATM-CFCA an DATM-CTCA is use to set an optimum elaye acknowlegement timeout for respective source noes in a multi-hop linear cluster-base two-tier topology as in Figure

117 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Cluster hea in Even path Sequence no.2 Sequence no.2 Sequence no.2 Sequence no.5 Sequence no.4 Sequence no.3 Sequence no.1 Sequence no.1 Sequence no.1 Sequence no.6 Sequence no.5 Sequence no.4 Cluster hea in O path Sequence no.1 Sequence no.1 Sequence no.1 Sequence no.6 Sequence no.5 Sequence no.4 E1 Nn N5 Sequence no.2 Sequence no.2 Sequence no.2 Sequence no.5 Sequence no.4 Sequence no.3 O1 1 N4 N3 N2 1 ND N1 Figure 4-4: The DATM-C with source noes (Nn) an estination noe (ND) Moel One; TCP Delaye Acknowlegement Timeout Moel for Fairness Critical Application In the propose moel one; TCP elaye acknowlegement timeout moel for a flat topology for a Fairness critical application (DATM-FFCA) consiers the travel time cost between respective source noes in only one path (either o or even whichever has the greatest hop count) from the estination noe in a network. Since the ual interleaving routing concept introuce in Chapter 3 is base on parallel ata transmission that eliberates the assumption of all ata packets are transmitte in a given cycle time (t). The accumulate hop count which is known as travel cost is a very important parameter to appropriately etermine the elaye acknowlegement timeout for each source noes in both o an even path as illustrate in Figure 4-3. The RTT value in DATM-FFCA is multiplie with the accumulate hop count for respective source noes (only in a single path) with an aition value of n which is a transition factor the perio between all receive packets before the acknowlegement process in a certain network. The transmission perio or can also be calle as the cooling perio between a ata an 98

118 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization acknowlegement packet phase. The transition phase is esigne to clear any outstaning packets which ha overflow from the ata packet phase in the ata stream. The factor of n is etermine by the number of connection which is in parallel to the main calculate hop count. The RTT value is use as a multiplying factor in eriving an optimum elaye acknowlegement timeout for each respective source noes in the network that is base on istance to travel (calculate by the hop count neee to reach the estination noe). The one path cycle time uration in DATM-FFCA is as formulate in equation 4-1. ( ) = (( ( )) + ) 4-1 Where Fair cycle time (t) is one cycle time taken for nn number of source noes in the network to have a fair ata transmission rate to the estination noe, Hi is the hop count for nn number of source noes (either in o or even path), n is the factor for transition perio between en of ata packet an start of acknowlegement packet at the en noe an RTT is the roun trip time (ms). The mathematical formulation in DATM-FFCA as in equation 4-2 erives respective elaye acknowlegement timeout base on a noe sequence in a network to achieve optimum network fairness in the process of sening ata packets an receiving an acknowlegement packet. = ( ) 4-2 Where DACKnn is the TCP elaye acknowlegement timeout for nn number of source noes in both paths (o an even) base on the hop count, where Fair cycle time (t) is from equation 4-1 an RTTnn is the roun trip time for respective source noes base on the separation of noes from the estination noe. In a cluster-base topology, the RTTnn is replace with RTTnn+1 value is in sequential increment accoring to the cluster heas (o an even) that isregar the hop count. Base on equation 4-2, there is a variance between a minimum an a maximum value of elaye acknowlegement timeout, but these variances are not obvious as all ata an acknowlegement packet transmission will be retaine in a Fair cycle time (t). The common ifference between all four propose moels is typically in terms of how the travel cost is erive in hop count an the factor of transition perio (n). Generally, the propose DATM-FFCA has higher travel cost 99

119 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization in a highly ensest network compare to DATM-FTCA. The selection of a RTT value has a crucial impact towars the practical concept of the propose moels. The set value of RTT in all the above formulation is 14.1 ms which is obtaine through a simple simulation setup of two noes istribute in a istance of 100 meters (lineof-sight) using DI-LSR from an average value of ten runs [28]. Each routing algorithm has a unique time factor which contributes towars the RTT value. Theoretically, with a lower value of RTT woul result in a higher throughput value for a given wireless scenario [27, 146]. Thus, the Fair cycle time (t) value obtaine from all four propose moels are base on the suggeste value of RTT. The ata an acknowlegement packet process flow in all four moels are generally similar. However, each moel was esigne with a unique formulation factor with a slightly ifferent technique between each moel that contributes towars the network fairness an overall network performance. In a scenario with conventional TCP agent, the RTT is efine as one cycle time from On/En to sen a ata packet to ND an receives an acknowlegement packet in return. Thus, both ata an acknowlegement packet takes place in one RTT value unlike in propose moels where a series of ata packets are generate from source noes to the estination noe an then followe by a series of acknowlegement packets generate to the respective source noes in a certain network. The propose moels are a comprehensive moel for a multi-hop linear topology to achieve optimum throughput fairness among all source noes in varying number of source noes. With a planne timely sequence of generate ata packets an precalculate elaye acknowlegement packets between a source an a estination noe ensures optimum network fairness in the long run. A multi-hop linear network with six source noes (o path: O1 to On, even path E1 to En) an a estination noe as ND with the propose DATM-FFCA in a controlle environment is shown in Figure 4-5. Figure 4-5 illustrates the entire ata packet sening sequence an receiving the acknowlegement packet sequence with a transition waiting perio in DATM-FFCA. 100

120 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization 1 Sum H cycle time En Sequence no. 1 Data packets Data-En Sequence no. 2 Sequence no. 3 Sequence no. 4 Sequence no. 5 Sequence no. 6 Acknowlegment packets Ack-En Data-En On Data-On Ack-On Data-On E2 O2 Data-E2 Data-O2 Data-En Data-On Transition waiting perio Ack-En Ack-On Ack-E2 Ack-O2 Data-E2 Data-O2 E1 Data-E1 Data-E2 Data-En Ack-En Ack-E2 Ack-E1 Data-E1 O1 Data-O1 Data-O2 Data-On Ack-On Ack-O2 Ack-O1 Data-O1 ND 1 st cycle Data-E1 Data-E2 Data-O1 Data-O2 1 hop 2 hops Data-En Data-On 3 hops Ack-On Ack-O2 Ack-En Ack-E2 4 hops 5 hops 6 hops Ack-O1 Ack-E1 2 n cycle Figure 4-5: The timing iagram of DATM-FFCA Referring to Figure 4-5, all source noes (O1/E1/O2/E2/On/En) initiates the first ata packet (Data-O1/Data-E1/Data-O2/Data-E2/Data-On/Data-En) transmission at time t+1 to the estination noe (ND) at the start of each 1 Sum H cycle (t) or fair cycle time (t). The generate ata packets (Data-O1/Data-E1) from source noes (O1/E1) requires one hop to the esignate estination noe (ND) in sequence no. 1 at a uration of one RTT where else for the generate ata packets (Data-O2/Data-E2) from source noes (O2/E2) requires two hops through the intermeiate noe (O1/E1) to reach the estination noe (ND) in sequence no. 2 with two RTT value. The generate ata packets (Data-On/Data- En) from source noes (On/En) requires three hops through the intermeiate noe (O2/E2 an O1/E1) to reach the estination noe (ND) in sequence no. 3 with three RTT value. Base on the mathematical formulation of DATM-FFCA, the first acknowlegement packets (Ack-On/Ack-En) at sequence no. 4 will be generate at the estination noe (ND) with three hops require through the intermeiate noe (O1/E1 an O2/E2) to reach the respective source noes (On/En) at the ege of seven RTT value. In the secon acknowlegement packets at sequence no. 5 (Ack-O2/Ack-E2) from the estination noe (ND) requires two hops to reach the respective source noes (O2/E2) at the ege of seven RTT value. For the thir an final acknowlegement packets at sequence no. 6 (Ack-O1/Ack-E1) from the estination noe (ND) requires only one hop to reach the respective source noes (O1/E1) at the ege of seven RTT value. 101

121 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Start: Acknowlegment packets A Ack flow process at estination noe *TCP elaye ack timeout formulation For noe On an noe En Data packets arrive at estination noe Default ack packets set at factor of two *Start the DACKnn elaye ack timer Before the elaye ack timer expires The elaye ack timer expires (timeout) TCP ack packets generate for each ata packets at estination noe TCP ack packets generate for 2 ata packets at estination noe Preefine ual reverse path routing table O sequence of noes Even sequence of noes TCP ack packets sen from estination noe Ack flow process at intermeiate noes TCP ack packets sen from estination noe Repetitive process for n number of intermeiate noes Is the biirectional packets Queue size? No Is the biirectional packets Queue size? No Yes Yes TCP ack packets roppe ue to queue overflow TCP ack packets in noe On queue TCP ack packets receive at the respective source noes (On) Ack flow process at source noes TCP ack packets in noe En queue TCP ack packets receive at the respective source noes (En) En: Acknowlegment packets * Refer to pseuucoe algorithm 1 (DATM-FFCA)/algorithm 2 (DATM-FTCA) Figure 4-6: Delaye acknowlegement process in DATM-FFCA an DATM-FTCA From the ual path sequence (o an even route) shown in Figure 4-3, this woul take seven RTT value e.g. 3RTT + 2RTT + 1RTT (o or even source noe) + 1RTT (factor of transition perio) for 1 Sum H cycle (t) or fair cycle time (t) to achieve 102

122 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization the equal ata packet transmission opportunity for all source noes (O1/E1/O2/E2/On/En) in the network. The briefly elaye acknowlegement timeout moel for both DATM-FFCA an DATM-FTCA is shown in the flow chart in Figure 4-6 with significant variant as escribe in Figure 4-7 an Figure 4-9 respective. The propose metho is implemente at the estination noe base on a series of ata an acknowlegement sequence with specific elaye acknowlegement timeout that is preefine before the acknowlegement packets are sent out to the respective source noe as shown in flow chart in Figure 4-6.The mathematical formulation process flow in DATM-FFCA is as briefly escribe in the pseuocoe of Figure 4-7. Algorithm 1: Delaye acknowlegement time for DATM-FFCA Suppose number of noes is nn, o sequence of noes is On, even sequence of noes is En an estination noe is ND = 0 At time t: 1: For On or En-2 0 then: Accumulate require hop(s) for every increment in a loop for one path else IF On-1 0 then: Require hop(s) for one path 2: For (HopCountOn or HopCountEn) (RTT) then: +1 cycle time per ata transmission block (n) 3: For (1 cycle time) - (On or En RTT) nn then: Respective elaye acknowlegement time for nn number of source noes 4: At time t+1: Source noes On/En start sening ata packets to ND 5: At time t+n: ND start sening TCP elaye acknowlegement packets to source noes On/En base on values in step 3. Figure 4-7: Generating elaye acknowlegement timeout in DATM-FFCA 103

123 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Moel Two; TCP Delaye Acknowlegement Timeout Moel for Throughput Critical Application The propose moel two; TCP elaye acknowlegement timeout moel for a flat topology for a Throughput critical application (DATM-FTCA) consiers only the maximum travel time cost on a respective source noe locate furthest in both o an even paths from the estination noe in a network. Since the ual interleaving routing concept introuce in Chapter 3 is base on parallel ata transmission that eliberates the assumption of ata packets are transmitte in cycle time (t) eicate to a single path which accommoates all the source noes within the allocate uration. The maximum number of hop count per-path is accumulate with an aition of n which is a factor of the transition perio between all receive packets phase an the acknowlegement process in a certain network as previously explaine in the propose moel. In the transmission perio phase of DATM-FTCA, any outstaning ata packets in the ata stream is given a lag time to reach the estination noe if neee, especially in a network with lower cycle time (t). The factor of n is etermine by the number of connections to the originator path in calculation process (both o an even) in parallel to the maximum calculate hop count on both paths. The n factor is ae in each respective paths in a network. The travel cost is a very crucial parameter to appropriately etermine elaye acknowlegement timeout for each an every source noes (o an even path) as illustrate in Figure 4-3. The travel cost in DATM-FTCA is calculate from a maximum hop count (base on the istance in the network) for the respective path with an aition of n which is a factor of the transition perio for each path in a certain network. The RTT value is use as a multiplying factor in eriving an optimum elaye acknowlegement timeout for each respective source noes in the network base on istance to travel (calculate by the hop count neee to reach the estination noe in parallel ata transmission). The cycle time in DATM- FTCA is as formulate in equation 4-3. ( ) = ( + ) + ( + )

124 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Where Fair cycle time (t) is one cycle time taken for nn number of source noes in the network to have a fair ata transmission rate to the estination noe, Homax is the maximum istance hop count for nn number of source noes on o path, Hemax is the maximum istance hop count for nn number of source noes on even path, n is the factor of transition perio an RTT is the roun trip time (ms). Equation 4-2 erives the respective elaye acknowlegement timeout for DATM- FTCA base on the noe sequence in a network to achieve optimum network fairness in the process of sening ata an receiving acknowlegement packets. Since DATM-FTCA were esigne using a common technique an implemente on the same platform as in DATM-FFCA, the values obtaine from equation 4-3 can be use in the common mathematical formulation of DATM-FFCA from equation 4-2. Both DATM-FFCA an DATM-FTCA allows the users to achieve optimum network fairness with a minimum trae-off on overall performance in a flat onetier multi-hop linear topology. However, when it comes to performance outcome, that s where the two technique iffers slightly. The measurable ifference between the propose DATM-FTCA an DATM-FFCA is typically how the travel cost is erive in terms of hop counts with the n factor which will result in terms of the overall network performance. Generally, the propose DATM-FTCA has lower travel cost in a highly ensest network compare to DATM-FFCA. The sequence of ata an acknowlegement packet transmission in DATM-FTCA is similar to DATM-FFCA, where in the first phase a series of ata packets are generate from source noes to the estination noe. The secon phase is the transition waiting perio between ata an acknowlegement packets in one cycle time (t). The thir phase is followe by a series of acknowlegement packets which is generate base on the sequence of source noes in a certain network. A multi-hop linear network with six source noes (o path: O1 to On, even path E1 to En) an one estination noe as ND with the propose DATM-FTCA in a controlle environment is as shown in Figure 4-8. Figure 4-8 illustrates the entire ata packet sening sequence an receiving the acknowlegement packet sequence with a transition waiting perio in DATM-FTCA. 105

125 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization En On E2 O2 E1 O1 ND 1 st cycle Sequence no. 1 Data packets Data-En Data-On Data-E2 Data-O2 Data-E1 Data-O1 Data-E1 Data-O1 Data-E2 Data-O2 Data-E2 Data-O2 1 hop 2 hops 3 hops 1 Sum H cycle time Sequence no. 2 Sequence no. 3 Sequence no. 4 Sequence no. 5 Sequence no. 6 Acknowlegment packets Data-En Data-On Data-En Data-On Data-En Data-On Transition waiting perio Ack-On Ack-En 5 hops 6 hops Ack-En Ack-On Ack-O2 Ack-E2 7 hops Ack-En Ack-On Ack-E2 Ack-O2 Ack-O1 Ack-E1 Ack-En Ack-On Ack-E2 Ack-O2 Ack-E1 Ack-O1 2 n cycle Data-En Data-On Data-E2 Data-O2 Data-E1 Data-O1 Figure 4-8: The timing iagram of DATM-FTCA Referring to Figure 4-8, the first phase of the timing iagram is as same as in DATM-FFCA where all source noes (O1/E1/O2/E2/On/En) initiates the first ata packet (Data-O1/Data-E1/Data-O2/Data-E2/Data-On/Data-En) transmission at time t+1 to the estination noe (ND) at the start of each 1 Sum H cycle (t) or fair cycle time (t). The generate ata packets (Data-O1/Data-E1) from source noes (O1/E1) requires one hop to the esignate estination noe (ND) in sequence no. 1 at a uration of one RTT where else for the generate ata packets (Data-O2/Data-E2) from source noes (O2/E2) requires two hops through the intermeiate noe (O1/E1) in sequence no. 2 to reach the estination noe (ND) with two RTT value. The generate ata packets (Data-On/Data-En) from source noes (On/En) requires three hops through the intermeiate noe (O2/E2 an O1/E1) in sequence no. 3 to reach the estination noe (ND) with three RTT value. The ifference in the secon phase between DATM-FTCA an DATM-FFCA is the use of n factor of the transition perio. In DATM-FTCA the n factor of the transition perio is use in both o an even path unlike in DATM- FFCA is only use in one path. Base on the mathematical formulation of DATM- FTCA, the first the acknowlegement packets (Ack-On/Ack-En) at sequence no. 4 will be generate at the estination noe (ND) with three hops require through the intermeiate noe (O1/E1 an O2/E2) to reach the respective source noes (On/En) at the ege of eight RTT value. In the secon acknowlegement packets at sequence no. 5 (Ack-O2/Ack-E2) from the estination noe (ND) requires two hops to reach the respective source noes (O2/E2) at the ege of eight RTT value. For the thir an final acknowlegement packets at sequence no. 6 (Ack-O1/Ack- 106

126 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization E1) from the estination noe (ND) requires only one hop to reach the respective source noes (O1/E1) at the ege of eight RTT value. From the ual path sequence (o an even route) shown in Figure 4-3, this woul take eight RTT value e.g. 3RTT (o source noe) + 1RTT (factor of transition perio) + 3RTT (even source noe) + 1RTT (factor of transition perio) for 1 Sum H cycle (t) or fair cycle time (t) to achieve the equal ata packet transmission opportunity for all source noes (O1/E1/O2/E2/On/En) in the network. The mathematical formulation process flow in DATM-FTCA is as briefly escribe in the flowchart of Figure 4-6 an in the pseuocoe of Figure 4-9. Algorithm 2: Delaye acknowlegement time for DATM-FTCA Suppose number of noes is nn, o sequence of noes is On, even sequence of noes is En an estination noe is ND = 0 At time t: 1: For 0<(On an En+2) nn then: Maximum number of require hop(s) on each path 2: For MaxHopOn an MaxHopEn then: A to the n factor for both o an even maximum hops 3: For ((MaxHopOn+n)+(MaxHopEn+n)) (RTT) then: 1 cycle time per ata transmission block 4: For (1 cycle time)- (On or En RTT) nn then: Respective elaye acknowlegement time for nn number of source noes 5: At time t+1: Source noes On/En start sening ata packets to ND 6: At time t+n: ND start sening TCP elaye acknowlegement packets to On/En base on values in step 4. Figure 4-9: Generating elaye acknowlegement timeout in DATM-FTCA 107

127 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Using DATM-FFCA an DATM-FTCA in a TCP agent will ensure optimum throughput fairness with variation in wireless performance metrics for varying number of source noes in a multi-hop linear topology environment which will suit well in an oil an gas pipeline monitoring application. The key parameter to be efine is the total number of hops (travel cost) between respective source noes an the estination noe in a ual path with an appropriate RTT value for each noe which is base on istance will benefit from network equality. Changes in the scenario where a scalable WSN with selective source noes is applicable in both DATM-FFCA an DATM-FTCA by assigning the precise hop count referring to equation 4-1 an 4-3 respectively in orer to overcome issues with banwith reservation for repeater noes (non-source noes) in a long range network. Network Fairness an Performance Evaluation on Flat One-tier Multi-hop Linear Topology The network fairness an overall network performance were simulate in NS2 [103, 125] to visualise the outcome an impening factors when full eployment of DATM-FFCA an DATM-FTCA in a pipeline network. The static routing algorithm propose in Chapter 3 [12] was implemente with both DATM-FFCA an DATM-FTCA on NS2 where etail analysis was carrie on various wireless performance metrics. Both the propose moel was compare with ADAM+ an ADAM-sn [23] as iscusse in Chapter 2 that was implemente with AODV a reactive routing protocol an DSDV a proactive routing protocol [21, 22]. All the simulate results are from an average value of five runs per cycle on the manipulating variable (nn number of source noes) with ifferent see values in between 1 to 10 over a simulation uration of 500 secons. In a flat one-tier topology, the number of source noes in each cycle was uniformly increase by 12 noes starting from 12 to 120 source noes. To replicate a real-time application, the start time an start sequence of a source noe are generate using a custom ranom function in the simulator. The start time for a source noe is generate ranomly between secons with non-sequential starting noes. All noes inicate in the results were stationary for the entire simulation uration with one estination noe (ND). The configuration an key simulation parameters 108

128 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization employe in NS2 to visualise the effects of varying source noes ensity is as tabulate in Table 4-2. Table 4-2: NS2 simulation parameters for flat topology Channel type Key parameters Raio propagation moel Wireless channel Two Ray Groun Set value MAC type Interface queue type (ifq) Drop Tail/PriQueue Source noes (SN) 12, 24, 36, 48, 60, 72, 84, 96, 108, 120 Destination noe (ND) 1 Queue length (ifqlen) Agent type at source noe Agent type at estination noe Traffic type Transmission range (RX Thresh) Sensing range (CS Thresh) Packet size 50 (DATM) & 100 (other methos) TCP/Reno TCPSink/DelAck Constant bit rate (CBR) 100 meter 125 meter 512 bytes Simulation Results an Discussion for Flat One-tier Multi-hop Linear Topology The ADAM+ an ADAM-sn as iscusse in Chapter 2 was implemente with AODV an DSDV which is presente as AODV+, AODV-sn, DSDV+ an DSDV-sn. The network fairness, as well as the performance of DATM-FFCA an DATM-FTCA, is compare with AODV+, AODV-sn, DSDV+ an DSDV-sn for flat one tier topology in terms of the performance metrics in section

129 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Throughput Fairness Inex vs. Number of Source Noes Throughput fairness inex AODV+ DSDV+ DATM-FFCA 0.3 AODV-sn DSDV-sn DATM-FTCA Number of source noes in network Figure 4-10: Throughput fairness inex vs. number of source noes In a multi-hop linear topology with nn number of source noes with only one estination noe has a crucial implication towars network stability ue to unfair allocation of network resources. The highest esirable throughput fairness inex of 1.0 represents that a network has achieve 100% normalise throughput fairness among all source noes. Referring to Figure 4-10, the throughput fairness inex of both DATM-FFCA an DATM-FTCA outperforms all the other compare methos with a significant ifference of 0.11 to 0.66 between low network ensities with 12 source noes to larger network ensities with 120 source noes. With the propose DATM-FFCA an DATM-FTCA achieve a throughput fairness inex of 0.99 an above with a varying number of source noes in the simulate environment. The technique of elaye acknowlegement base on noe istance use in both moels has a significant improvement in terms of throughput fairness inex compare to a conventional TCP agent use in Chapter

130 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Receive Data Packet Variation vs. Number of Source Noes 100 Receive packet variation (%) AODV+ DATM-FFCA AODV-sn DATM-FTCA DSDV+ DSDV-sn Number of source noes in network Figure 4-11: Receive ata packet variation vs. number of source noes Variation of receive ata packets is another parameter in visualising fairness between source noes in a multi-hop linear network. Referring to Figure 4-11, the percentage of receive ata packet variation in the propose moels were relatively lower when compare with the other moels between low network ensities with 12 source noes to larger network ensities with 120 source noes. Generally, the ifference becomes greater with the increasing number of source noes in a network for all compare methos. The variation between source noe ata receiving rate is typically ue to the ata travel cost (total hop metric) in a multi-hop linear network in particular to the allocate proportion for respective noes. Another crucial factor which contributes towars this variation is roppe ata rate where it is consiere as waste network allocation in a bigger context. The propose moels have a better proportioning of network resources which narrow the variation gap that ensures optimum throughput fairness inex Packet Delivery Ratio vs. Number of Source Noes A percentage of receiving packets over sen packets in a network is a key performance inicator from the perspective of network reliability. The effect of 111

131 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization elivery ratio on varying number of source noes has inicate that both DATM- FFCA an DATM-FTCA has achieve a elivery ratio at a constant rate of above 99 % between low network ensities with 12 source noes to larger network ensities with 120 source noes as shown in Figure The propose moels have significantly reuce the number of roppe packets which is a esirable goal in any WSN compare to the other methos in the simulation environment. Generally, the elivery ratio of the compare methos is inverse proportional to the increasing number of source noes in varying network conitions. Incorporating the propose moels with DI-LSR woul a a great value towars successful ata packet elivery an further reuces unwante waste of network resources in a pipeline network. Delivery ratio (%) AODV+ DSDV+ DATM-FFCA 35 AODV-sn DSDV-sn DATM-FTCA Number of source noes in network Figure 4-12: Packet elivery ratio vs. number of source noes Throughput vs. Number of Source Noes The average throughput over all flows in the network is a performance inicator in a WSN that reassembles the ability of a network to achieve higher throughput. Generally, the throughput (Kbps) rate in a network with the propose moels outperforms all the other compare methos between low network ensities with 12 source noes to larger network ensities with 120 source noes as shown in Figure

132 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Throughput (Kbps) AODV+ DSDV+ DATM-FFCA AODV-sn DSDV-sn DATM-FTCA Number of source noes in network Figure 4-13: Throughput vs. number of source noes Since the formulation of the propose moels is base on ual interleaving route which reuces the ata travel cost (total hop metric) hence, increases the ata transfer rate between source noes to the estination noe in the simulate environment. The techniques use in both propose moels were aequate enough to increase the rate of throughput in a unique characteristic by retaining a constant throughput fairness inex on all varying environment as shown in Figure That s because both variants were esigne using a common technique an implemente on the same platform, however, there is a significant variation in terms of throughput performance. The DATM-FFCA will be practical for an application with a moerate rate of throughput while the DATM-FTCA will be practical for an application with a higher rate of throughput. The mathematical formulation of the propose moels has a ifferent effect in low network ensities an reflects towars a split irection from an intersection point as shown in Figure The other compare metho has emonstrate a significant ifference an egraing throughput rate with a varying number of source noes in a network. The propose moels enable higher ata transfer rate with the available network resources without fairness trae-off a pipeline network. 113

133 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Moel Three; TCP Delaye Acknowlegement Timeout for Fairness Critical Application In the propose moel three; TCP elaye acknowlegement timeout moel for cluster-base topology for a Fairness critical application (DATM-CFCA) consiers the travel time cost between nn number of source noes through a eicate leaer noe (cluster hea) followe by a series of intermeiate cluster heas in only one path (either o or even whichever has the greatest hop count) from the estination noe in a network. Since the ual cluster hea interleaving routing concept introuce in Chapter 3 is base on parallel ata transmission that eliberates the assumption of all ata packets are transmitte in cycle time (t) eicate for one path [96]. The accumulate hop count known as travel cost is a very important parameter to etermine appropriately elaye acknowlegement timeout for n number of source noes locate uner a eicate cluster hea in both o an even path as illustrate in Figure 4-4. The RTT value in DATM-CFCA is multiplie with the accumulate hop count for n number of source noes (respective to the total connection uner each leaer noe) in a eicate leaer noe (cluster hea) followe by a series of intermeiate cluster heas with an aition of cn which is a transition factor the perio between all receive packets before the acknowlegement process in a certain network. The factor of cn is etermine by the number of connection in parallel at a leaer noe to the main calculate hop count. The transition perio between a ata packet phase an acknowlegement packet phase is implemente to clear any outstaning ata packets in the ata stream. The RTT value is use as a multiplying factor in eriving an optimum elaye acknowlegement timeout for respective source noes in the network that is base on istance to travel (calculate by the hop count neee to reach the estination noe). Unlike the previous flat one-tier topology moels [28], the cluster-base moels have an increasing number of hop count ue to the implementation on a cluster-base two-tier topology as shown in Figure 4-4. The one path cycle time uration in DATM-CFCA is as formulate in equation 4-4. ( ) = ( ( + 1) )

134 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Where Fair cycle time (t) is one cycle time taken for ch number of cluster heas in a network, Hi is the hop count for ch number of cluster heas (either in o or even path), n is the number of source noes in a eicate leaer noe, cn is the factor of transition perio an RTT is the roun trip time (ms). Both in the cluster-base an flat topology moel allows the users to achieve optimum network fairness (equal share) with a minimum trae-off on the overall network performance, particularly on throughput. That s because both variants were esigne using a common technique an implemente on the same platform, without compromising on network fairness an overall network performance. Thus, the values obtaine from equation 4-4 can be use in the mathematical formulation of equation 4-2 ue to the similarity with the other propose moels. The throughput fairness inex of close to one can be achieve with the increasing number of source noes an cluster heas in a certain wireless network with the propose DATM-CFCA formulation shown in equation 4-4 an equation 4-2. The sequence of ata an acknowlegement packet transmission in DATM-CFCA is similar to the concept in the previous moels, wherein the first phase a series of ata packets are generate from source noes to the estination noe through a series of cluster heas. The secon phase is the transition waiting perio between ata an acknowlegement packets in one cycle time (t). The thir phase is followe by a series of acknowlegement packets which are generate respectively base on the sequence of source noes in a certain network. Unlike in the previous propose flat one-tier topology moels, all ata an acknowlegement process in the propose cluster-base two-tier topology moels occurs in a eicate ata stream with the propose ual cluster hea interleaving technique from Chapter 3 which can accommoate a higher amount of traffic in a network. A multi-hop linear cluster-base network with six source noes (N1 to Nn) connecte uner two cluster heas (o path: O1, even path E1) to a estination noe as ND with the propose DATM-CFCA in a controlle environment is shown in Figure 4-4. Referring to Figure 4-14, wherein a typical control scenario presumes that all source noes (N1/N2/N3/N4/N5/Nn) initiate the first ata packet (Data-N1/Data-N2/Data-N3/Data-N4/Data-N5/Data-Nn) transmission at 115

135 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization time t+1 to the estination noe (ND) at the start of each 1 Sum H cycle (t) or fair cycle time (t). Figure 4-14 illustrates the timing iagram for a series of ata an acknowlegement packets transfer flow with a transition waiting perio in the propose DATM-CFCA. Sequence no. 1 Sequence no. 2 Sequence no. 3 Sequence no. 4 Data packets Acknowlegment packets Data-Nn Ack-Nn Nn Sequence no. 5 Sequence no. 6 Data-Nn N5 Data-N5 Ack-N5 Data-N5 N4 Data-N4 Ack-N4 Data-N4 N3 Data-N3 Ack-N3 Data-N3 N2 Data-N2 Ack-N2 Data-N2 N1 Data-N1 Ack-N1 Data-N1 Data-Nn E1 Data-N5 Data-N4 Transition waiting perio Ack-Nn Ack-N5 Ack-N4 Data-N3 Data-N2 O1 Data-N1 Ack-N3 Ack-N2 Ack-N1 Data-Nn Data-N5 Data-N4 Data-N3 ND Data-N2 Data-N1 Ack-Nn Ack-N3 Ack-N5 Ack-N2 Ack-N4 Ack-N1 1 st cycle 1 hop 2 hops 5 hops 6 hops 7 hops 8 hops 2 n cycle Figure 4-14: The timing iagram of DATM-CFCA In the o cluster hea sequence, the generate ata packets (Data-N1/Data- N2/Data-N3) from source noes (N1/N2/N3) requires one hop to the eicate leaer noe (O1) in sequence no. 1 an another hop to reach the estination noe (ND) in sequence no. 2 with a total travel cost of two RTT value. Where else in the even cluster hea sequence, the generate ata packets (Data-N4/Data-N5/Data- Nn) from source noes (N4/N5/Nn) requires one hop to the eicate leaer noe (E1) in sequence no. 1 an another hop to reach the estination noe (ND) in sequence no. 2 with a total travel cost of two RTT value. All the ata transmission occurs simultaneously as the travel cost rate is same for the explaine scenario which is subjecte to the ata generation time in a network. In a resultant cycle, if there is a glitch in the ata sening sequence or latency, the ata packets can overflow to the non-activity region which is the transition waiting perio. 116

136 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Base on the mathematical formulation of DATM-CFCA, the first series of acknowlegement packets (Ack-Nn/Ack-N3) at sequence no. 3, (Ack-N5/Ack-N2) at sequence no. 4 an (Ack-N4/Ack-N1) at sequence no. 5 will be generate at the estination noe (ND) with one hop require to the respective intermeiate cluster hea or leaer noe (O1 an E1). There is a small variation of time (t) between the start time since the respective source noes are locate at a ifferent istance even with the same travel cost uner a leaer noe. This interval will also ease the bottleneck issues an control the volume of traffic at a certain cluster hea ue to concurrent acknowlegement packet transfer. The acknowlegement packets (Ack-Nn/Ack-N3) at sequence no. 4, (Ack-N5/Ack-N2) at sequence no.5 an (Ack-N4/Ack-N1) at sequence no.6 will be forware to one hop require to the respective source noes from the intermeiate cluster hea or leaer noes E1 an O1 respectively. The acknowlegement packet from sequence no. 4 will reach the estination noe at the ege of seven RTT value. Followe by the acknowlegement packet from sequence no. 5 will reach the estination noe at the ege of eight RTT value. While the acknowlegement packet from sequence no. 6 will reach the estination noe at the ege of nine RTT value. Since the elaye acknowlegement timeout formulation in DATM-CFCA has a small variation between source noes, the next cycle ata packet generation will be base on this interval. In a controlle environment, the chronology will retain throughout the ata an acknowlegement packets transfer process with minimum changes ue to cross over at intermeiate cluster heas particularly for a network with high ensities of source noes. 117

137 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Start: Acknowlegment packets A Ack flow process at estination noe Data packets arrive at estination noe The elaye ack timer expires (timeout) TCP ack packets generate for each ata packets at estination noe O sequence of cluster hea *Start the DACKnn elaye ack timer Prefixe cluster hea return route to the source noes *TCP elaye ack timeout formulation for source noes in cluster heas (On an En) Default ack packets set at factor of two Before the elaye ack timer expires TCP ack packets generate for 2 ata packets at estination noe Even sequence of cluster hea TCP ack packets sen from the estination noe TCP ack packets sen from the estination noe Ack flow process at intermeiate cluster hea Is the biirectional packets Queue size? Repetitive process for ch number of cluster hea No No Is the biirectional packets Queue size? Yes TCP ack packets roppe at intermeiate cluster hea ue to queue overfow Yes Ack flow process at source noes Is the biirectional packets Queue size? No No Is the biirectional packets Queue size? Yes Yes TCP ack packets in noe n queue TCP ack packets roppe ue to queue overfow TCP ack packets in noe n queue TCP ack packets receive at the respective source noes TCP ack packets receive at the respective source noes En: Acknowlegment packets * Refer to pseuucoe algorithm 3 (DATM-CFCA)/algorithm 4 (DATM-CTCA) Figure 4-15: Delaye acknowlegement process in DATM-CFCA an DATM-CTCA The brief elaye acknowlegement timeout moel for both DATM-CFCA an DATM-CTCA is shown in the flow chart in Figure 4-15 with a significant 118

138 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization performance variant as escribe in Figure 4-16 an Figure 4-18 respectively. From the ual cluster hea sequence (o an even) shown in Figure 4-4, this woul take nine RTT value e.g. (o or even cluster hea: 2RTT + 2RTT + 2RTT) + 3RTT (factor of transition perio) for 1 Sum H cycle (t) or fair cycle time (t) to achieve the equal ata packet transmission opportunity for all source noes (N1/N2/N3/N4/N5/Nn) in the network. The cn factor creates stability when applie to the propose moel to counter the effects of concurrent ata an acknowlegement packet transfer at a certain intermeiate cluster hea. The mathematical formulation process flow in DATM-CFCA is as briefly escribe in the pseuocoe of Figure Algorithm 3: Delaye acknowlegement time for DATM-CFCA Suppose number of source noes is nn, number of cluster hea is ch, o sequence of cluster hea is CHOn, even sequence of cluster hea is CHEn an estination noe is ND = 0 At time t: 1: For CHOn or CHEn-2 0 then: (Accumulate require hop(s) + 1 hop) cn for every increment in a loop for one path else IF CHOn-1 0 then: Require hop(s) for one path + 1 hop 2: For (HopCountCHOn or HopCountCHEn) (RTT) then: +1 cycle time per ata transmission block (cn) 3: For (1 cycle time)- (Nn RTT) nn then: Respective elaye acknowlegement time for nn number of source noes 4: At time t+1: Source noes nn start sening ata packets to ND 5: At time t+n: ND start sening TCP elaye acknowlegement packets to source noes nn base on values in step 3. Figure 4-16: Generating elaye acknowlegement timeout in DATM-CFCA 119

139 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Moel Four; TCP Delaye Acknowlegement Timeout for Throughput Critical Application The propose moel four; TCP elaye acknowlegement timeout moel for a cluster-base topology for a Throughput critical application (DATM-CTCA) consiers only the maximum travel time cost on a respective source noe locate furthest in both o an even sequence of cluster hea path in a network. Since the ual cluster hea interleaving routing concept introuce in Chapter 3 is base on parallel ata transmission that eliberates the assumption of ata packets are transmitte in cycle time (t) eicate for a path which accommoates all the source noes within the allocate uration [96]. The accumulate hop count known as travel cost is a very important parameter to etermine appropriately elaye acknowlegement timeout for n number of source noes locate uner a eicate cluster hea in both o an even path as illustrate in Figure 4-4. The travel cost in DATM-CTCA is the maximum number of hop count per path or referre as the maximum hop count of the n number of source noes (respective to the total connection uner each leaer noe) in a eicate leaer noe (cluster hea) to a estination noe. Then, the maximum hop count in each path is accumulate with an aition of cn which is a factor of the transition perio between all receive packets before the acknowlegement process in a certain network as explaine in the previously propose moels. The factor of cn is etermine by the number of preefine paths (both o an even sequence of cluster heas). The travel cost is a very crucial parameter to etermine appropriately elaye acknowlegement time for each an every source noe in both interleaving paths as illustrate in Figure 4-4. The RTT value is use as a multiplying factor in eriving an optimum elaye acknowlegement timeout for respective source noes in the network that is base on istance to travel (calculate by the hop count neee to reach the estination noe). The cycle time uration in DATM-CTCA is as formulate in equation 4-5. ( ) = (((( + 1) ) + ) + ((( + 1) ) + ))

140 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization Where Fair cycle time (t) is one cycle time that accommoates; chomax is the number of maximum cluster heas in o sequence, chemax is the number of maximum cluster heas in even sequence in a network, n is the number of source noes in a eicate leaer noe, cn is the factor of transition perio an RTT is the roun trip time (ms). The common ifference between the propose DATM- CTCA an DATM-CFCA is typically how the travel cost is erive in terms of hop count with the cn factor. Generally, the propose DATM-CTCA has lower travel cost in a highly ensest network compare to DATM-CFCA. The throughput fairness inex of close to one can be achieve with the increasing number of source noes an cluster heas in a certain wireless network with the propose DATM-CTCA formulation shown in equation 4-5 an equation 4-2. The sequence of ata an acknowlegement packet transmission in DATM-CTCA is similar to the concept in the previous moels, wherein the first phase a series of ata packets are generate from source noes to the estination noe through a series of cluster heas. The secon phase is the transition waiting perio between ata an acknowlegement packets in one cycle time (t). The thir phase is followe by a series of acknowlegement packets which are generate respectively base on the sequence of source noes in a certain network. A multi-hop linear cluster-base network with six source noes (N1 to Nn) connecte uner two cluster heas (o path: O1, even path E1) to a estination noe as ND with the propose DATM-CTCA in a controlle environment is shown in Figure 4-4. Referring to Figure 4-17, all source noes (N1/N2/N3/N4/N5/Nn) initiates the first ata packet (Data-N1/Data-N2/Data-N3/Data-N4/Data- N5/Data-Nn) transmission at time t+1 to the estination noe (ND) at the start of each 1 Sum H cycle (t) or fair cycle time (t). In the o cluster hea sequence, the generate ata packets (Data-N1/Data-N2/Data-N3) from source noes (N1/N2/N3) requires one hop to the eicate leaer noe (O1) in sequence no. 1 an another hop to reach the estination noe (ND) in sequence no. 2 where with a total travel cost of two RTT value. Where else in the even cluster hea sequence, the generate ata packets (Data-N4/Data-N5/Data-Nn) from source noes (N4/N5/Nn) requires one hop to the eicate leaer noe (E1) in sequence no. 1 an another hop to reach the estination noe (ND) in sequence 121

141 Effective TCP Delaye Acknowlegement Timeout Moel for Throughput Fairness Optimization no. 2 where with a total travel cost of two RTT value. All the ata transmission occurs simultaneously as the travel cost rate is same for the explaine scenario which is subjecte to the ata generation time in a network. In a resultant cycle, if there is a glitch in the ata sening sequence or latency, the ata packets can overflow to the non-activity region which is the transition waiting perio. Figure 4-17 illustrates the basic timing iagram for a series of ata an acknowlegement packets transfer flow with a transition waiting perio in DATM-CTCA. Sequence no. 1 Sequence no. 2 Sequence no. 3 Sequence no. 4 Data packets Acknowlegment packets Data-Nn Ack-Nn Nn Sequence no. 5 Sequence no. 6 Data-Nn N5 Data-N5 Ack-N5 Data-N5 N4 Data-N4 Ack-N4 Data-N4 N3 Data-N3 Ack-N3 Data-N3 N2 Data-N2 Ack-N2 Data-N2 N1 Data-N1 Ack-N1 Data-N1 Data-Nn E1 Data-N5 Data-N4 Transition waiting perio Ack-Nn Ack-N5 Ack-N4 Data-N3 Data-N2 O1 Data-N1 Ack-N3 Ack-N2 Ack-N1 Data-Nn Data-N5 Data-N4 Data-N3 ND Data-N2 Data-N1 Ack-Nn Ack-N3 Ack-N5 Ack-N2 Ack-N4 Ack-N1 1 st cycle 1 hop 2 hops 10 hops 11 hops 12 hops 13 hops 2 n cycle Figure 4-17: The timing iagram of DATM-CTCA Base on the mathematical formulation of DATM-CTCA, the first series of acknowlegement packets (Ack-Nn/Ack-N3) at sequence no. 3, (Ack-N5/Ack-N2) at sequence no. 4 an (Ack-N4/Ack-N1) at sequence no. 5 will be generate at the estination noe (ND) with one hop require to the respective intermeiate cluster hea or leaer noe (O1 an E1). There is a small variation of time (t) between the start time since the respective source noes are locate at a ifferent istance even with the same travel cost uner a leaer noe. This interval will also ease the bottleneck issues an control the volume of traffic at a certain cluster hea ue to concurrent acknowlegement packet transfer. The acknowlegement packet from sequence no. 4 will reach the estination noe at the ege of twelve RTT value. Followe by the acknowlegement packet from sequence no. 5 will 122