Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Modeling and Analyzing Anycast Routing in
Reactive and Proactive Networks
Dissertation
Submitted to the Department of Computer Science
Bahria University, Islamabad, Pakistan
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Submitted by: Fazl-e-Hadi
Supervisor: Professor. Dr. Abid Ali Minhas
August 2013
BAHRIA UNIVERSITY, ISLAMABAD
APPROVAL SHEET
SUBMISSION OF HIGHER RESEARCH DEGREE THESIS
Fazl-e-HadiPhD ScholarDepartment of Computer ScienceBahria University, Islamabad Campus,Islamabad.
I hereby certify that the above candidate’s work, including the thesis, has been completedto my satisfaction and that the thesis is in a format and of an editorial recognized by thedepartment as appropriate for examination.
Signature:
Supervisor: Dr. Abid Ali Minhas
Date:————————
The undersigned, certifies that:
1. The candidate presented at a pre-completion seminar, an overview and synthesis ofmajor findings of the thesis, and that the research is of the standard and extentappropriate for submission as a thesis.
2. I have checked the candidate’s thesis and its scope, format, and editorial standardsare recognized by the department as appropriate.
Signature:
Head of Department.
Date:————————
i
DECLARATION OF AUTHENTICATION
I hereby, confirm and declare that the PhD Dissertation submitted by me is solely myown work except where otherwise stated. I also confirm that I have not submitted thisresearch work as a PhD Dissertation partially or fully to any other University. I declarethat I have not done any sort of plagiarism and take sole responsibility of my research work.
Signature:————————
Date:————————
Fazl-e-HadiPhD ScholarDepartment of Computer ScienceBahria University, Islamabad Campus,Islamabad.
ii
ACKNOWLEDGEMENTS
In the name of ALLAH (SWT), The Beneficent, The Merciful. Verily all praise, graceand sovereignty belong to my ALLAH (SWT). I would not have been capable of complet-ing this dissertation without blessings of Allah (SWT).
I am really thankful to my supervisor Dr. Abid Ali Minhas for his valuable contribu-tions and continuous support throughout my research work.
I am thankful to the Higher Education Commission of Pakistan for granting me the schol-arship and also for awarding me the travel grant for presenting one of our research papersin China.
I am highly obliged to my respectable teachers especially to Dr. Mehboob Yasin, Dr.Afaq Hussain Syed, Dr. Mansoor Alam Ansari and Dr. Muhammad Sher and Mr. FazleWahab.
I am highly obliged to the co-authors in some of my papers. They are Dr. Nadir Shah, Dr.Mehboob Yasin, Dr. Fahad Bin Muhaya, Dr. Afaq Hussain Syed, Dr. M. Yunas javed,Mr. Shakir Ullah Shah and Mr. Atif Naseer. They have contributed in the literaturesurvey and formatting of the papers.
My best friends is my asset, they supported me in one way or another during my re-search work especially Dr. Nadir Shah, Dr. Tamleek Ali Tanveer, Dr. Amir Zaman, Dr.Manzoor Ahmed, Dr. Sajid Anwar and Mr. Adeel Akhtar.
I am also thankful to my best students like Sajjad Hussain and Lubna Zafar.
My prime thanks to my parents, brothers and sisters who always prayed for me andsupported me throughout my whole life, especially to my mother. I lost her on March 1,2010. She wished to see this day, may her soul rest in peace, Ameen.
I am thankful to my beloved wife. She extremely supported me during my research work.
May ALLAH (SWT) give reward to all those who helped me and prayed for me dur-ing my research. Ameen.
Fazl-e-Hadi
iii
DEDICATION
I dedicate this work to my loving parents, my beloved wife and children.
iv
Abstract
Anycast routing is an important service which is used for various interesting applications
in different networks. In this dissertation anycast routing protocol has been modeled, im-
plemented and analyzed in various reactive and proactive networks like Delay/Disruption
Tolerant Networks (DTNs), Wireless Mesh Networks (WMNs), Computational Grid envi-
ronment and Wireless Sensor Networks (WSNs).
DTNs are characterized by frequent and long duration partitions and end-to-end
connectivity may never be present between the source and the destination at the message
origination time. Anycast is used for many applications in DTNs such as information
exchange in hazardous/crisis situation, resource discovery etc. In this dissertation, new
classification of DTNs has been proposed first time in the literature, namely: Message
Ferries Networks (MFN), Interplanetary Networks (IPN) and Intermittently Connected
Mobile Ad hoc Networks (ICMAN). The dissertation presents a mobile model for anycast
routing in DTNs. Furthermore, a novel forwarding scheme for ICMANs has been proposed
called Receivers Based Forwarding (RBF). Extensive simulation results with respect to
link availability; group size and buffer size show that the RBF performs better than the
Shortest Path Forwarding (SPF) in term of data delivery ratio, average end-to-end delay
and overall data efficiency.
WMNs are self healing, built through a number of distributed and redundant nodes
to support variety of applications and to provide reliability. Similarly, anycasting is an
imperative service that might be used for a variety of applications in WMNs. Anycast
routing has been modeled, implemented and analyzed in the WMNs. In this dissertation
main focus is to model the anycast scheme considering the traffic from the gateway to the
mesh network having multiple anycast groups. Geocast traffic, in which the packets reach
to the group head via unicast traffic and then broadcasted inside the group, has also been
implemented and analyzed in this dissertation. Moreover the intergroup communication
between different anycast groups has also been analyzed. The network is modeled, sim-
ulated and analyzed for the routing delay and packet delivery ratio. Simulation results
v
shows that proposed scheme is better in terms of routing delay and packet delivery ratio
Since the emergence of WSNs, energy constraint always affects its design and opera-
tions. Many techniques have been proposed to lessen the consumption of energy in WSNs.
By discovering the loop holes in the existing techniques an energy aware anycast routing
protocol (EAA) has been proposed, implemented and analyzed for WSNs. EAA distribute
the sensed data among the existing nodes in a cost efficient manner to gain maximum net-
work lifetime. Through extensive simulation results in TOSSIM, it is demonstrated that
the EAA outperforms the existing technique in terms of energy consumption which leads
to maximum network lifetime.
Computational Grid can perform the computationally extensive jobs by utilizing
the wide spread processing capabilities of volunteer processors. In order to utilize the
wide spread resources, failure options cannot be ignored. Detailed implementation of
fault tolerant techniques using anycast has been done in this dissertation. With reference
to computational grid, the main contribution of this dissertation is the implementation
of fault tolerant techniques using anycasting with modified forwarding mechanism and
its analytical analysis. The communication cost involved in the proposed scheme has
been investigated and compared with the previous techniques in this dissertation. The
implementation of the computational grid is carried out in real testbed based on Alchemi
toolkit.
vi
Extended Abstract
Modeling and analyzing anycast routing in reactive and proactive networks is the main
theme of this dissertation. Various networks are considered like Delay/Disruption Tolerant
Networks (DTNs), Wireless Mesh Networks (WMNs), Computational Grid environment
and Wireless Sensor Networks (WSNs). Anycast routing is an important service which is
used for various interesting applications in the above mentioned networks. Other routing
strategies for mobile ad hoc networks like multicast has also been studied for getting the
thorough knowledge of this emerging and hot area of research.
DTNs are characterized by frequent and long duration partitions. End-to-end con-
nectivity may never be present between the source and the destination at the message
origination time. Anycast is an important service used for many applications in DTNs
such as information exchange in hazards/crisis situation, resource discovery etc. Previ-
ously proposed techniques lacks the mobility of the nodes and that is why an adaptive
anycast routing protocol for DTNs has been proposed in this dissertation. A novel for-
warding scheme for DTNs called Receivers Based Forwarding (RBF), which considers the
number of anycast receivers available through a link as well as the path length to the
nearest receiver through that link in deciding the next hop while forwarding an anycast
bundle, has also been proposed. The effect of group size on the said approach has also been
analyzed. Furthermore classification of DTNs into three subcategories, namely: Message
Ferries Networks (MFN), Interplanetary Networks (IPN) and Intermittently Connected
Mobile Ad hoc Networks (ICMAN) have also been proposed in this dissertation. Effect
of link availability, group size and buffer size has been studied for the ICMAN. Extensive
simulations are conducted and the results show that the RBF performs better than the
Shortest Path Forwarding (SPF) in term of data delivery ratio, average end-to-end delay
and overall data efficiency.
WMNs are self healing, built through a number of distributed and redundant nodes to
support variety of applications and provide reliability. Similarly, anycasting is an imper-
ative service that might be used for a variety of applications in WMNs. Anycast routing
vii
has been modeled, implemented and analyzed in the WMNs. In this dissertation main
focus is to model the anycast scheme considering the traffic from the gateway to the mesh
network having multiple anycast groups. Geocast traffic in which the packets reach to
the group head via unicast traffic and then broadcasted inside the group has also been
implemented and analyzed in this dissertation. Moreover the intergroup communication
between different anycast groups has also been analyzed. The review of the related liter-
ature shows that no one has studied the anycasting and geocasting from gateway to the
mesh network while considering multiple anycast groups and intergroup communication.
The network is modeled, simulated and analyzed for the routing delay and packet delivery
ratio.
Although security was out of the scope of this dissertation but in addition to above
contributions in WMNs, the said protocol has also been analyzed for various security
attacks. The reason being that the field based routing protocols (FBR) are prone to
different active and passive attacks launched by external illegal nodes and internal selfish
nodes. First of all external intruders are eliminated by issuing proper IDs to the actual
mesh clients. A secure field based routing protocol (SFBR) has been proposed for WMNs.
Internal selfish nodes have been detected by a simple technique and as a result the network
performance became better in routing delay and packet delivery ratio. According to the
simulation results only 25% packets reached to the destination in normal routing and rest
of the packets are dropped due to the intervention of the intruders. As in SFBR intruders
are identified and removed from the forwarding list. In case of intruders, alternate path
has been selected to rout the packets to the destination. As a result almost 90% of the
traffic reached to the destination.
Due to limited battery and processing speed, energy constraints always remain an issue
in WSNs. It has an effect on its design and operations. This dissertation also contributes
towards the same issue. By discovering the loop holes in the existing techniques, an energy
aware anycast routing protocol (EAA) has been proposed, implemented and analyzed for
WSNs. EAA distributes the sensed data among the existing nodes in a cost efficient man-
viii
ner to gain maximum network lifetime. Through extensive simulation results in TOSSIM,
it is demonstrated that the EAA outperforms the existing technique in terms of energy
consumption which leads to maximum network lifetime. According to the average results
EAA save almost 9.1% energy at every node of the sensor network. It minimizes the
header to payload ratio which leads to maximum network lifetime and 40% improvement
has been recorded.
For many computationally extensive tasks different volunteer processors may be uti-
lized, the phenomenon known as computational grid. Managing computational grid re-
sources while based upon the vulnerable network media is a challenging task. Moreover
failure options cannot be ignored. Detailed implementation of fault tolerant techniques
using anycast has been done in this dissertation. With reference to computational grid the
main contribution of this dissertation is the implementation of fault tolerant techniques
using anycasting with modified forwarding mechanism and its analytical analysis. The
communication cost involved in the proposed scheme has been investigated and compared
with the previous techniques. The implementation of the computational grid is carried
out in real testbed based on Alchemi toolkit.
ix
LIST OF PUBLICATIONS ORIGINATED FROM THEDISSERTATION
1. Fazl-e-Hadi, Abid Ali Minhas, Nadir Shah, Mehboob Yasin, ” A Novel ForwardingScheme for Adaptive Anycasting in Delay/Disruption Tolerant Networks ”, IndianJ.Sci.Technol. Vol. 3, Issue 12, pp: 1165- 1172, ISSN: 0974-5645 [Indexed by ISI]
2. Fazl-e-Hadi, Abid Ali Minhas, ” EAA: Energy Aware Anycast Routing in WirelessSensor Networks”, Journal of Engineering and Applied Sciences (JEAS), Vol: 30 No.(1), January-June 2011.
3. Fazl-e-Hadi, Abid Ali Minhas, Shakir Ullah Shah, ” A Novel Cost-Based Frameworkfor Communica- tion in Computational Grid using Anycast Routing ”, InternationalJournal of the Physical Sciences, Vol. 6(10), pp.2348-2355, 2011. ISSN 1992-1950
4. Fazl-e-Hadi, Abid Ali Minhas, Atif Naseer, ”Modeling and Analyzing Anycast andGeocast Routing in Wireless Mesh Networks ”, Accepted for publication in Advancesin Electrical and Computer Engineering Journal - ISSN 1582-7445, 2010 [Indexed byISI and SCIE] [Impact Factor: 0.509]
5. Fahad Bin Muhaya, Fazl-e-Hadi, Atif Naseer, ”ESFBR: Enhanced secure field basedrouting protocol in Wireless Mesh Networks”, Indian J.Sci.Technol. Vol. 4, Issue 6,pp: 613-617, ISSN: 0974-5645 [Indexed by ISI]
6. Fazl-e-Hadi, Nadir Shah, Dr. Afaq Hussain Syed, Dr. Mehboob Yasin, ”AdaptiveAnycast: A new Anycast protocol for performance improvement in Delay Toler-ant Networks”, Proceedings of the IEEE International Conference on IntegrationTechnology , Shenzhen , China, 2007, pages 185 - 189
7. Fazl-e-Hadi, Abid Ali Minhas, ”Modeling and analyzing anycast routing in reactiveand proactive net- works”, Proceedings of the 6th International Conference on Fron-tiers of Information Technology, Pakistan, 2009, http://doi.acm.org/10.1145/1838002.
8. Fazl-e-Hadi, Nadir Shah, Dr. Afaq Hussain Syed, Dr. Mehboob Yasin, ”Effect ofgroup size on anycasting with receiver base forwarding in Delay Tolerant Networks”,Proceedings of the IEEE International Conference on Electrical Engineering, Pak-istan, 2007, pages 1 - 4, http://dx.doi.org/10.1109/ICEE.2007.4287301
9. Atif Naseer, M. Yunas javed, Fazl-e-Hadi, ”SFBR: Secure Field Base Routing inWireless Mesh Net- works”, Proceedings of the IEEE 7th International Confer-ence on ICT and Knowledge Engineering, 2009, Thailand, 2009, pages 106 - 111,http://dx.doi.org/10.1109/ICTKE.2009.5397328
10. Fahad Bin Muhaya, Fazl-e-Hadi, Atif Naseer, ”Selfish Node Detection in WirelessMesh Networks”, Proceedings of the IEEE International Conference on Networkingand Information Technology, Philippines, 2010
11. Fazl-e-Hadi, Fahad Bin Muhaya, Atif Naseer, ”Secure Multimedia Communicationin Wireless Mesh Networks”, Proceedings of the 5th International Conference forInternet Technology and Secured Trans- actions (ICITST-2010), London, UK, 2010
x
Contents
APPROVAL SHEET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iAUTHENTICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiDEDICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivEXTENDED ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiLIST OF PUBLICATIONS ORIGINATED FROM THE DISSERTATION . . . . . . . xLIST OF ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivLIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvLIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
CHAPTERS
1 INTRODUCTION 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Scientific Focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 RELATED WORK 72.1 Delay Tolerant Networks (DTNs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Wireless Mesh Networks (WMNs) . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.3 Wireless Sensor Networks (WSNs) . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.4 Computational Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.5 Thesis Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 ANYCAST ROUTING IN DELAY TOLERANT NETWORKS 173.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 Intermittent Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.1.2 Variable Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.1.3 Asymmetric Data Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.1.4 High Error Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Applications of DTNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.3 DTN Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.1 Message Ferries DTNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.3.2 Interplanetary DTNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.3.3 Intermittently Connected Mobile ad hoc Networks (ICMAN) . . . . . . . . 22
3.4 Anycasting in DTNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.5 Anycasting Applications in DTNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.6 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.7 Proposed Adaptive Anycast Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 24
xi
3.7.1 Situation Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.7.2 Group Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.7.3 Message Buffering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.7.4 Bundle Forwarding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.7.5 Receiver Based Forwarding (RBF) . . . . . . . . . . . . . . . . . . . . . . . 273.7.6 Pseudo code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.7.7 Scenario where SPF fails and RBF still works . . . . . . . . . . . . . . . . . 303.7.8 Bundle retransmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.8 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.8.1 Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.8.2 Varying link availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.8.3 Varying group size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.8.4 Varying buffer size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4 ANYCAST ROUTING IN WIRELESS MESH NETWORKS 384.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.2 Wireless Mesh Network Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2.1 Basic Mesh Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.2.2 802.16 based WMNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.2.3 Infrastructure/Backbone WMNs . . . . . . . . . . . . . . . . . . . . . . . . 404.2.4 Client meshing (client WMNs ) . . . . . . . . . . . . . . . . . . . . . . . . . 414.2.5 Hybrid WMNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Wireless Mesh Networks Applications . . . . . . . . . . . . . . . . . . . . . . . . . 424.4 Wireless Mesh Networks advantages . . . . . . . . . . . . . . . . . . . . . . . . . . 434.5 Routing in WMNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.5.1 Traditional ad hoc routing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444.5.2 Field Based routing (FBR) . . . . . . . . . . . . . . . . . . . . . . . . . . . 444.5.3 Group communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.5.4 Anycast routing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.6 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464.7 Proposed Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.7.1 Proposed Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.7.2 Routing Packet to Any Group (Anycasting on Gateway) . . . . . . . . . . . 474.7.3 Geocasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.7.4 Load balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.7.5 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.7.6 Flow chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.7.7 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.7.7.1 Routing delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544.7.7.2 Packet delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.7.7.3 Anycast based Geocasting . . . . . . . . . . . . . . . . . . . . . . 56
4.8 Securing the field based routing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574.8.1 Mitigating the external intruders . . . . . . . . . . . . . . . . . . . . . . . . 58
4.8.1.1 Flow chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584.8.1.2 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584.8.1.3 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.8.2 Mitigating the internal selfish nodes . . . . . . . . . . . . . . . . . . . . . . 634.8.2.1 Flow chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634.8.2.2 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
xii
5 ANYCAST ROUTING IN WIRELESS SENSOR NETWORKS 685.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.2 Recent Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705.3 Optimal Base Station Selection using Anycast Routing . . . . . . . . . . . . . . . . 715.4 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725.5 Proposed EAA: Energy Aware Anycast Routing . . . . . . . . . . . . . . . . . . . 735.6 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.6.1 Simulator selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.6.2 Simulation setup and performance evaluation . . . . . . . . . . . . . . . . . 77
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6 ANYCAST ROUTING IN COMPUTATIONAL GRID 826.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 826.2 Grid Middleware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2.1 Architecture of Alchemi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856.2.2 Programming Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 876.2.3 Fault Tolerance in Alchemi . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.3 Multicast and Anycast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 886.3.1 Multicast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 886.3.2 Anycast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.4 Mathematicl Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 906.4.1 Multicasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 906.4.2 Anycasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 906.4.3 Comparison of Anycast and Multicast with respect to its delay . . . . . . . 916.4.4 Anycast with RBF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.4.5 Probability of anycast receiver’s availability without using RBF . . . . . . . 926.4.6 Probability of anycast receiver’s availability with using RBF . . . . . . . . . 92
6.5 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7 CONCLUSIONS AND FUTURE WORK 957.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 957.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Bibliography 99
xiii
LIST OF ABBREVIATIONS
• DTNs: Delay Tolerant Networks.
• WMNs: Wireless Mesh Networks.
• WSNs: Wireless Sensor Networks.
• MFN: Message Ferries Network.
• IPN: Interplanetary Networks.
• ICMAN: Intermittently Connected Mobile Ad hoc Networks.
• RBF: Receivers Based Forwarding.
• SPF: Shortest Path Forwarding.
• EAA: Energy Aware Anycast.
• TOSSIM: Tiny Operating System Simulator.
• TCP/IP: Transmission Control Protocol/ Internet Protocol.
• MANETs: Mobile Ad Hoc Networks.
• DNS: Domain Name Service.
• AODV: Ad hoc On-demand Distance Vector.
• DSR: Dynamic Source Routing.
• GSR: Gateway Source Routing.
• DDOS: Distributed Denial of Service.
• PDAs: Personal Digital Assistants.
• PDR: Packet Delivery Ratio.
• FBR: Field Base Routing.
xiv
List of Figures
3.1 A simple DTN network model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.2 Bundle Layer (DTN Layer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.3 High Level System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.4 Flow diagram of RBF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.5 Working of Receiver Based Forwarding (RBF) . . . . . . . . . . . . . . . . . . . . . 303.6 Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver Based For-
warding (RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.7 Overall efficiency of Shortest Path Forwarding (SPF) and Receiver Based Forwarding
(RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.8 Average end-to-end delay of Shortest Path Forwarding (SPF) and Receiver Based
Forwarding (RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.9 Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver Based For-
warding (RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.10 Overall efficiency of Shortest Path Forwarding (SPF) and Receiver Based Forwarding
(RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.11 Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver Based For-
warding (RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.12 Overall efficiency of Shortest Path Forwarding (SPF) and Receiver Based Forwarding
(RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.13 Average end-to-end delay of Shortest Path Forwarding (SPF) and Receiver Based
Forwarding (RBF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Basic mesh architecture [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.2 IEEE 802.16 based wireless mesh networks [1] . . . . . . . . . . . . . . . . . . . . . 404.3 Infrastructure/Backbone based WMN [1] . . . . . . . . . . . . . . . . . . . . . . . 414.4 Client WMNs [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.5 Hybrid WMNs [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.6 Ad hoc Routing Protocols hierarchical classification . . . . . . . . . . . . . . . . . 454.7 Architecture of mesh network for anycast routing . . . . . . . . . . . . . . . . . . . 484.8 Anycasting on Gateway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.9 Routing Packet from Internet to Group heads . . . . . . . . . . . . . . . . . . . . . 504.10 Flow chart of anycast, unicast and geocast routing with load balancing . . . . . . . 544.11 Packet Delay comparison of Anycast and Unicast traffic . . . . . . . . . . . . . . . 554.12 Packet delivery comparison of Anycast and Unicast traffic . . . . . . . . . . . . . . 564.13 Packet Delay comparison of Anycast and Unicast base geocasting . . . . . . . . . . 574.14 Developed Scenario of Mesh Network with Intruder . . . . . . . . . . . . . . . . . . 584.15 Flow chart of mitigating the external intruder . . . . . . . . . . . . . . . . . . . . . 594.16 Packets captured by intruders at various levels . . . . . . . . . . . . . . . . . . . . 614.17 Packet delivery comparison of NR and SFBR . . . . . . . . . . . . . . . . . . . . . 614.18 Packet delay time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.19 Comparison of various routing protocols with respect to Packet Delivery . . . . . . 63
xv
4.20 Flow chart for mitigatoing the internal selfish nodes . . . . . . . . . . . . . . . . . 644.21 Packets captured by inruders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.22 Packet Delivery Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.23 Packet Delay Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664.24 Comparison of various routing protocols . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1 Physical topology of the sensor network. . . . . . . . . . . . . . . . . . . . . . . . . 715.2 An examples illustrating anycast between the sensor nodes and the base station. . 715.3 Topology of the CASE-I splitting data at every sensor node . . . . . . . . . . . . . 745.4 Topology of the CASE-II splitting data only at the sensing node . . . . . . . . . . 755.5 Snapshot of CASE-I splitting data only at the sensing node in TinyViz . . . . . . . 785.6 Header to payload comparison of CASE-I and CASE-II. . . . . . . . . . . . . . . . 795.7 Total energy consumption in CASE-I and CASE-II. . . . . . . . . . . . . . . . . . 795.8 Node wise energy consumption for CASE-I and CASE-II. . . . . . . . . . . . . . . 805.9 Network lifetime comparison of both CASE-I and CASE-II . . . . . . . . . . . . . 80
6.1 Grid Architecture Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856.2 Block Diagram of Alchemi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866.3 Backup selection using Multicast . . . . . . . . . . . . . . . . . . . . . . . . . . . . 886.4 Backup selection using Anycast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.5 Scenario Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926.6 Cost Analysis using Multicast and Anycast . . . . . . . . . . . . . . . . . . . . . . 94
xvi
List of Tables
6.1 Nodes participating in the computational grid with their specification . . . . . . . 94
xvii
Chapter 1
INTRODUCTION
1.1 Background
A network is a set of devices (often referred as nodes, which are capable of sending and/or
receiving data generated by other nodes within the network) connected by communica-
tion links. Probably the most important characteristics of a computer network are its
generality, ability to carry many different types of data and support of a wide and ever-
growing range of applications [1][2]. Based upon the variety of applications and type of
hardware used, the types of networks currently are the Internet, Mobile Ad hoc networks
(MANETs), Delay tolerant Networks (DTNs), Wireless Sensor Networks (WSNs) and
Mesh networks etc.
The Internet, a wide area network based upon the Transmission Control Protocol/
Internet Protocol (TCP/IP) protocol suite. These protocols are used for routing data
and ensuring the reliability of messages exchanged. Communication on the Internet is
based on packet switching. A message that is to be transmitted is divided into blocks
called Packets. Packets travel independently from source towards the destination through
network connected by routers. The source hosts, routers and destination hosts are called
nodes. Each packet of same message can take a different path through the network de-
pending upon link availability. The packets of same message may arrive at destination
out of order. It is the responsibility of destination transport layer protocol to arrange the
packets in correct order.
1
Ad hoc networks are wireless networks where fixed infrastructure is not available. Each
node can act both as host and router. If nodes are in communication range of each other
then they can communicate with each other directly. If the nodes are not in communication
range, then other nodes which lie between the source and the destination, act as routers
and forward the packet to the destination. The nodes in ad hoc networks may be mobile
thus these networks are called Mobile Ad Hoc Networks (MANETs). These networks
have the characteristics of high error rates and frequent topological changes as compared
to wired networks [3][4].
Delay tolerant networks (DTNs) are those where end-to-end connection may not be
present. These networks are characterized by frequent and long duration partitions. In
these networks, the packets are transferred in store-forward fashion rather than a commu-
nication flow. The DTN nodes may have limited buffer. The intermediate nodes receive
the packets, store it, and as the opportunity arises the packets are forwarded to the destina-
tion. DTN may be opportunistic or periodically/predictable connected. In opportunistic
DTN the contact schedule between nodes is not known while in predictable DTN the
contact schedule between nodes is known [5][6].
To collect and propagate environmental data, Wireless Sensor Networks (WSNs) use
small, low-cost sensors. Wireless sensor networks make possible observing and controlling
of physical environments from distant locations with better precision. Various interesting
applications of WSNs are environmental monitoring, military surveillance and gathering
the information from the uncongenial locations. Due to its limited battery life time and
computational capabilities, WSNs face serious issues. Various researchers are attracted
to solve these issues through more energy efficient routing protocols, Fault-Tolerance,
localization algorithms and hardware system design [7],[8][9].
Computational grid is a kind of distributed and parallel computing. It is composed of
heterogeneous, geographically distributed resources connected by unreliable network me-
dia. It is used to solve the complex problems in sophisticated and user friendly manner,
which require large amounts of computing resources. A computational grid is a hardware
2
and software infrastructure that provides dependable, consistent, pervasive, and inexpen-
sive access to high-end computational capabilities [10],[11]. It is used to increase the
performance and reduce the cost of computer hardware and software in a variety of ways.
The components of a grid may be heterogeneous and dynamic in nature. For volunteer
participation in the grid environment, grid middleware software suite is deployed on every
volunteer machine. In order to perform user’s tasks each participating machine offer some
essential functionalities [12][13].
A self healing wireless network which is built through a number of distributed and
redundant nodes to support variety of applications and provide reliability is known as
mesh network. The basic aim of the wireless mesh networks (WMNs) is the guaranteed
connectivity. Wireless mesh networks are gaining popularity for its wide range of applica-
tions. Wireless mesh networks have gained substantial consideration as an unconventional
solution to applications such as community networks, enterprise networks, and last mile
access networks to the Internet [1][14].
Anycast is an important service in addition to unicast and multicast routing. As stated
by the original proposal [15], ”an anycasting service provides stateless best effort delivery
of an anycast datagram to at least one host, and preferably only one host, which serves
the anycast address.” The purpose of the anycast protocol is to connect any desirable
node (machine) without caring about any particular one. Anycast has many interesting
applications which includes resource discovery, locating and communicating with any one
among the distributed servers, service access point within a network and file sharing in
any distributed system. Being on internet it has a significant use in DNS (Domain Name
Service), NTP (Network Time Protocol) and reliable multicast protocols. It has also some
vital applications in disaster rescue fields. To support all of the above applications anycast
is an important service. The anycast routing objectives are:
• Efficient power management in the energy constrained networks
• Maximizing message delivery ratio
• Minimize message latency
3
• Maximizing routing efficiency
• Fair forwarding responsibilities among the nodes.
1.2 Motivation
The importance of anycast routing and its use in variety of applications has been briefly
discussed in the above section. It will be discussed in detail for every mentioned network
in chapter 3-6. An extensive literature survey (presented in chapter two), authors are of
the opinion that there is an adequate room to do research for modeling and analyzing
the anycast service in the above mentioned networks, which is the most important service
used in variety of applications.
There are some real issues in the anycast routing like nonexistence of a mobile model
for anycast service in delay tolerant networks, fair forwarding mechanism, introducing
the concept of anycast groups in wireless mesh networks, security issues in wireless mesh
networks, cost analysis of anycast routing in computational grid environment and power
aware anycast routing in WSNs. These issues need serious attention and that’s why the
focus of this research is to solve these issues related to anycast routing in the mentioned
networks.
In simple words, the effort made in this thesis will reduce the energy consumption in
the energy constrained networks. Using the techniques presented in this dissertation will
minimize the delay between packets. The forwarding scheme will maximize the packet de-
livery, which is very important in emergency situations. These outcomes are very beneficial
in hazardous situations.
1.3 Scientific Focus
The main issues which are mentioned in the motivation section is the scientific focus of
this research, these are:
Mobile model for the anycast routing in delay tolerant networks include the assumption
4
by the previous work that nodes in the network are stationary. The connectivity among
the nodes was the mobile devices that act as carrier to deliver messages for the nodes.
The mobile devices do not generate messages themselves, i.e. message carriers are fixed.
Also the moving patterns of these mobile carriers can be obtained.
Fair forwarding mechanism includes the concept of the optimization in the forwarding
policy which leads to the maximum packet delivery ratio, minimal delay etc.
Introducing the concept of anycast groups in wireless mesh networks is related to the find-
ings that group communication in general and anycast group communication in particular,
should be introduced in the wireless mesh networks. It also includes the need of securing
these protocols in wireless mesh networks.
Performance analysis of anycast routing in computational grid environment reveals the
implementation of the anycast routing in real environments and its performance compar-
ison with other proposed mechanism.
Power aware anycast routing in wireless sensor networks deals with the critical issues of
power management, header to payload ratio, data integrity, collision at MAC layer etc.
To solve the above mentioned problems, the anycast routing has been modeled and ana-
lyzed in all of the above mentioned networks and are presented in this dissertation. The
simulation tools are also verified for the known inputs and known outputs. The results
are also compared with the previously proposed schemes.
1.4 Thesis Organization
The remainder of the thesis is organized as follows.
• Chapter 2 briefly describes the related work in all of the above mentioned networks
i.e. Delay Tolerant Networks (DTNs), Wireless Mess Networks (WMNs), Computa-
tional Grid and Wireless Sensor Networks (WSNs). It includes the state of the art
work related to anycast routing in the above mentioned networks.
• Chapter 3 focuses on the presentation of mobile anycast model for DTNs. The
5
classification of DTNs into three subcategories has also been proposed, namely:
Message Ferries Network (MFN), Interplanetary Networks (IPN) and Intermittently
Connected Mobile Ad hoc Networks (ICMAN). Furthermore, a novel forwarding
scheme for DTNs called Receivers Based Forwarding (RBF) has been proposed.
The simulations results in NS-2 are also included.
• In chapter 4, the concept of anycast groups in wireless mesh networks has been
introduced and analyzed for its performance issues. The simulation results for the
proposed mechanism in OMNet++ have been included. Furthermore the specified
protocols have also been analyzed for their various security issues.
• Chapter 5 starts with the energy issues related to anycast routing in WSNs. By
discovering the loop holes in the existing techniques, an energy aware anycast routing
protocol (EAA) has been proposed for WSNs. EAA distribute the sensed data among
the existing nodes in a cost efficient manner to gain maximum network lifetime.
Through extensive simulation in TOSSIM, it has been demonstrated that the EAA
outperforms the existing techniques in terms of energy consumption which leads to
maximum network lifetime.
• Chapter 6 gives the detailed implementation of fault tolerance techniques in compu-
tational grid. The main contribution is the implementation of fault tolerant tech-
niques using anycasting and its analytical analysis for the computational grid. The
communication cost involved in the proposed scheme and its comparison with the
previous techniques has been presented. The implementation of the computational
grid has been carried out in Alchemi toolkit which is based Microsoft .Net framework.
• Chapter 7 concludes the dissertation by highlighting the important issues in modeling
and analyzing the anycast routing in reactive and proactive networks. Possible future
directions in the mentioned networks have also been discussed in this chapter.
6
Chapter 2
RELATED WORK
This chapter describes the available literature regarding the anycasting routing. Detailed
study of the available literature in the reactive networks like DTNs which is based upon
the reactive routing protocol i.e Dynamic Source Routing (DSR), proactive network like
WMNs which is based upon the proactive Field Base Routing (FBR), WSNs and Compu-
tational Grid, leads to the fact that the anycast routing in the above mentioned networks
needs serious attention. Solving the problems found in the available literature in the above
mentioned networks was the focus of this research. Rest of the chapter portrays the state
of the art work in anycast routing in the above mentioned networks.
2.1 Delay Tolerant Networks (DTNs)
The delay tolerant network architecture, proposed by the K. Fall [16] for interoperability
among challenged networks, describes the peculiar connectivity behavior of Terrestrial Mo-
bile Networks and Exotic Media. Other authors discussed the DTN as partially connected
ad-hoc networks, opportunistic connectivity and ’challenged networks’ in [17][18][19].
Delay tolerant network is characterized by frequent and long duration partitioning of
the network. Thus, end-to-end delay may become too long to be supported by the TCP/IP
model. Further, often owing to frequent partitioning, end-to-end connectivity between the
source and destination nodes may not be present at the time of message generation. In
literature, little agreement among researchers over labels used to refer to different kinds
7
of networks which fall under the category of delay/disruption tolerant networks has been
found. For example, K. Fall named Terrestrial Mobile Networks as delay tolerant networks
[16]; Vahdat et al. talk about partially-connected networks [19]; while highly partitioned
networks have been studied in [20][21]; and Zhao et al. refer to it as message ferrying
[22]. Researchers have also studied similar phenomena under the title of intermittently
connected mobile networks [23], delay tolerant mobile networks [24], challenged networks
[25], disruption tolerant networks [26] and ad hoc relay wireless networks [27].
After studying the salient features and distinguishing characteristics of these networks,
which have been referred in the literature as DTNs, broad categories of delay/disruption
tolerant networks have been proposed. According to our classification, DTNs may be
further classified into three subcategories (explained in the next chapter).
Furthermore, Anycast routing has been studied extensively for Internet and mobile
ad hoc networks. Shah et al. [28] proposed a selective anycast service to choose best
mirror server among the many available servers, depending upon some application-specific
metrics, like path length and server load. This is an application-layer anycasting on the
Internet, where the connectivity is guaranteed and the end-to-end delay is within the range
of TCP/IP protocol suite.
Katabi et al. [29] have designed an IP-anycast protocol. Inter-domain anycast has been
divided into two phases. In the first phase anycast routes are built without consuming
much bandwidth and storage space. In the second phase anycast routes are generated
according to the beneficiary domain’s interests. Park et al. [30] discuss the anycast routing
in the context of mobile ad hoc networks. They have proposed that how multiple classes
of unicast routing can be utilized for the anycast route construction and maintenance.
In [31], Park et al. explored various military applications of anycast routing . Wang
et al. [32] present the anycast routing protocol for MANETs on the basis of ad hoc on-
demand distance vector (AODV) protocol. For each entry in the routing table, the authors
record anycast group number and confirms that the corresponding destination is member
of an anycast group or not. Forwarding of anycast bundles at a node is done as follows:
8
it retrieves the entries from its routing table for the desired anycast group and selects the
one having shortest path in term of hop count. Peng et al. [33] propose anycast routing
technique based upon DSR (dynamic source routing protocol) for MANETs (Mobile ad
hoc Networks) that returns route-error message to the source in case of link failures. Zheng
Xie et al. in [34] propose an efficient way for handling link failures in MANETs. If link
failure occurs at the intermediate node, then the intermediate node discovers an alternate
route for the destination instead of sending route error to the source, causing reduction in
control overhead.
Nitin et al. [35] propose a MAC layer anycasting with consultation of unicast routing
protocol. If there is more than one route to the destination available at routing layer
then MAC layer forwards to one of the neighbors on either path according to signal
strength. They have also discussed in detail its effect on various categories of unicast
routing protocols.
Recently, Yili Gong et al. [36] discussed the DTNs with message ferries where cor-
respondent nodes in the network are stationary. The connectivity among the nodes is
provided by mobile devices that act as carriers to deliver messages to the destination
nodes. Specifically, the mobile devices do not generate messages themselves and the mov-
ing patterns of these mobile carriers are well defined.
The above detailed literature survey reveals that there is a need to categorize the
DTNs and present a mobile model for the anycast routing in DTNs with fair forwarding
mechanism. All these issues are addressed for the DTNs in this research and presented in
chapter 3.
2.2 Wireless Mesh Networks (WMNs)
The basic aim of the wireless mesh networks (WMNs) is the guaranteed connectivity.
Wireless mesh networks are gaining popularity for its wide range of applications. Wireless
mesh networks have gained substantial consideration as an unconventional solution to ap-
plications such as community networks, enterprise networks, and last mile access networks
9
to the Internet [1]. Akyildiz et al. [1] state that the wireless mesh network is a class of
networks where some nodes are fixed for serving as gateway for the Internet connectivity
and others are mobile nodes which gives access to the mobile nodes in multihop fashion.
Due to redundant links, connectivity is not an issue in these networks. Routing in the
wireless mesh networks always attracts the researchers. Most of the recent studies regard-
ing the routing in the wireless mesh networks focus on the traffics flows from mesh nodes
to the mesh gateways using AODV [37] or OLSR [38]. Various unicast routing algorithms
have been proposed in different studies like [38][39][37][40][41].
Because of its scalability and robustness many researchers like Lenders et al. [42] and
V. Park et al. [43] presented the field based routing algorithms. Recently, Baumann et al.
[44] presented a field based routing algorithm for routing the packets from the mesh nodes
to the gateway in anycast fashion. The authors presented a field based routing algorithm,
HEAT, which computes the temperature field keeping the gateway as the source of heat.
In their later work, Baumann et al. [45] presented the gateway source routing (GSR)
algorithm for routing the packets into the wireless mesh network. The authors use the
routing path in backward direction, the path which is build up by the mesh clients by
sending the packets to the gateway. In order to route the packets from gateway to the
mesh nodes, it is necessary that the mesh clients first send the data to the gateway which
seems to be very anomalous limitation of the proposed scheme. The concept of the field
based routing algorithms is very straight forward. In these algorithms the data moves along
the steepest path towards its destination. Khan et al. [46] introduced a hybrid wireless
mesh protocol (HWMP) which is the marriage of two seemingly opposite technologies i.e.
flexibility of on-demand route discovery and enabling efficient proactive routing as well.
Field base routing algorithms are ingeniously used for various applications like load
balancing in wide area networks [47], data gathering in sensor networks [48], placement of
sensor nodes [49] and, routing in MANETs [43][44].
After investigating the existing literature in WMNs, authors are of the point of view that
there is a need to present the concept of anycast groups in WMNs. An anycast model
10
considering the traffic from the gateway to the mesh clients, having different anycast
groups has been proposed. To the best of our knowledge, the idea of anycast group
communication for wireless mesh networks has not been studied before. Another related
concept is the geocasting, it is a phenomenon in which packets are delivered to a particular
group belonging to a specific geographical location. There are various geocast algorithms
available in the literature e.g. [50] and [51] which exclusively depend upon the exact
geographical information of source and the destination which are very hard to obtain [52].
The dissertation presents a geocast model which carries the traffic from the gateway till
the group head in unicast fashion following the gradient based routing and then broadcasts
it inside the group. Furthermore, the protocols presented in WMNs are also considered
for some security issues, i.e. detection of internal and external intruders. Detailed work
in WMNs is presented in chapter 4.
2.3 Wireless Sensor Networks (WSNs)
In the recent past, many researchers are attracted towards the critical issues related to
WSNs, for example, Abid et al. [53] have studied various protocols for its energy efficiency.
Many other techniques have been proposed to lessen the consumption of energy in WSNs
[54]. Sensor nodes have various energy and computational constraints because of their
inexpensive nature and ad hoc method of deployment. Significant research has been
focused at overcoming these deficiencies through more energy efficient routing, localization
algorithms, system design, and load balancing [9].
In the literature various authors uses single base station (sink node) and the sensor
nodes gather the data and deliver that data to this single node [55], [56], [57]. In some
studies the authors consider various base stations and the sensed data may be split into
various pieces and deliver it to various base stations [54], [58]. The research community
has studied the anycast routing extensively for the Internet and MANETs (Mobile ad
hoc networks) [29], [28], [30] and specifically for the delay tolerant networks [36], [59]
and [60]. But the Internet and MANETs scenarios are very different from the wireless
11
sensor networks (WSNs) specifically because of the strict energy limitations. The power
awareness is the major concern in WSNs. It has attracted the research community to
optimize the existing techniques and present novel energy aware techniques.
Rich literature is available for the energy efficiency in medium access control (MAC)
protocols [61] and multicast routing [62], [63], [64]. Unicast and broadcast routing proto-
cols got the similar attention [65] and [66]. The anycast routing did not get the desired
attention. Abid et al. [53] have focused on the energy efficiency of various protocols in
WSNs but did not study the anycast routing protocol.
The authors in [67] proposed the first anycast protocol for WSNs. The authors pro-
posed the idea of delivering the packets to the nearest sink node, which does not perform
well always. The authors of [68] also proposed another anycast routing technique. They
built the source trees; this technique is similar to the nearest sink node scheme presented
in [67] because both approaches are based upon the selection of minimum energy path.
As stated earlier minimum energy paths does not always guarantee the maximum network
lifetime.
The most recent work on anycast routing is the optimal base station selection algorithm
for anycast routing which has been proposed by Hou et al. [69]. The authors consider the
surveillance video example and proposed an analytical evaluation for maximization of the
network lifetime. In the surveillance video example it is necessary to forward all sensed
bit streams to a single base station rather than forwarding to various base stations. This
is because partial bit streams may be decoded properly. For load balancing the proposed
technique splits the bit streams and various sub flows and forward them to a single base
station through various paths. The authors have used a routing mechanism based on
linear programming to maximize the network life time, but splitting the data at every
intermediate node puts extra burden on the network and affects the goal of the authors.
Following limitations in the given scheme have been identified which need to be addressed
to maximize the network life time which is an important measurement parameter for
WSNs.
12
• If we divide the same flow into multiple sub-flows and it is routed through different
paths then many sensor nodes should take part in this routing which will be otherwise
in sleep mode. The overall energy consumption should increase which is very critical
for the sensor network.
• Different sub-flows should follow different paths so any change in the topology causes
data loss which will lead to the retransmission of the data which increases the over-
head and energy loss.
• If we involve more nodes for routing in wireless sensor networks there will be more
problems at the MAC layer due to packet collision.
• Header to payload ratio will increase for more flows.
• If any segment of the data is lost then the whole packet will be dropped at the base
station.
• Proper reassembly mechanism should be adopted at the base station.
• Jitter will occur for multiple flows, so a playback buffer is required at the base station.
• The scheme is forwarding the data in the form of sub-flows, which is the result
of division of incoming bits to be forwarded on different paths. This involves more
nodes in the routing and consequently more energy is used and lifetime of the network
decreases.
• No limits on the sub-flows, i.e. how many sub-flows should be there, or it will divide
the flow into sub-flows at every intermediate sensor node which will involve almost
all the sensor nodes in the network which is not energy efficient.
After investigating the literature for anycast routing in WSNs it is observed that exist-
ing approaches for optimal selection of the base station using anycast routing [69] have
serious issues of power management, header to payload ratio, data integrity, collision at
MAC layer, playback buffer and jitter. These issues need to be resolved and hence this
13
dissertation focuses on the presentation of energy aware anycast routing in WSNs (EAA).
The detailed work is presented in chapter 5.
2.4 Computational Grid
Grid can be classified, according to the types of shared resources and their functionali-
ties, into computational Grid, Data Grid, Storage Grid, Equipment Grid, knowledge Grid,
Interaction Grid [70] etc. Computational grid is a kind of distributed and parallel com-
puting. It is based upon heterogeneous, geographically distributed resources connected by
unreliable network media. It is used to solve the complex problems, in a sophisticated and
user friendly manner, which require large amounts of computing resources. A computa-
tional grid is a hardware and software infrastructure that provides dependable, consistent,
pervasive, and inexpensive access to high-end computational capabilities [71] [72]. It is
used to increase the performance and reduce the cost of computer hardware and software
in a variety of ways. The components of a Grid may be heterogeneous and dynamic in
nature. Similarly packet loss is also common in a long range of geographically distributed
network and heterogeneous in nature, so user assigned jobs are always prone to different
type of failures, errors and faults [73].
Fault tolerance is an important service of Grid, which ensures the delivery of a service
despite the presence of fault. The issue of fault tolerance in Grid computing is higher
than traditional parallel computing [74], [75], [76], which includes a wide range of errors,
failures and faults and shows fragility of grid environment. So the fault tolerance becomes
very important.
Nazir et al. [77] proposed a proactive approach for scheduling jobs in computational
grid. In this technique, history is maintained about the grid resources and jobs are sched-
uled according to their history. Check pointing [78], [79] is a common and an efficient
technique to save the state of the computation on stable storage periodically. It is used
to resume the job execution from the previous consistent stored state rather from the be-
ginning. It increases the application response time and hence the the efficiency of system
14
improves. It helps in load balancing by migrating jobs from loaded machine to less loaded
machines. Similarly, it helps in fault resiliency by migrating a job from faulty machine to
stable machine. It is mainly used for long running jobs to save the work to be recomputed
from the beginning. The idea proposed in [80] uses multicast technique, which multicasts
address of executer machine in order to select backup machine. This technique causes the
following problems:
• Data loss or delivered out of order will increase unreliability
• Increases the network traffic delays
• Many multicast servers do not distinguish any client
Therefore it is easy to join a group and watch the data that is being sent to it i.e. Dis-
tributed Denial of Service (DDOS) attack is possible. The authors of [80] used multicast
for backup selection of the executor machines by transferring a single packet to multiple
recipients which causes the above mentioned problems. The idea of anycast is available
in [13] but the authors did not analyze the anycast communication against the multicast.
An anycast backup selection mechanism with Receiver Based Forwarding (RBF) has been
proposed. The involved communication cost in multicast and anycast scenarios has been
analyzed. The analysis has been done by creating a computational testbed using Alchemi
middleware. Detailed model and analysis has been presented in chapter 6.
2.5 Thesis Contribution
By thoroughly investigating the related work it was observed that there are various is-
sues related to anycast routing in various networks. This thesis contributes towards the
following points related to anycast routing in various networks.
1. The classification of DTNs into three subcategories, namely: Message Ferries Net-
works (MFN), Interplanetary Networks (IPN) and ICMAN i.e Intermittently (Oc-
casionally) Connected Mobile Adhoc Networks.
15
2. Presentation of mobile model for anycast routing in DTNs.
3. Proposing a novel forwarding scheme for DTNs called Receivers Based Forwarding
(RBF), which considers the number of anycast receivers available through a link as
well as the path length to the nearest receiver through that link.
4. Presentation of an anycast routing model for wireless mesh networks.
5. Geocast traffic in which the packets reach to the group head via unicast traffic and
then are broadcasted inside the group.
6. Intergroup communication between different anycast groups in WMNs.
7. Security issues in WMNs.
8. The study proposed an energy aware anycast routing protocol (EAA) for wireless
sensor networks. EAA smartly distributes the sensed data among the existing nodes
in a cost efficient manner to gain maximum network lifetime.
9. Detailed implementation of fault tolerance techniques using anycast.
10. Cost analysis of various techniques for communication cost in computational grid.
16
Chapter 3
ANYCAST ROUTING INDELAY TOLERANTNETWORKS
In this chapter anycast routing in DTNs has been presented. Chapter starts with the
introduction of DTNs and then its applications and architecture have been described.
New classification of DTNs into three subcategories has been proposed. After this anycast
routing in DTNs has been described. Problems in the existing anycast routing have been
identified. Proposed adaptive anycast routing has been presented. A novel forwarding
scheme known as Receiver Base Forwarding (RBF) has been presented. Flow chart, pseudo
code and working of RBF have been presented. At the end performance evaluation and
simulation results of RBF against the Shortest Path Forwarding (SPF) has been presented.
The chapter ends with the concluding remarks.
3.1 Introduction
In Internet and ad hoc network, routing is always based on the notion of end-to-end
connectivity. However still there are many scenario of ad hoc networks where these end-
to-end delays may be for longer times i.e several minutes, hours or days. This delay is
longer than round trip time (RTT) of protocols such as Transmission Control Protocol
(TCP). In these scenarios, fully connected paths i.e end-to-end paths, may never or rarely
exists. Such communication networks are commonly grouped as delay/disruption tolerant
17
networks (DTNs). These networks are characterized by intermittent connectivity, variable
delay, asymmetric data rates and high error rates. A little detail is as follows:
3.1.1 Intermittent Connectivity
If there does not exist end-to-end path between the source and destination at the time of
message generation, then packet is forwarded to next hop (according to routing algorithm).
The intermediate nodes store the packets and as opportunity arises, the packets are for-
warded to the next node, according to a particular routing algorithm. In this situation,
for end-to-end communication the TCP/IP protocols do not work.
3.1.2 Variable Delay
Long propagation delays between nodes, variable queuing delays at nodes and long du-
ration partition of the network cause large end-to-end path delays. The Internet and
MANETs protocols that require quick return of acknowledgment would fail in DTNs.
3.1.3 Asymmetric Data Rates
The Internet supports moderate asymmetries of bi-directional data rate for users with
cable TV or asymmetric DSL access. But in case of DTNs asymmetries are large, so the
conventional protocols may not work [81].
3.1.4 High Error Rates
When data received has errors, then error is either corrected or retransmission of data
occurs (according to error detection and correcting coding scheme). For a given link-error
rate, fewer retransmissions are needed for hop-by-hop in DTNs than for end-to-end re-
transmission (linear increase vs. exponential increase).
Connectivity among the nodes in DTNs:
Basically there are two types of connectivity in DTNs.
18
• Opportunistic/Unpredictable Connectivity: Opportunistic/unpredictable con-
nectivity among the nodes is not scheduled. That is the connectivity between nodes
is probabilistic. For example in Vehicular communication [6], vehicles with comput-
ing and communication capability move randomly, independently and at the speed
of their own choice (within range). Data can be sent or received through wireless
Personal Digital Assistants (PDAs), if it came in to the communication range, or if
it passes through an information kiosk [81].
• Periodical/predictable Connectivity: In this case the connectivity schedule is
known or can be obtained. To provide communication services to the sparse nodes
in the network special set of mobile nodes are used which is called message ferries
(MF)[22]. Message ferries move according to the scheduled contacts. It introduces
the concept of non-randomness in the mobility of nodes and exploits such non-
randomness for communication.
3.2 Applications of DTNs
There are many applications of DTNs, few of them are described here briefly:
1. The best example of DTNs is the deep space communication, because the round-trip
duration between the planets is several minutes. This is because the long distances
between them. So DTNs are applicable in such scenarios. Another aspect is the
disconnection in satellite communication, which causes due to its position. The con-
tacts between different planets are scheduled and the idea of DTNs is fully applicable
in space communication.
2. Remote access to the Internet resources is another interesting application of DTNs.
Due to the lack of infrastructure like GSM, satellite and wired network, remote
areas are often disconnected from the Internet. In such scenarios the Internet ac-
cess can be provided to these remote areas by exploiting the existing resources and
human/vehicle mobility. [82].
19
3. Sensor network with intermittent connectivity is another application of DTNs. A
large number of sensors with little storage, power and limited wireless coverage can
be deployed in a region to collect the data of concern. If sensor nodes are either
mobile or may not have permanent connectivity then it is the example of DTNs. For
example collection of oceanographic data from the devices installed for it [83].
4. In crisis environments, like disaster recovery or battle field, the nodes may be sparsely
distributed. The network partitioning may be frequent and for long time. In such
hazards situations DTNs are fully applicable [19].
3.3 DTN Architecture
The delay tolerant network architecture is based upon asynchronous messages (bundle)
forwarding model. It is an overlay built upon the existing underlying structure. The nodes
having the DTN module installed on it can receive and forward the bundles, these nodes
are called DTN nodes. It can operate with other normal routing nodes. DTN node may
have redundant links to multiple hops. Figure 3.1 shows a simple DTN network model.
In this figure the nodes n1, n3, n4, n5, n7 and n8 are the DTN nodes. Other nodes are
called regular nodes. So the DTN nodes create overlay architecture. The DTN nodes
transmit the bundles in store-and-forward manner. For bundle storage and forwarding
process every DTN node has a fixed size of buffer. [84].
As discussed in the literature survey that, little agreement among researchers has been
found over the name used to refer different kinds of networks which fall under the cate-
gory of DTNs . For example, ‘Terrestrial Mobile Networks’ is used by K. Fall et al. [16],
‘Partially-connected networks’ is used by Vahdat et al. [19], ‘Highly partitioned networks’
is used by [20], [19], ‘message ferrying’ is used by Zhao et al. [22]. ‘Intermittently con-
nected mobile networks’ is used by [23], ‘Delay tolerant mobile networks’ is used by [24],
‘Challenged networks’ is used by [25], ‘Disruption tolerant networks’ is used by [26] and
‘Ad hoc relay wireless networks’ is used by [27]. After studying the salient features and dis-
tinguishing characteristics of the example networks, broad categories of delay/disruption
20
Figure 3.1: A simple DTN network model
tolerant networks have been proposed in this research. The DTNs may be further classified
into the following three subcategories:
3.3.1 Message Ferries DTNs
In Message Ferries DTNs, there are special nodes, called message ferries (MFs) that move
in the region on well-defined paths and are responsible for data transfer among the regular
nodes. MFs do not generate data by themselves. MFs are equipped with more functions
and capabilities, like bundle storage capacity, longer battery lifetime etc. DTNs with
message ferries are described by Zhao et al. [22]
3.3.2 Interplanetary DTNs
Interplanetary DTNs are characterized by very long propagation delay due to long distance
[85], not supported by TCP/IP suite. For example, the round-trip time of a bundle is about
8 and 40 minutes from Earth to Mars, depending on the orbital positions of the planets
[82]. The connectivity may be periodical/predictable (scheduled) using the position and
21
speed.
3.3.3 Intermittently Connected Mobile ad hoc Networks (ICMAN)
In ICMAN, the mobile nodes communicating via limited radio range, experience frequent
partitions and the end-to-end route may not be present at the time of message generation.
Their difference from the message ferries is that each node in ICMAN may acts as a router,
source or receiver and message carriers are not fixed. The connectivity among the mobile
nodes is often opportunistic which clearly distinguishes them from periodical/predictable
nature of interplanetary networks.
3.4 Anycasting in DTNs
To deliver a message to any one member among a group, anycast service is used. It
delivers the packets to at least one and in most cases only one member in an anycast
group. For example several servers provide the same service, in this case a client wants to
get service from any available nearest server. It does not care about getting the service
from a specific one. Resource discovery mechanism is a building block for many distributed
systems. Anycast service may be used in various distributed systems. Similarly anycast
service can be used with DTN nodes for a variety of applications like in disaster rescue
field, finding a device or person without knowing exact IDs and location. To support
these important applications in DTNs, an efficient anycast service is needed. This service
helps to form robust systems and it also make the network configuration easy. Anycast in
DTNs needs special attention due to the unique challenges of delay, storage capacity and
unpredictability. It requires an adaptable model and routing algorithms. Unlike unicast
which has a predefined destination, in anycast the destination may be chosen dynamically
according to the current topological structure. The anycast routing objectives in DTNs
are:
• Maximizing message delivery ratio
22
• Minimizing message latency
3.5 Anycasting Applications in DTNs
There are various applications in DTNs which require anycasting services. Some of them
are as follow:
In Disaster recovery the rescue workers need to give information about the victims, haz-
ards and other related information to one of the server out of a set of servers.
In military deployment there is need to deliver fast and reliably information from a mobile
field commander to a nearest mobile command center or another mobile field commander
out of many mobile command centers or commanders.
In DTN Geocasting (Delivering of message to all hosts in a geographical location) the effi-
cient approach will be that first use the anycast and then broadcast within that particular
geographical region.
Since ICMANs have limited resources such as link bandwidth, storage capacity and con-
nectivity among the nodes; an efficient anycast service is necessary for supporting the
aforementioned applications.
3.6 Problem Statement
After conducting the detailed literature survey it has been observed that in the existing
literature the anycast nodes are static and fixed. Message carriers are fixed and do not
generate messages themself. Moving pattern of the carriers is known in advance. These
constraints are very crucial for a class of DTNs known as intermittently connected mobile
ad hoc networks (ICMAN).
Moreover the existing forwarding mechanisms forward the bundle to next hop purely
based upon the path length. In tie cases the forwarding decision is made randomly or
based upon the node ID (e.g lower node ID wins). In the DTNs forwarding mechanism is
of great importance because in emergency situations message drop is very crucial.
23
3.7 Proposed Adaptive Anycast Algorithm
According to the classification proposed in this research, the anycast model has been pro-
posed for intermittent connected mobile ad hoc networks (ICMANs). In the proposed
model the mobile nodes are sparsely distributed and communicating via short wireless
radio range. The nodes experience frequent and long duration partitions of the network.
Nodes move with limited storage capacity and connectivity among the nodes is opportunis-
tic. In contrast to the existing mechanisms every node may be a source, receiver or the
carrier which carries the message and forwards the message when the opportunity arises.
Situation awareness mechanism made it adaptive to the current topological changes.
A novel packet forwarding mechanism named Received Based Forwarding (RBF) has been
proposed, which improves the performance of anycasting when compared to the widely
adopted shortest path forwarding (SPF). Anycast model of DTNs is given below.
Anycast Model: Anycast in DTNs is defined as bundle transmission to any of the mem-
bers in a group of DTN nodes. Every DTN node has a name and the translation between
DTN name and underlying network is done by DTN routing agent. This bundle layer
is responsible to buffer the bundles, routing at bundle layer and other functionalities at
bundle layer. It operates above the transport layer and below the application layer of the
network. Below the transport layer, protocol of own choice can be used depending on the
network environment while a single DTN bundle layer is used across all the DTN regions,
as shown in figure 3.2 [86]. Anycast source uses the explicit name of the anycast receiver
as the destination address. Extensive simulations have been conducted to check the effect
of buffer size, effect of link availability percentage and group size on the proposed anycast
mechanism. The components of Adaptive Anycast with RBF are given below:
3.7.1 Situation Awareness
The underlying network is operational for a long time so that caches are built up and
sufficient knowledge about the network topological condition is available. This is because
that the network is opportunistic and there are interest based contacts. So the sufficient
24
Figure 3.2: Bundle Layer (DTN Layer)
knowledge about the interest based contacts is built. To achieve situational awareness, each
DTN routing agent periodically sends request to underlying routing agent (responsible to
perform appropriate routing in underlying networks) for the current network condition,
Dynamic Source Routing (DSR) [40] has been used as an underlying routing protocol.
The underlying routing protocol responds to the DTN routing agent with the current
topological condition of the underlying network, includes the paths from current node to
the intended destinations, the high level system model shown in figure 3.3.
3.7.2 Group Formation
Any DTN node that wants to join anycast group sends a join request (JRqt). The join
request is flooded so that it can reach the anycast source as early as possible. Each
DTN anycast source updates its members list on receiving the join request. The group
joining model proposed in [29] has been used. Members per unit area (group size) plays
25
Figure 3.3: High Level System Model
an important role in DTN anycasting. The group size may varies in different scenarios,
like in PDAs (Personal Digital Assistants) networks the anycast group size can be larger
such as the people having PDAs or cellular phones and want to share the audio/video or
other files within a cluster. In military battle field /sensor network, the anycast group
may comprise a small number of nodes. The effect of group size has been analyzed on the
proposed scheme by varying the group size [60].
3.7.3 Message Buffering
In the DTN scenario considered for this research, each DTN node has a limited buffer
to store the in-transit packets. The packet is forwarded to the next hop and is removed
from the buffer on receiving the acknowledgment from the next hop i.e., the custodian
transmission is enabled to ensure the reliable delivery between two hops. When the buffer
is about to overflow, the DTN node sends the buffer information to the sender for flow
control. In different networks the buffer size varies for example in wireless sensor networks
the nodes have a very limited buffer while in vehicle ad hoc networks the nodes have a
larger storage capacity. The proposed scheme has been tested by varying the buffer size.
26
3.7.4 Bundle Forwarding
Each anycast bundle has the explicit list of receivers. So each DTN node forwards the
packet according to its local knowledge about the underlying network provided by under-
lying networking protocol.
3.7.5 Receiver Based Forwarding (RBF)
A novel forwarding scheme named Receivers Based Forwarding (RBF) has been proposed
for anycasting. When a node wants to forward the bundle, it gets the current topological
information from the underlying routing protocol. The DTN routing agent sends a route
request to the underlying routing protocol (RoutRqt); the route request contains the any-
cast receiver list. The underlying routing protocol sends the discovered paths information
(from the current node to the receivers in the list) to the DTN agent. The DTN routing
agent removes the currently unavailable outgoing links. The DTN routing agent finds
the path to the closest anycast member. In a situation, when there is more than one
such paths available to the anycast receivers, earlier approaches, took the first discovered
shortest path [33], [32]; we refer to this scheme as the Shortest Path Forwarding scheme.
In contrast, this resarch proposed RBF scheme, which works in the following way.
1. For each link find the number of anycast receivers accessible through it.
2. Find the path length to the nearest anycast receiver accessible through this link, it
has been named as the minimum path length.
3. Choose the next hop which has the shortest minimum path and through which
largest number of receivers are reachable.
The flow diagram of RBF is shown in figure 3.4. Authors proved that the probability of
the bundle to reach any member of the anycast group has been increased. This assertion
has been supported by extensive simulation which shows an increase in the delivery ratio
and decrease in the end-to-end delay. Thus, the proposed scheme has achieved higher
overall efficiency.
27
Figure 3.4: Flow diagram of RBF
3.7.6 Pseudo code
The pseudo code of the RBF is as follow:
// DTN routing agent is working above the transport layer which requests the underlying
//routing agent for path information to the anycast receivers.
DTN routing agent
Do
{
Paths Info = handler underlying routing agent (list of anycast recievers)
If (Paths Info)
28
{
//Choose the paths with least number of hops
Shortest Paths = Select Shortest Paths (Paths Info)
//The Select Shortest Paths will return list of shortest paths
If (Shortest Paths >= 1)
{Forward the bundles to the next hop of the shortest path through which more receivers are reachable.}
}
}Until(NoPath Info)
// the underlying network is operational and the underlying routing agent will give
the paths to the DTN routing agent.
Working of RBF is depicted in Figure 3.5. Node 0 is a source node and node 5, 7 and
9 are members of anycast group. When node 0 wants to send anycast bundle to any of the
anycast members then node 0 will send request message to underlying routing agent in
order to get the network condition (paths) to anycast receivers. Now node 0 will analyze
the discovered paths to anycast members as shown in figure 3.5(a). The dashed lines
show available links. As shown in figure 3.5(b), node 0 will delete all those outgoing links
that are currently not available. The node 0 has two shortest paths of equal length so
it has two choices to forward the bundle, either to node 3 or to node 1. The Shortest
Path Forwarding algorithm solves the tie situation by forwarding the bundle to the first
one or randomly. But RBF selects the node through which more anycast receivers can be
reached, so the probability of available receivers will increase. Therefore, node 0 forwards
the bundle to node 1 as shown in figure 3.5(c). Along the message, the list of anycast
receivers is also included. Similar operation will be done at each DTN node for each
bundle. Node 1 gets paths to node 5 and node 7 and then forwards the bundle to node 4
as shown in figure 3.5(d). Node 4 finds the path to node 5 so it will forward the bundle
to node 5 as shown in figure 3.5(e).
29
Figure 3.5: Working of Receiver Based Forwarding (RBF)
3.7.7 Scenario where SPF fails and RBF still works
There are various situations where SPF fails to operate and RBF still works. For example
refere to the scenario shown in figure 3.5, according to SPF node 0 will have to forward
the bundle to node 3 for intended receiver node 9. If link between node 6 and node 9
breaks then all packets forwarded to node 3 will be dropped at node 6. According to the
basic property of DTNs, the partitioning of the network may be for indefinite time, so
the receiver 9 may not be reached forever. On the other hand, according to RBF, if node
0 forwards the bundle to node 1 from which more than one receivers are reachable. In
this case if the link between node 4 and node 7 breaks with the same probability then the
bundle can reach to an alternate receiver. The situation became more worse when the
data transfer contains one important bundle and it is dropped due to non-availability of
receiver.
30
3.7.8 Bundle retransmission
DTN node checks periodically its local buffer for unacknowledged packets, provided its
existence, they are retransmitted according to anycast receiver list. In order to reduce the
overhead there are controlled retransmissions for a threshold (e.g we have threshold of 5).
Once the acknowledgment is received, the packet is deleted from the buffer.
3.8 Performance Evaluation
For performance evaluation the adaptive anycasting in DTNs has been implemented and
analyzed with both SPF and RBF using network simulator-2 (NS-2). NS-2 is a non-
commercial, publicly available, open source, object-oriented simulator written in C++
with an OTcl interpreter as a front-end. NS-2 consists of several components which help
to obtain a more complete simulation in terms of programming flexibility, visualisation,
and understanding. It is the most widely used simulator in the world for the networking
research [87]. Following metrics have been set as the performance parameters:
1. Message Delivery Ratio: It is defined as the ratio of total number of unique anycast
bundles received by any anycast group member to the total number of bundles
transmitted by the anycast source.
2. Data Efficiency: It is defined as the ratio of total number of unique anycast bundles
received by any group member to the total traffic generated, both data bundles and
control packets.
3. Average Message Delay: It is defined as the ratio of total delay for received anycast
bundles to the total number of anycast bundles.
3.8.1 Simulation Parameters
A specific type of DTN known as ICMANs has been considered for simulation. Simulation
has used 20 nodes over 800 x 800 meter area with 110 m transmission range. DSR is used
as routing approach for underlying routing in network. For situation awareness, DTN
31
routing agent collaborates with the underlying routing agent. MAC layer is IEEE 802.11.
One anycast session has been carried out for 60 seconds. Node 1 is fixed as the anycast
source. While the other DTN nodes randomly join the anycast group by sending join
request. The anycast source sends the bundle at the rate of 1 bundle per second. At every
2 second each DTN agent checks its buffer to see bundle waiting for retransmission. The
performance of the proposed scheme has been tested with respect to:
• Varying link availability
• Varying group size
• Varying buffer size
3.8.2 Varying link availability
The performance has been observed by varying the link availability of each link uniformly
for 10% to 90%. The buffer capacity of each DTN node is 30 bundles. Figure 3.6 shows
the packet delivery ratio (PDR) of adaptive anycast algorithm with both SPF and RBF
schemes. Varying the link availability from 10% to 90% PDR has been recorded for Data
Packets (DP), PDR with DP and control packets (CP) has been recorded separately.
When the availability is low, RBF achieves better delivery ratio because, the probability
of receivers’ availability increases. The delivery ratio of both schemes SPF and RBF are
almost equal at 90% or more link availability.
Figure 3.7 shows that when link availability is low the overall data efficiency of RBF is
high because the probability of number of receivers is high and if one receiver is not avail-
able, the bundle may be forwarded to another receiver, and no extra traffic is generated
for retransmission and checking the underlying network condition. Here the PDR contains
both data packets and control packets. At high link availability the difference between
SPF and RBF data efficiency decreases.
Figure 3.8 shows the average end-to-end delay performance using both SPF and RBF.
At low link availability the average end-to-end delay of both schemes abruptly increases
due to the long duration partitions, which causes packets to be buffered for longer time.
32
Similarly it has been noted that RBF experiences comparatively low delay due to the al-
ternate receivers’ availability and that the propagation delay is less than the long duration
partition delay. At high link availability the average end-to-end delay of both schemes is
minimal. Above the 90% link availability the average end-to- end delay of both schemes
is almost equal.
Figure 3.6: Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
Figure 3.7: Overall efficiency of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
33
Figure 3.8: Average end-to-end delay of Shortest Path Forwarding (SPF) and ReceiverBased Forwarding (RBF)
3.8.3 Varying group size
The effect of group size has been observed on the proposed scheme.
Figure 3.9 shows the average delivery ratio of adaptive anycast algorithm with both SPF
and RBF schemes for various group sizes. Varying the group size from 3 to 15 nodes, the
delivery ratio increases. When the group size is small, the RBF achieves better delivery
ratio as compared SPF because the probability of receivers’ availability increases. The
difference between delivery ratio of SPF and RBF decreases as the group size increases.
Figure 3.10 shows that Overall efficiency of Shortest Path Forwarding (SPF) is lower than
the Receiver Based Forwarding (RBF) for all group sizes; however, the difference grows
smaller for larger group sizes.
3.8.4 Varying buffer size
The size of buffer plays an important role in the disruption tolerant network. The perfor-
mance of the proposed scheme has been observed by varying the buffer size.
Figure 3.11 shows that when buffer size is limited, the RBF performance is better than
SPF. This is because of increased probability of anycast receivers’ availability. However
when buffer size grows larger the difference between RBF and SPF performance decreases.
34
Figure 3.9: Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
Figure 3.10: Overall efficiency of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
This is because the nodes can hold the data packets for longer time.
Figure 3.12 shows that overall efficiency of RBF and SPF decreases due to retransmis-
sion of dropped/lost packets. It is also shown that the overall efficiency of RBF and SPF
is slightly different on average by varying buffer size. There is no significant difference
between the performances of RBF and SPF by varying the buffer size. It is due to the
implementation of custodian transfer, reliable hop by hop transfer which introduces more
traffic while the packets reside in the buffer.
Figure 3.13 shows the average end-to-end delay of RBF and SPF by varying the buffer
size. It is shown that increasing buffer size leads to increase in the end-to-end delay. This
35
is because the packets are buffered for longer time. However it increases the delivery ratio.
Figure 3.11: Packet delivery ratio of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
Figure 3.12: Overall efficiency of Shortest Path Forwarding (SPF) and Receiver BasedForwarding (RBF)
36
Figure 3.13: Average end-to-end delay of Shortest Path Forwarding (SPF) and ReceiverBased Forwarding (RBF)
3.9 Conclusion
In this chapter classification of delay/disruption tolerant networks (DTN) has been pro-
posed by identifying their distinguishing characteristics. Firstly, DTNs with special nodes
working as message ferries DTNs; secondly, DTNs with periodic/scheduled connectivity
and finally, the intermittently connected Mobile ad hoc networks. Each class has unique
routing/buffering requirements of its own. Further, after presenting detailed discussion
of anycasting schemes proposed in literature for the intermittently connected Mobile ad
hoc networks, a new anycast scenario with Receiver Based Forwarding (RBF) scheme has
been proposed. RBF considers the number of anycast receivers accessible through a link as
well as the path length to the nearest receiver through that link in forwarding the anycast
bundle to the next hop. Extensive ns-2 simulation results with respect to link availability,
group size and buffer size show that the RBF performs better than SPF in terms of data
delivery ratio, average end-to-end delay and overall data efficiency.
37
Chapter 4
ANYCAST ROUTING INWIRELESS MESH NETWORKS
In this chapter anycast routing in wireless mesh networks (WMNs) has been investigated.
Chapter starts with the introduction of WMNs, and then its various architectures have
been presented. WMNs applications and WMNs advantages are described in section 4.3
and section 4.4 respectively. Various routing algorithms including anycast routing have
been presented in section 4.5. After that the problem statement in the existing anycast
routing has been presented in section 4.6. Proposed solutions which includes proposed
network model, anycasting, geocasting, load balancing, flow chart, algorithm and simula-
tion results have been demonstrated in section 4.7.
During the research it has been observed that the field base routing (FBR) is prone to
various security attacks. Solutions to protect the FBR from external intruders and inter-
nal selfish nodes, along with the flowchart and simulation results have been presented in
the section 4.8. Finally section 4.9 concludes the chapter.
4.1 Introduction
A self healing wireless network which is built through a number of distributed and re-
dundant nodes to support variety of applications and provide reliability is known as mesh
network. The basic aim of the wireless mesh networks (WMNs) is the guaranteed connec-
tivity. Wireless mesh networks are gaining popularity for its wide range of applications.
38
Wireless mesh networks have gained substantial consideration as an unconventional solu-
tion to applications such as community networks, enterprise networks, and last mile access
networks to the Internet [1]. There are several factors, which affect the throughput ca-
pacity in WMNs, such as characteristics of MAC protocol, hidden/exposed node problem
and broadcasting problem [88].
4.2 Wireless Mesh Network Architecture
4.2.1 Basic Mesh Architecture
The basic architecture of mesh network is proposed by [1] is shown in figure 4.1. In contrast
to ad hoc networks, wireless mesh networks do not impose the strict infrastructureless
property. The basic architecture of the wireless mesh network is shown in Figure 4.1[1]. It
has some fixed nodes (Gateways) which provide access to the Internet acting as backbone;
other nodes may be fixed access points which provide connectivity to the wireless mobile
nodes in multihop fashion. Every node in the network acts as router.
Figure 4.1: Basic mesh architecture [1]
39
4.2.2 802.16 based WMNs
802.16 mesh networks are primarily used to provide the access for distribted and sparsely
connected areas shown in figure 4.2. This internet access to the remote areas is cost
effective. Normally the network topology is a tree based with a root at the base station.
In WMNs most of the nodes are normally stationary or move occasionally. Normally the
nodes are well equipped and power is not as issue in these networks, so the concentration of
various routing algorithms is on improving the performance of individual transfers which
enhances the overall network capacity, rather than handling the mobility or reducing the
power usage. Reduction of the traffic resulted from the interfered transmission is one of
the main problems.
Figure 4.2: IEEE 802.16 based wireless mesh networks [1]
4.2.3 Infrastructure/Backbone WMNs
Infrastructure/Backbone WMNs is a network in which some nodes like the gateways are
connected to the Internet on one side and have a wired or wireless connectivity to the
mesh clients on the other side. The Infrastructure/Backbone WMNs is presented in figure
4.3
40
Figure 4.3: Infrastructure/Backbone based WMN [1]
4.2.4 Client meshing (client WMNs )
It consists of mesh clients which can perform routing and can support various user’s
applications. As shown in figure 4.4, client nodes communicate directly with each other
in the absence of mesh routers.
Figure 4.4: Client WMNs [1]
41
4.2.5 Hybrid WMNs
By combining the client meshing and infrastructure based networks hybrid WMNs ar-
chitecture is formed shown in figure 4.5 . In hybrid WMNs nodes can get access to the
Internet, WiFi, WiMAX, and other networks through the infrastructure and can also
communicate to the other mesh clients directly without mesh routers[1].
Figure 4.5: Hybrid WMNs [1]
4.3 Wireless Mesh Networks Applications
There are many interesting application of WMNs, some of them are highlighted below:
• City wide broadband internet access
• Emergency and disaster situations
• Video on demand and IP-TV
• Wireless backhaul
• Backup network
• Security surveillance system
• Building automation
42
4.4 Wireless Mesh Networks advantages
WMNs offer several benefits in contrast to other technologies. Some benefite are listed
below:
• Increased reliability: as there are more than one paths available from source to the
receiver, thus the WMNs provide communication reliability by eliminating single
point failures and potential bottleneck links.
• Large coverage area: WMNs provide long distance communication with their multi-
hop feature. WMNs provides non line of sight connectivity among the dispersed
users. Furthermore, it exploits other techniques like spatial reuse or multi-channel
communications which allow long distance communications.
• Automatic network connectivity: the basic characteristic of WMNs is that they
are dynamically self-organized and self-configured i.e. the mesh clients and other
nodes like gateways and routers establish the network connections and preserve this
connectivity. The scalability is guaranteed by adding the new routes in case new
nodes join the network. The restructuring process of the network is initiated when
new nodes join the network.
• Low installation costs: the self-organization and self-configuration nature of the
WMNs make them comparatively cheaper than the other technologies. Manual in-
tervention is minimal in these networks. It also provide the cost effective internet
connectivity to the sparsely distributed areas by providing the multihop communi-
cation rather than installation of large number of Wi-Fi Access Points (APs) and
expensive cabling.
4.5 Routing in WMNs
Routing in the wireless mesh networks always attracts the researchers. Irregular environ-
ment of the wireless networks leads to the design challenges of a good routing protocol. In
43
order to get long term route stability and opportunistic performance routing design has
to be modified. Its has to make sure that the technique is robust against the variety of
attacks, soft and hard failures, channel breakdown and various denial of service (DOS)
attacks. These are the few primary challenges in routing over wireless mesh networks
(WMNs). Some other challenges are the topological changes due to mobility, high error
rates and low bandwidth.
4.5.1 Traditional ad hoc routing
Due to similarities between MANETs and WMNs, routing protocols that are designed for
MANETs may be considered for WMNs. MANET is a set of wireless nodes, these nosed
may be mobile which can form peer-to-peer networks dynamically with no centralized
control or wired infrastructure. On the other hand WMNs is a subset of ad hoc network.
These networks have a hybrid architecture i.e. it has some fixed nodes which provides
the internet access and may have some mobile nodes. Due to limited range of wireless
communication, multihop communication became very popular in these networks, where
every node act as router to route the message and at the same time it may be the source
or the destination. That is why routing of packets between any pair of nodes is challenging
and attracts many researchers to focus on it [89]. Figure 4.6 shows the classification of ad
hoc routing protocols. The field base routing has been added in this classification, which
is gaining popularity for the WMNs.
4.5.2 Field Based routing (FBR)
FBR is getting attention from the research community due to its simplicity and robustness.
It is based upon a routing field which is exchanged among the routing nodes. The nodes
having larger routing field value got a larger chunk of data to forward. In literature the
routing field is described as heat (The value which is computed for every node considering
the gateway as a source of heat) [45]. It is a small value that builds the routing tables.
The FBR approach has been studied in a variety of applications. It is used for anycast
44
Figure 4.6: Ad hoc Routing Protocols hierarchical classification
routing [44]. Some times it is used as a density based routing. In [44] authors show that
FBR is applicable by getting the best performance which is a trade off between proximity
and density. Moreover FBR is used for service discovery in MANETs [90]. The service
providers are considered as the ultimate source and the routing field is established in the
network. The data moves along the path having larger routing field value. It has also been
applied in sensor networks [91] because of its applicability and simplicity. It increases the
life time of the sensor network.
4.5.3 Group communication
Group communication is an important paradigm to study. Considering the citywide mesh
network there might be many groups like the group of educational institutes, industries,
vehicular network clusters etc. Citywide multi hop network which has a fixed infrastructure
on one side in form of gateways and mobile/fixed wireless nodes on the other side. The
gateways have neither the mobility nor the power issues. There might be a series of fixed
points which relay the traffic to the sparsely distributed nodes. Gateways provide the
internet access to the wireless nodes which is the most common application of the wireless
mesh networks. Optimally reaching to a gateway for the internet access is the focus of the
45
contemporary studies.
4.5.4 Anycast routing
Anycast is an important service which always applies the greedy approach to deliver the
packets. It may choose the nearest or first optimal destination. Anycast (one-to-any) is a
scheme of routine which allows the packets to be routed to the nearest or best destination.
Anycast communication is between a single sender and the nearest receiver of the group,
rather than the whole group as in case of multicasting. If there are various groups of
the same category, than the anycast traffic will be forwarded to the next hop towards
the group head having larger calculated parameter (Temperature Field), considering the
group head as the heat source. The major contribution of this study is the proposal of an
anycast model for the traffic from gateway to the mesh nodes considering various anycast
groups. Moreover geocast and unicast communication have also been studied. Various
interesting applications of anycast routing have already been discussed in chapter 1.
4.6 Problem Statement
After conducting the detailed literature survey it has been observed that while considering
the data traffic from gateway to the mesh nodes the idea of anycast group communication
is not present. Currently the traffic moves in unicast fashion. So the main problem is
that the idea of group communication especially the anycast group communication is not
available in the existing literature. This research proposes the concept of anycast group
communication for the traffic moving from gateway to the mesh nodes. Field base routing
has been used for the anycast group communication. Secondly, the concept of intergroup
communication is not available, so that has been made possible with proposed anycast
group communication. Thirdly, the geocast routing has been handled through the unicast
communication in WMNs. This research proposes the anycast communication for reaching
to the group head and broadcast communication inside the group. It is important to note
that this thesis in not concerned with the geographical location of the geocast nodes rather
46
it deals with the delivery of packets using anycast instead of unicast. In addition to the
above mentioned contributions, during the research it has been observed that the anycast
group communication using FBR is prone to various security attacks. This research also
proposes the techniques to mitigate the external intruders and the internal selfish nodes.
4.7 Proposed Solution
4.7.1 Proposed Network Model
A scenario for the anycast routing in wireless mesh networks has been developed in this dis-
sertation. Figure 4.7 shows the architecture of wireless mesh network for anycast routing.
There are three types of nodes in this architecture:
• Gateway: The gateway provides the Internet connectivity to the mesh clients. The
traffic from the gate way to the mesh clients has been considered, so the gateway is
the main source of traffic in this scenario.
• Routing Nodes: These nodes have the capability to route the packet towards its
destinations. The routing is based upon the ”Gradient Based Routing” keeping the
destination as the ultimate source for calculating the routing field.
• Groups: These are different groups having their own group members. Group mem-
bers are registered with the group head. The group heads are registered with the
gateway.
Routing packets from Internet (gateway) to groups has been studied in two forms i.e. 1)
anycasting at the gateways and 2) geocasting.
4.7.2 Routing Packet to Any Group (Anycasting on Gateway)
Following steps are required to route anycast traffic:
Step I:
All the group heads send join messages to Gateway for creation of Group on Gateway.
47
Figure 4.7: Architecture of mesh network for anycast routing
Step II:
Gateway receives the request and accepts/rejects the request message and sends acknowl-
edgment to group. This acknowledgment is sent by using the same mechanism as discussed
in section of field base routing.
Step III:
Once a group is created on gateway and has certain messages to send to any of the group
on the gateway, so it broadcasts the message to all the groups that join the gateway. The
step III is explained with the help of scenario in Figure 4.8
4.7.3 Geocasting
There are two steps involved in geocasting; 1) unicast (in case it is known the packet is
for a specific group) or anycast (in this case the packet may be sent to any nearest group)
48
Figure 4.8: Anycasting on Gateway
and 2) The broadcasting inside the group.
Gateway to Group Head (Unicasting)
To route the unicast packet from gateway to group head using field based routing, first
find out the field value of the destination (Group Head) and then send the message to
that group.
Intra-group Communication (Broadcasting)
When a packet reaches at group head, the next step is to send the packet to every node in a
group. The packet is now broadcasted from head to all the group members and group head
maintains a list of all the group members. The communication up to the burning points
(Group Head in this case) has been done unicast/anycast manner, as shown in Figure 4.9.
49
The group head than broadcast the packet inside the group. This whole process is called
Geocasting. So in short, if there is anycast traffic at the gateway, it will be delivered to
Figure 4.9: Routing Packet from Internet to Group heads
any nearest group head via gradient based routing mechanism. On the other hand, if there
is geocast traffic for any group, it will be route following the gradient based routing till
the group head in unicast fashion and will be broadcasted inside the group. Any group
head or a routing node may be considered for the unicast traffic following the same routing
mechanism. There might be various interest groups intended for various types of traffic,
so intergroup communication has also been made possible. The calculation of the routing
field value is shown in the following equation:
50
Let there are m number of nodes i.e. y1, y2, ..., ym in the network, and there are k number
of possible destinations/receivers i.e. R1, R2, ..., Rk and dmk is the distance between the
node and reciever, then:
Total distance of every node i to the destination where i = 1...m, is:
Di =k∑
i=1
di (4.1)
Field value of every node i is a function f over the distance di where i = 1...m, is:
Wi = f(di) (4.2)
4.7.4 Load balancing
As the forwarding nodes make decision on the basis of field values of its neighbors. If
proper load balancing technique is not followed then the entire traffic will follow the same
route which is not suitable for the energy constrained nodes. The load balancing has been
ensured by putting a threshold on the traffic flow; if the traffic exceeds from the threshold
value the forwarding algorithm will select the next sub-optimal path.
4.7.5 Algorithm
The algorithm for the anycast, unicast and geocast routing following the gradient base
routing technique with load balancing is given below.
Every node in the network calculates its field value considering the field value of its neigh-
bors and then advertize the value to populate the routing table of the neighbors. It is
worthy to note that this field value is used for the routing and the value is calculated at
the routing table building phase. The forwarding is based upon this value but the value is
not calculated at the forwarding time. The packet is forwarded to the next hop having the
highest field’s value until the traffic flow exceeds the threshold, which has been considered
for load balancing. If the traffic flow exceeds the threshold the algorithm selects the next
51
hop with the next highest field’s value. For geocast traffic it carries the data in unicast
fashion till the group head and then broadcast it inside the group.
Algorithm:
Gm- Group member
Tt - Traffic type
Fl- Field value
Nn- Neighbor Node
Th- Threshold value
Nh- Nearest Head
repeat
Initialize Fl of every Nn
Calculate Fl for every Nn
Check the Tt of every message
if Tt = Anycast then
Select the nearest head Nh
if traffic− flow − ofNn ≤ Th then
Select Nn having highest field value.
Forward the message to Nn
else
Select next hop having next highest value.
Forward the message to the next hop
end if
if Tt = Unicast then
if traffic− flow − ofNn ≤ Th then
Select Nn having highest field value.
Forward the message to Nn
else
Select next hop having next highest value.
52
Forward the message to the next hop
end if
if Tt = Geocast then
if traffic− flow − ofNn ≤ Th then
Select Nn having highest field value.
Forward the message to Nn
else
Select next hop having next highest value.
Forward the message to the next hop
end if
if Node = GroupHead then
Broadcast the message to Gm
end if
until Destination Reached
4.7.6 Flow chart
The flow of data is depicted in Figure 4.10
4.7.7 Simulation results
For the performance analysis OMNet++ [92] simulator has been used. The selection
of OMNet++ is due to its better graphical user interface. As the test scenario is
static and thorough analysis of the delivered packets and modeling of the anycast
groups were required, so the OMNet++ is selected instead of NS-2 for this module
(chapter). A scenario has been developed in the OMNet++ simulator to simulate
the anycast, unicast and geocast traffic. The scenario contains twenty four nodes
which include one gateway, which relays the data from and to the Internet and
three anycast groups represented by the group heads. In this research, the traffic
from the Internet towards the mesh clients has been analyzed, so the gateway is
53
Figure 4.10: Flow chart of anycast, unicast and geocast routing with load balancing
the ultimate source of the traffic which relays the Internet traffic towards the mesh
clients and groups.
4.7.7.1 Routing delay
Figure 4.11 depicts the comparison of packet delays experienced by unicast and
anycast traffic. As the idea of anycast group communication is proposed for the
54
first time considering the traffic from gateway towards the mesh clients. In this
particular analysis it has been demonstrated that introducing the concept anycast is
always better than the unicast. Unicast traffic is carried out between two particular
nodes, in anycast the packets are forwarded to the nearest (optimal in terms of
cost) node. The delays have been analyzed which is experienced by unicast and
anycast type traffic, because before the idea of anycasting the traffic flows were of
the unicast type. The results show that with the increase of the volume of data
to be delivered, the delay experienced by unicast traffic is considerably high than
the anycast traffic. This is because the anycast traffic always chooses the best
possible path and delivers the packet to the nearest group head. Although the load
balancing mechanism has been implemented, explained in section 4.7.4.
Figure 4.11: Packet Delay comparison of Anycast and Unicast traffic
4.7.7.2 Packet delivery
The packet delivery is always an important parameter to study. Figure 4.12 shows
the comparison of the number of packets delivered between the unicast and any-
cast type of traffic. Packets may be dropped uniformly in routing due to buffer
overflow and the intermediate devices, link failure or any unseen problem. It has
55
been observed that anycast type of traffic always gives better packet delivery as
compared to the unicast traffic. This is because that the unicast traffic has to
follow a predefined path towards a predefined destination. Aycast communication
is adaptable and selects the best path while forwarding the packet to the next hop
towards ’any’ optimal destination.
Figure 4.12: Packet delivery comparison of Anycast and Unicast traffic
4.7.7.3 Anycast based Geocasting
Geocasting is the phenomenon of delivering the data packets to a particular geo-
graphical location. The entire group members have been considered to be the part
of a particular geographical location. Now one way is to deliver the data packets to
the group members in unicast fashion, named as unicast based geocasting (UG). In
UG every group member has a distinct connection and receives the packets directly
from the source through the intermediate nodes. The other way to deliver a packet
to the group head (which is possible in the proposed group communication) and
then broadcasts it within the group. It has been named as anycast based geocast-
ing (AG). AG provides the luxury to select the group head based on its optimality
(nearer). A single packet has to be sent to the group head, which broadcasts the
packet among the group members through its own broadcasting mechanism. The
figure 4.13 portrays the packet delay analysis experienced by UG and AG. The
56
traffic delay in case of UG is directly proportional to the number of group members
and in case of AG, the delay is to reach to the group head plus the delay involved
in the broadcast phase. The intra-group broadcasting delay has been considered in
which the group members are normally in the direct range of the group head.
Figure 4.13: Packet Delay comparison of Anycast and Unicast base geocasting
4.8 Securing the field based routing
During the research work it has been observed that anycast based upon the field
based routing is prone to various security attacks. In addition to the anycast
routing, security issues related to FBR have also been investigated in this research.
Securing FBR is an additional contribution of this reaerch study. Remaining part
of this chapter is related to the security of FBR in WMNs.
Three types of traffic (i.e. any cast, unicast and geocast) have been analyzed for
possible attacks. To route the packet in the network, field based routing is used as
it considers a small value to exchange information between nodes. This feature of
FBR can easily be exploited by any intruder if it advertises its routing field value
as maximum. By doing this it can divert all the traffic and can launch active and
passive attacks. The scenario is depicted in Figure 4.14 with two intruders.
As every node forwards the packet only to the neighbor whose field value is highest
57
Figure 4.14: Developed Scenario of Mesh Network with Intruder
among the neighbors, and if any of the neighbors is an intruder and advertises
its field value to maximum, all the traffic routes towards that node. As shown in
the above scenario, there are two intruders having the highest field value, so all the
traffic always passes towards these nodes and hence dropped or changed. To remove
this issue, a secure field based routing approach is proposed, that authenticates each
node before forwarding the packet.
4.8.1 Mitigating the external intruders
4.8.1.1 Flow chart
Flow chart of mitigating the external intruder is shown in figure 4.15.
4.8.1.2 Algorithm
The algorithm for the said approach is presented here. It describes the initialization
and calculation of the routing field value for every node. The node considers its
neighbor field value for calculating its own field value. It then advertize it; which
is now base for the routing. The nodes have to be authenticated before getting the
58
Figure 4.15: Flow chart of mitigating the external intruder
packets. It is because the intruder may advertise wrong and highest routing field
value to divert the traffic.
Algorithm
ht - Field table
fi - Neighbor node
fp - Parent node
hl - Field level
59
do
au(fi)← check node authenticity
if (au(fi) = NO)
remove fi from the neighbor list
else
calculate hl level
exchange field level info with fi, fp
update ht(fi)
calculate the hl, fi, fp
endif
forward message
reached destination
until (destination reached)
4.8.1.3 Simulation results
In this section the simulation results are presented which reflect the effect of intrud-
ers in the WMNs. The results are based upon the placement of intruders at various
levels. The results are gathered while varying the number of intruders at the upper
and lower levels. If the intruder is capable to place itself near to gateway then it
will get larger amount of data. At lower level the amount of captured data is low.
The simulation is tested by sending 200 data packets and it is worthy to note that
if the number of intruder is high i.e. 5 and they are able to place themselves at
upper layer then it captured up to 61.5% of data packets. Figure 4.16 depicts the
results.
Figure 4.17 shows the comparison of normal routing (NR) with secure field based
routing (SFBR) algorithm. Intruders can launch active and passive attacks. The
packet are dropped by the intruders, according to the scenario implemented only
60
0
20
40
60
80
100
120
140
1 2 3 4 5
Num
ber
of
Pac
ket
s ca
ptu
red
Number of Intruders
Upper Layer
Lower Layer
Figure 4.16: Packets captured by intruders at various levels
25% packets reached to the destination in normal routing and rest of the packets
are dropped due to the intervention of the intruders. As in SFBR intruders are
identified and removed from the forwarding list. In case of intruders alternate path
is selected to rout the packets to the destination in result almost 90% of the traf-
fic reached to the destination. Due to the alternate paths packet may experience
longer delay but ensures its delivery. Some packets in the secure algorithm are also
dropped, it is due to the random link or node failure.
0
20
40
60
80
100
120
140
50 75 100 125 150
Nu
mb
er o
f p
acke
ts d
eliv
ered
Number of packets
NR
SFBR
Figure 4.17: Packet delivery comparison of NR and SFBR
As the SFBR perform authentication and the packets are forwarded only to the
registered nodes. It also follows the alternate rout in case of intruders, that is why
it experienced a bit longer delay than the NR. But it ensure maximum delivery.
61
The results are depicted in 4.18
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
50 75 100 125 150
Del
ay (
µs)
Number of Packets
NRSFBR
Figure 4.18: Packet delay time
While simulating the proposed scheme, the concept of load balancing is also kept in
mind. A threshold is fixed and after exceeding that threshold the data is diverted
to another alternate route. It might cost a larger delay but it distributes the com-
munication load uniformly among all the nodes in the network. The packets may
be dropped due to buffer overflow which can be avoided by having prior information
of the available buffer on the next hop.
Comparison of Various routing protocol’s packet delivery
The performance of SFBR is also compared with other routing protocols like Re-
active hop by hop Routing, Proactive Field Based Routing, Enhanced Proactive
Field Based Routing, Wireless Mesh Gateway Routing and Enhanced Wireless
Mesh Gateway Routing. The SFBR technique shows a better performance with
lower number of packets. As compared to Reactive hop by hop routing, SFBR
shows greater packet delivery. Proactive Field Based Routing shows better packet
delivery as compared to SFBR when the number of packets is high. Due to its min-
imal overhead and protection against the intruders, SFBR is an efficient approach.
Figure 4.19 shows the comparative results.
62
Figure 4.19: Comparison of various routing protocols with respect to Packet Delivery
4.8.2 Mitigating the internal selfish nodes
The SFBR approach is also used for mitigating the internal selfish nodes as well.
The temperature field value is exploited to detect the internal selfish nodes. The
basic idea is detecting the selfish node by comparing its advertized routing field
value with the advertized value of its neighbors. If there is a difference greater
than the set threshold the node is detected to be selfish node. The flowchart and
simulations results for mitigating the selfish node are given below.
4.8.2.1 Flow chart
Figure 4.20 shows the flow of control for mitigatoing the internal selfish nodes.
4.8.2.2 Simulation results
Selfish node detection (SND) has been analyzed against SFBR. The results of both
these approaches show that internal attacks are more powerful and the number
of packets received by the internal intruders is comparatively higher. Figure 4.21
shows the packet delivery analysis of SFBR and selfish node detection approach. As
discussed earlier, SFBR is an external intruder’s detection approach. Selfish node
63
Figure 4.20: Flow chart for mitigatoing the internal selfish nodes
detection approach enhances the security by capturing the intruder node and never
sends the packet towards those nodes. These selfish nodes advertise maximum field
value and thus all the traffic routed towards that selfish node.
Figure 4.22 shows a comparative result of packets delivery between selfish node
detection and SFBR, as selfish node detection approach is more efficient mechanism
to detect the intruders hence achieve relatively higher packet delivery. The delay
time for both the approaches is almost same.
64
0
20
40
60
80
100
120
50 75 100 125 150
Num
ber
of
Pac
ket
s ca
ptu
red
Number of Packets
SND
SFBR
Figure 4.21: Packets captured by inruders
0
20
40
60
80
100
120
50 75 100 125 150
Num
ber
of
Pac
ket
s d
eliv
ered
Number of Packets
SND
SFBR
Figure 4.22: Packet Delivery Comparison
Figure 4.23 shows packet delay comparison between selfish node detection and
SFBR.
In secure routing packet routing packets may follow longer path to avoid the intrud-
ers, so face more delay as compared to other approaches which always the shortest
path to deliver the packet. Figure 4.24 shows the comparison of Proactive Field
Based, Proactive Enhanced Field Based Routing, Secure Field Based Routing and
Selfish Node Detection Approach. This comparison shows that Selfish Node De-
tection Approach shows higher packet delivery when packet count reaches above
100. This approach is better than SFBR due to internal detection of intruders, so
65
Figure 4.23: Packet Delay Comparison
messages pass securely from source to destination and shows larger packet delivery.
Figure 4.24: Comparison of various routing protocols
4.9 Conclusion
Anycast and geocast routing in wireless mesh networks have been modeled, imple-
mented and analyzed in this research. The concept of the anycast routing consid-
66
ering the traffic from gateway towards the mesh clients has been proposed. The
proposed technique has been studied against the unicast traffic by varying the
volume of traffic from gateway towards the anycast groups. The anycast commu-
nication outperforms other type of communication in terms of packet delays and
packet delivery. Intergroup communication is also made possible. Moreover geocast
communication technique has been proposed in which the data travels in unicast
fashion till the group head and then broadcasted inside the group. It lessens the de-
lay time experienced by the geocasting based upon unicasting. Analyzing the field
base routing for internal and external intruders is an additional contribution of this
research. More realistic scenarios which will contain mobile nodes and adaptive
group formation and head selection will be considered in future work.
67
Chapter 5
ANYCAST ROUTING IN
WIRELESS SENSOR
NETWORKS
In this chapter anycast routing in wireless sensor network has been investigated. In-
troduction to the wireless sensor network and anycast routing in WSNs is given in
section 5.1. Related work is given in the section 5.2. Section 5.3 explains the optimal
base station selection technique using anycast routing. Problems in the existing ap-
proach have been described in section 5.4. Energy Aware Anycast Routing (EAA) is
explained in section 5.5. Section 5.6 demonstrates the simulation results and finally
section 5.7 concludes the chapter with future directions.
5.1 Introduction
Since the emergence of wireless sensor networks (WSNs) energy constraint has always
remained an issue which affects its design and operation. In literatures various tech-
niques are available which have been proposed to lessen the energy consumption in
WSNs. It has been explored that there are loop holes in the existing techniques, thus
68
an energy aware anycast routing protocol (EAA) for wireless sensor networks has been
proposed, implemented and analyzed. EAA distributes the sensed data among the
existing nodes in a cost efficient manner to gain maximum network lifetime. Through
extensive simulation in TOSSIM, it is demonstrated that the EAA outperforms the
existing routing techniques in terms of energy consumption which leads to maximum
network lifetime.
WSNs are crowded networks of small, inexpensive sensors. These sensors can collect
and distribute environmental data. WSNs help in monitoring applications. It can
help in remote monitoring of physical environments with improved accuracy. WSNs
have various other interesting applications such as habitat monitoring, environmen-
tal monitoring, military battlefield and to gather sensitive information in unreceptive
locations. It can be used to facilitate the mankind in various real time emergency
situations such as fire, earthquake, flood, roadside accident etc. Since its emergence
the main focus was the design of processor and computing, but the energy constraint
routing techniques still needs improvement because they affect the design and opera-
tions of WSNs. In the recent past many researchers are attracted towards this critical
issues for example Abid et al. [53] studies various protocols for its energy efficiency.
Many other techniques are also available to lessen the consumption of energy in WSNs
for example the technique prposed by Gabow et al.[54]. Sensor nodes have various
energy and computational constraints. Significant research has been focused to over-
come these deficiencies through more energy efficient routing, localization algorithms,
system design, and load balancing [9].
In the literature various authors uses single base station (sink node) and the sensor
nodes gather the data and deliver that data to this single node [55], [56], [57]. In some
studies the authors consider various base stations and the sensed data may be split
into various pieces and deliver it to various base stations [54], [58]. For some types of
data such as multimedia sensing application in real time (surveillance video), the data
is sensed by the dispersed nodes and needs to be delivered to a specific (or single)
69
base station. An optimal base station selection algorithm for anycast routing has been
proposed by Hou et al. [69]. Hou et al. [69] has presented an enhanced model for the
optimization of the network life time. But the technique presented in their research
also invites the researchers to consider it for the most suitable solution for the anycast
routing in WSNs. Based upon the work of [69], this research proposes EAA: Energy
Aware Anycast Routing in Wireless Sensor Networks, which is be explained in the
section 6.3.
5.2 Recent Work
The anycast routing has been studied extensively for Internet and mobile ad hoc
networks [29], [28], [30] and specifically for the delay tolerant networks [36], [59] and
[60]. But the Internet and mobile ad hoc networks scenarios are very different from
the wireless sensor networks. The power awareness is the major concern in WSNs
and it has attracted the research community to focus on.
Rich literature is available for the energy efficiency in medium access control (MAC)
protocols [61] and multicast routing [62], [63], [64]. Unicast and broadcast routing
protocols got the similar attention [65] and [66]. The anycast routing did not get the
desired attention. Abid et al. [53] focuses on the energy efficiency of various protocols
in WSNs but did not study the anycast routing protocol.
The authors in [67] proposed the first anycast protocol for WSNs. The authors pro-
posed the idea of delivering the packets to the nearest sink node, which does not
perform well always. The authors of [68] also proposed another anycast routing tech-
nique. They built the source trees; this technique is similar to the nearest sink node
scheme presented in [67] because both approaches are based upon the selection of
minimum energy path. As stated earlier minimum energy paths does not always
guarantee the maximum network lifetime.
70
5.3 Optimal Base Station Selection using Anycast Rout-
ing
The most recent work on anycast routing is the optimal base station selection algo-
rithm using anycast routing has been proposed by Hou et al. [69].
Figure 5.1: Physical topology of the sensor network.
Figure 5.2: An examples illustrating anycast between the sensor nodes and the basestation.
The physical topology of the sensor network shown in figure 5.1 and the base station
selection using anycast is shown in figure 5.2. The authors have proposed this tech-
71
nique for surveillance video example. In this example it is necessary to forward all bit
streams generated by a forwarding node to the same base station instead of forward-
ing to different base stations, because partial data stream from a video source may
not properly decoded and processed at a base station. In the base station selection
and data forwarding, the proposed technique splits the data stream into sub-flows and
sending them to the same base station through different paths. Hou et al. [69] did
a great job but the general perception while studying the topologies given by them
is that every forwarding node splits the data which leads to many problems affecting
the network lifetime. All issues are stated in the next section.
5.4 Problem Statement
Some flaws in the scheme proposed by Hou et al. [69] have been identified which need
to be addressed to maximize the network life time which is an important parameter
for WSNs. The finding are as below:
• If the same flow is divided into multiple sub-flows and it is routed through
different paths then many sensor nodes should take part in this routing which
will be otherwise in sleep mode. The overall energy consumption should increase
which is very critical for the sensor network.
• Different sub-flows should follow different paths so any change in the topology
causes data loss which leads to the retransmission of the data that increases the
overhead and energy consumption.
• If more nodes are involved for routing in wireless sensor networks there will be
more problems at the MAC layer due to packet collision.
• Header to payload ratio increases for more flows.
• If any segment of the data is lost then the whole packet is dropped at the base
station.
72
• Proper reassembly mechanism should be adopted at the base station.
• No limits on the sub-flows, i.e. how many sub-flows should be there, or it
will divide the flow into sub-flows at every intermediate sensor node which will
involve almost all the sensor nodes in the network which is not energy efficient.
After studying the literature for anycast routing in wireless sensor networks it is ob-
served that existing approach for optimal base station selection using anycast routing
[69] has serious issues of power management, header to payload ratio, data integrity,
collision at MAC layer, playback buffer and jitter. These issues need to be resolved
and hence this research focuses on the presentation of energy aware anycast routing
in WSNs (EAA).
5.5 Proposed EAA: Energy Aware Anycast Routing
The proposed energy aware anycast routing (EAA) minimizes the number of trans-
missions and at the same time distributes the load among the sensor nodes efficiently.
Figure 5.3 depicts the topology of the existing approach (will be called as CASE-I
hereafter) which divides the sensed data on the originator (sensor node which sensed
the data) and at all the intermediate nodes which forwards the data towards the op-
timally selected base station. The topology even becomes more complex if the sensor
nodes are multiple hops away from the base station. The division also grows expo-
nentially with the increased number of intermediate nodes. It is important to note
that the main focus of EAA is to minimize the number of transmissions for its various
benefits cited above rather than the selection of base station. Selection of an optimal
base station using EAA might be an interesting future direction.
The CASE-1 scenario depicted in figure 5.3 show that the sensing nodes are only
three hops away from the base station but the algorithm still leads to the riotous
traffic at the second hop. So there is a need to optimize the division of the data
at every node to save the energy consumption which will solve the problems quoted
73
in the problem statement up to some extent. This will ultimately lead to maximize
the network lifetime. The proposed energy aware anycast (EAA) scenario is depicted
Figure 5.3: Topology of the CASE-I splitting data at every sensor node
in figure 5.4 (will be called as CASE-II hereafter). EAA splits the data into two
segments at the source (sensing node) to utilize the available neighbors for the load
balancing. The forwarding nodes (intermediate nodes) forwards the data according
to their routing strategy. It does not split the data again to prevail over the problems
up to some extent quoted in the problem statment and eventually leads to maximize
the network lifetime. The most important is the header to payload ratio which ex-
ponentially increases and causes more energy consumption. As energy consumption
in propagation in greater than the energy consumption in processing, so this division
at every intermediate node causes problems and leads to minimal network lifetime.
MAC layer collisions lead to the retransmission of the packets which will consume
74
Figure 5.4: Topology of the CASE-II splitting data only at the sensing node
more energy and will minimize the overall network lifetime. Let n be the number of
nodes in the sensor network, a packet p is divided into s segment from the sensing
node (originator) till the base station, h is the header size on each segment then the
total data size and header size are given by the equation 5.1 and 5.2 respectively.
Total data size =
n∑i=1
Si n is the total number of nodes (5.1)
Total header size =
n∑i=1
hi n is the total number of nodes (5.2)
If any packet pi is divided into si segments then the header to payload ratio of CASE-I
and CASE-II will be:
75
Let A be the header to payload ratio of CASE-I and B be header to payload ratio of
CASE-II, then:
CASE-I
A =
∑ni=1 hi∑ni=1 Si
n is the total number of nodes (5.3)
CASE-II
B =
∑mi=1 hi∑mi=1 Si
(5.4)
Where m is the total number of neighbors directly connected to sensing node in EAA,
m has been limited to be at most 2.
From 5.3 and 5.4
A >> B (5.5)
From equation 5.3 and 5.4 it is clear that the header to payload ration for CASE-I
is greater than the CASE-II. As the energy consumed in the propagation is greater
than the energy consumed in processing, so saving data transmitted in the form of
header, definitely, maximize the network lifetime. The argument is proved with the
help of extensive simulations.
Network life time refers to maximum time till any one of them is dead. In order to
maximize the network lifetime distribution of the load is important. However there
should be a proper mechanism to avoid the infinite division of data. EAA distributes
the load among the available paths keeping the tolerable header to payload ratio. As
all the traffic passes through the node 1, 2 in both scenarios of figure 5.3 and 5.4, so
it drains off its energy quickly. The proposed EAA also distributes the data for load
balancing.
76
5.6 Performance Evaluation
5.6.1 Simulator selection
As discussed in [53], there are many simulators available for wireless networks for
example Avrora [93], ATEMU [94], EmStar [95], SWANS [96], QualNet [97], OPNet
[98] and SENSE [99] but tiny operating system simulator (TOSSIM) [100] is chosen
due to its exceptional features. Tiny Operating System (TinyOS) [101] is an open-
source, dedicated event driven operating system and is developed by a consortium
led by the University of California, Berkeley in cooperation with Intel Research. It is
designed for embedded WSNs. It is component based operating system that helps in
minimizing the code and speedy implementation. Due to its compact style of coding
TinyOS exceptionally perform better for energy efficiency. TOSSIM is a discrete
event simulator for TinyOS WSNs. It is the salient feature of TinyOS integrated
environment that same code verified through simulation can be burn in real motes to
deploy a physical real network of wireless sensor nodes. Therefore it is very beneficial
to debug the code before burning in real mote. While using TOSSIM on a PC, users
can examine TinyOS code under consideration using debuggers and other development
tools [53].
5.6.2 Simulation setup and performance evaluation
A scenario with twelve nodes has been developed; any number of nodes can be tested.
The nodes having ID i.e. from 0-11. Node 0 is the base station and it has been
assumed that it is connected with continuous power supply and energy is not the
main concern of the base station. Both the cases i.e. CASE-I and CASE-II have been
tested by developing an application in WSNs [102].
TinyViz [103] offers an extensible graphical user interface for debugging, envisioning,
and interacting with TOSSIM for TinyOS applications. Screenshot of the CASE-I in
TinyViz is shown in figure 5.5.
As in CASE-I the sensed data is split into fragments according to number of reachable
77
Figure 5.5: Snapshot of CASE-I splitting data only at the sensing node in TinyViz
neighbors and then the processes of division of data goes on till it reaches to the base
station, the division of data has been limited only to the sensing nodes (originator) in
CASE-II. The comparison of header to payload ratio is given in the figure 5.6. As in
the scenario depicted in figure 5.3 and 5.4, node 9, 10 and 11 are the originator nodes.
At this stage header to payload ratio is almost equal in both cases because in both
techniques the data is divided into pieces; however the proposed scheme restricted the
division of stream into at most 2 sub streams. As the data travels toward the base
station in CASE-1 the next hop again splits the data into sub streams and header to
payload ratio grows exponentially. So the header to payload ratio at the node 1 and
2 which are directly in the range of the base station is intolerable. While in CASE-II
the division take place only at the originators and the header to payload ratio remains
steady until it reaches to the base station.
The header to payload ratio depicted in figure 5.6 has a great impact on the network
lifetime because a lot of energy is wasted for the transmission of the non valuable
header part. The energy consumption dramatically increases which minimizes the
overall network lifetime. Total energy consumption and node wise energy consumption
78
0
50
100
150
1 2 3 4 5 6 7 8 9 1011
CASE-I
CASE-II
Hea
der
-to
-pay
load
Rat
io
Header-to-payload Ratio of CASE-I & II
Figure 5.6: Header to payload comparison of CASE-I and CASE-II.
comparisons are shown in figure 5.7 and 5.8 respectively. The node wise energy
0
1000
2000
3000
Ener
gy c
onsu
med
(m
j)
Data splitting schemes
CASE-I
CASE-II
Figure 5.7: Total energy consumption in CASE-I and CASE-II.
consumption is depicted in figure 5.8, the variation of energy consumption among the
nodes is obvious depending upon the position of the node in the network. Readers
are referred to the scenarios depicted in figure 5.3 and 5.4. Nodes 9, 10 and 11 are
the sensing nodes in both CASE-I and CASE-II. In CASE-I it divides the sensed data
among the reachable neighbors and in CASE-II it divides it into at most two pieces.
Node 3 and 8 consumes comparatively less energy due to their location. Nodes 1 and
2 which are in the direct range of the base station are the critical nodes for measuring
the network lifetime. CASE-II consumes comparatively less energy because of not
79
dividing the packet at every intermediate node. On average the CASE-I consumes
9.1% more energy at every node as compared to CASE-II. The division of packets at
0
200
400
600
1 2 3 4 5 6 7 8 9 1011
CASE-I
CASE-II
Ener
gy c
on
sum
ed
(m
j)
Nodewise energy consumtion
Figure 5.8: Node wise energy consumption for CASE-I and CASE-II.
every intermediate node is a cause of increasing the header to payload ratio which
has a serious impact on the network lifetime. Network lifetime comparison of both
CASE-I and CASE-II is depicted in figure 5.9. In CASE-I the critical nodes i.e. node
1 and 2 drain off its full energy rapidly. The network lifetime of the CASE-II is
comparatively high than the CASE-I and leads to 40% increase in average.
Figure 5.9: Network lifetime comparison of both CASE-I and CASE-II
80
5.7 Conclusion
Anycast routing protocols have been analyzed for its energy efficiency in wireless sen-
sor networks. EAA has been proposed which utilizes the available network resources
(nodes) by adopting the optimal load balancing mechanism. In contrast to previously
proposed techniques the EAA restricts the division of the data into sub flows on the
intermediate hops. Through extensive simulation in TOSSIM, and by analyzing the
results it has been proved that EAA is more energy efficient than the previous tech-
niques. According to the average results EAA save almost 9.1% energy per node of
the sensor network. It minimizes the header to payload ratio which leads to maxi-
mum network lifetime and 40% improvement has been recorded. Although an effort
has been made to make the anycast routing more energy efficient but there is still a
room to optimize it by adopting the energy saving techniques at the critical nodes.
In addition to EAA, some more measures may be taken at the intermediate hops and
specifically at the nodes which are nearer to base station in the future work.
81
Chapter 6
ANYCAST ROUTING IN
COMPUTATIONAL GRID
In this chapter anycast routing has been investigated in computational grid. In the
literature, backup machine selection for fault tolerance was based upon the multicast.
The idea of anycast has been introduced for backup machine selection. Chapter starts
with the introduction to computational grid. Grid middleware has been described in
section 6.2. Multicast and anycast techniques for backup machine selection have been
presented in section 6.3. Section 6.4 presents the mathematical model of multicast,
anycast and their comparison with respect to delay and probability of backup machine
availability. Section 6.5 demonstrates the performance evaluation and finally section
6.6 concludes the chapter.
6.1 Introduction
Computational grid can perform the computationally intensive jobs in order to uti-
lize the wide spread processing capabilities of volunteer processors. In order to utilize
the wide spread resources, failure options cannot be ignored. The Grid Technologies
have greatly been applied to a number of different sectors especially for flexible and
secure resources/power sharing/aggregation in a coordinated manner. It can be used
82
over the Internet and Intranet for the development of complex applications e.g. col-
laborative scientific simulation, distributed mission training, analysis of elementary
particle physics etc. Grid computing fulfills the requirements of these applications
such as high speed of parallel computing, storage capacity and network bandwidth.
The basic idea behind the grid is not to buy additional resources but to use free cycles
of existing underutilized grid resources. Grid can be classified, according to the types
of shared recourses and their functionalities, into Computational Grid, Data Grid,
Storage Grid, Equipment Grid, knowledge Grid, Interaction Grid [104].
Computational grid represents a distributed and a parallel computing architecture.
It is composed of heterogeneous, geographically distributed resources connected by
unreliable network media. It is used to solve complex problems, in sophisticated and
user friendly manner, which require large amounts of computing resources. A com-
putational grid is a hardware and software infrastructure that provides dependable,
consistent, pervasive, and inexpensive access to high-end computational capabilities
[10],[11]. It is used to increase the performance and reduce the cost of computer
hardware and software in a variety of ways. The components of a Grid may be
heterogeneous and dynamic in nature.
Packet loss is common in geographically distributed and heterogeneous network, so
user assigned jobs are always prone to different type of failures, errors and faults [73].
Fault tolerance is an important service of Grid, which insures the delivery of a service
despite the presence of fault(s). The issue of fault tolerance in grid computing is
more significant than in traditional parallel computing [74] [75][76] because of a wide
range of errors, failures and faults [105] and fragility of the grid environment. Fault
tolerance techniques can be categorized into reactive and proactive ones.
Retry: If a job fails its execution on a machine due to any type of failure, error or
fault, it can be completed by another machine by submitting the job from the start.
Submitting a failed job from start is called retry. This technique may not be suitable
for jobs which require huge computational resources. Nazir et al [77] proposed a
83
proactive approach for scheduling jobs in a computational grid. In this technique,
history is maintained about the grid resources and jobs are scheduled according to
their history.
Checkpointing: Checkpointing [78, 79] is a common and an efficient technique to
save the state of the computation on stable storage spaces periodically. It is used to
resume the job execution from the previous consistent stored state rather than from
the beginning. It increases the application response time and improves the efficiency
of a system. It helps in load balancing by migrating jobs from loaded machines to
less loaded machines. Similarly it helps in fault resiliency by migrating a job from a
faulty machine to a stable machine. It is mainly used for long running jobs, to save
the work to be recomputed from the beginning. The idea proposed in [80] is that of a
multicast technique, which multicasts address of executer machine in order to select
a backup machine. This technique causes the following problems:
• Data loss or out of order delivery will increase unreliability.
• Increases the network traffic delays.
• Most multicast servers do not discriminate any clients.
Therefore it is easy to join a group and watch the data that is being sent to it i.e.
Distributed Denial of Service, D (DOS) attack is possible. The authors used multi-
cast for backup selection of the executor machines by transferring a single packet to
multiple recipients which cause the above mentioned problems. The idea of anycast is
available in [13] but the authors did not analyze the anycast communication against
the multicast. Anycast based backup selection mechanism with Receiver Based For-
warding (RBF) has been proposed in this research. This research also analyzes the
communication cost involved in multicast and anycast scenarios.
The analysis has been done by creating a computational testbed using Alchemi mid-
dleware. The test bed results and the mathematical proof shows that the anycast
communication is far better for the selection of backup machines. Moreover anycast
84
provides not only better efficiency but also enhances the security. A test bed of Grid
has been developed for the analysis of communication cost. Alchemi executors were
installed on different geographically distributed PCs. Alchemi Manager was installed
on a local powerful machine. They were connected through Local Area Network
(LAN) having speed of 100 MB/sec.
6.2 Grid Middleware
The Grid Middleware plays a vital role in a grid infrastructure and is to be deployed
on each machine to make it a part of grid. There are various grid middleware which
provides different functionalities. Globus Toolkit [106] and Alchemi[107][108] are
famous example of middleware among open source middlewares. Grid architecture is
divided into different layers, each perform a specific functionality as depicted in Figure
6.1. The upper layer is Application layer and is user-centric. It is used for a variety
of applications: engineering, science, business etc. Users interact with this layer via
browsers. The Middleware layer brings different elements intelligently. The lower
layer is hardware centric, used to establish the connectivity among grid resources.
Figure 6.1: Grid Architecture Layers
6.2.1 Architecture of Alchemi
Alchemi is .NET based desktop grid framework running on Windows-based environ-
ment and is very user friendly. It is used for flexible, platform independent, high
85
throughput, object-oriented, thread-based applications (fine grained abstraction) and
file based jobs (course grained abstraction). Thread is basic unit of grid application.
Manager, Executor, owner and cross-platform manager are the different components
of Alchemi. Manager: Execution is the responsibility of the manager. Executors
dedicatedly or non-dedicatedly register with manager. It creates threads for a job
and distribute them for execution on available executors. It submits the result of
executed threads to the owner. Executor: Executor executes the grid threads ded-
icatedly or non-dedicatedly. It receives threads from manager and executes them.
Owner: Owners submit jobs to the manager and inquire the status of their jobs.
Figure 6.2: Block Diagram of Alchemi
Alchemi supports a .NET based distributed system which is a collection of indepen-
dent and distributed processing components i.e. nodes connected via LAN/WAN as
depicted in Figure 6.2. Nodes can share their resources among themselves through
centralized authority i.e. Manager. The following steps describes the functionality of
a grid application:
1. Owner(n) will make a request to the manager to solve a problem
2. Manager will create threads and distribute them to all available Executors
3. Executors will send the result of a thread to the manager
86
4. Manager will send the result of all threads to the owner(n)
6.2.2 Programming Model
Alchemi support two types of parallel programming models i.e. a Job Model and a
Thread Model [109]. A brief description is given below:
Job Model: Alchemi supports highly level of abstraction in which the smallest unit
of parallel programming is a Process. It is called course-grained abstraction. In this
approach of parallel programming, grid application deals with files i.e. input files,
output files and executable files(processes). This models of parallel programming is
very complex and is not flexible. This model support to develop legacy tasks using
different languages like C, C++, Java. Cross Platform Manager manages the legacy
tasks using ASP.NET web services.
Thread Model: The primary programming model of Alchemi is a multi-threaded
parallel programming model which is more low level and is called fine-grained abstrac-
tion. In this approach of parallel programming the basic unit is a thread (independent
unit of work). This approach is more powerful, easy to use and flexible.
There are different reasons for the of Alchemi for creating computational grid. Some
of these are given:
• Globus toolkit is Linux-based and Alchemi is windows based. Most of the system
in our research are windows based, so for achieving better results, Alchemi was
preferred.
• It is an open source Middleware, so one can bring changes as required.
6.2.3 Fault Tolerance in Alchemi
Due to heterogeneous and dynamic nature of resources in a computational grid, faults,
errors and failure may be likely to occur. Many middlewares have different level of
fault tolerance but most of them are not fully fault tolerant. Heart beating is the basic
fault tolerance technique of Alchemi. Executor sends heartbeat signals to Manager
87
periodically. When Manager receives these signals, it predicts that executor is alive
and is working. It is not necessary that executor is in a good operational form.
6.3 Multicast and Anycast
6.3.1 Multicast
Figure 6.3: Backup selection using Multicast
Figure 6.3 shows the backup machine selection using a multicast technique. Alchemi
manager distributes the executing threads to the executor 1, 2 and 3. In order to
select a backup executor, the executing machines multicasts the ’backup request’
to the available volunteers i.e. Backup 1, 2 and 3 Alchemi executors. The backup
executors will reply back to sending machines. The executor will select a backup
machine based upon the machine’s configuration. This activity creates a lot of traffic
which includes backup requests and redundant backup replies. The scenario is also
vulnerable to D(DOS) attack in which hacker will spoof the IP of requesting executor
machine and will send it to multiple backup machines in the network, who will reply
back to the victim executor machine and will launch a DDOS attack. As the network
media is not reliable, huge traffic will cause the media failure which will stop future
communication in the computational grid.
88
6.3.2 Anycast
Figure 6.4 shows the backup machine selection using the anycast technique. Alchemi
manager distributes the executing threads to the executor 1, 2 and 3. The Alchemi
manager and executor will coordinate with a separate utility for finding the distance
(number of hops) of the available backup machines. In order to select a backup
executor, the executing machines will select the nearest backup machine e.g. ’Backup
1’ will be a backup machine for ’Executor 1’ rather than ’Backup 2 and 3’ based upon
the distance (number of hops) from ’Executor 1’.
In anycast communication, the executor machines will coordinate with their relevant
backup machines to create redundant multicast requests and replies. The one to
one anycast communication will minimize the chances of a DDOS attack. As far as
the communication cost is concerned, it is substationally decreased in case of anycast
communication. In order to maximize the probability of backup machine’s availability
due to unreliable network media, author’s own proposed receiver based forwarding
(RBF) [18] mechanism has been used. The RBF selects that next hop through which
more anycast receivers are reachable.
Figure 6.4: Backup selection using Anycast
89
6.4 Mathematicl Model
The generalized mathematical model depicting the communication cost for multicast,
anycast and anycast with RBF is given below:
Let P = {p} : set of all nodes in population
N= Number of nodes in P
n = Number of groups available in P
g ={g1,g2,g3,....,gi,....,gn} : set of groups= where i=1,2,3,..n
For any group gi
gi={d1,d2,....,dm} : set of available nodes in group gi, where d1....to ...dm are nodes
in gi
m=total number of nodes in gi
ti= Time required to reach a packed at di
Di = Distance of a node di from source
6.4.1 Multicasting
As all the members will receive the packet so in this case
Delay = Max(ti)
or
Delay= Max{t1,t2,....tm} = tMax (6.1)
6.4.2 Anycasting
In Anycasting any one member of the groups will receive the packet.
Select i : Di is min [send packet to di]
Case 1: select a node randomly then
Delay = Average{t1,t2,....tm} = tAvg (6.2)
90
Case 2: select the nearest node then
Delay = Min{t1,t2,....tm} = tMin (6.3)
6.4.3 Comparison of Anycast and Multicast with respect to its delay
From equation 6.1, 6.2 and 6.3 the following relationship between multicast and any-
cast is obtained:
Min(ti) ≤ Average(ti) ≤ Max(ti) (6.4)
Where equality holds only when
• All nodes in gi are at equal distance from source but even in this case network
traffic is substantially increased
• A group may have only one node
As it clear from equation 6.4 that anycast communication cost will always be less
than the multicast communication.
6.4.4 Anycast with RBF
Receivers Based Forwarding (RBF) considers the number of anycast receivers available
through a link as well as the path length to the nearest receiver through that link, in
deciding the next hop while forwarding an anycast bundle [59].
Let H = {h1,h2,h3,....hL}: set of next hops which satisfy the stated condition
Let Ri = No. of receiver available at hi
Where hi∈ H and i =1,2,3.....L
Select i : Ri is Max [Send the packet to hi∈H]
91
6.4.5 Probability of anycast receiver’s availability without using RBF
The following three cases are possible for the probability of anycast receiver’s availi-
bility without using RBF:
Average case:
R
ΣRi(6.5)
Worst case:
MinR
ΣRi(6.6)
Best case:
MaxR
ΣRi(6.7)
6.4.6 Probability of anycast receiver’s availability with using RBF
The following equation shows the probability of anycast receiver’s availability using
RBF.
MaxR
ΣRi(6.8)
Equation 6.8 shows that it is always the best case of anycasting without using RBF.
In figure 6.5, six executors are connected with a single manager. The weight on edges
Figure 6.5: Scenario Visualization
92
shows the number of hops away from manager. The scenario is developed with central
manager at Riphah International University, Islamabad, Pakistan [110], with various
executers. The specifications are given in table 6.1.
6.5 Performance Evaluation
For performance evaluation Microsoft .Net framework has been used with Alchemi
open source toolkit [107]. Alchemi packages include Alchemi manager and Alchemi
executers with the backend database tools such as MySQL. As the Microsoft has a
major market share so the research focuses on development of the computational grid
using the window based toolkit to utilize the free cycles of widely spread Windows-
based machines. The involved communication cost has been analyzed by choosing the
Pi calculation program [15] for the grid topology shown in figure 6.5. The involved
communication costs have analyzed for the back up node selection in case of failure
while varying the number of hops of the executers. The study shows that as in
the multicast scenario the manager has to contact all the group members for their
willingness and specification so larger communication cost is involved in term of delay,
packet delivery ratio and reliability as compared to anycast. The results are shown in
figure 6.6. The results have been recorded by placing the back-up nodes at variable
distance (in this case number of hops from 1 to 6). Two schemes i.e. backup node
selection using multicast and backup node selection using anycast have been applied.
It has been observed that as the distance of backup node increases the communication
cost of multicast scheme increases exponentially, While the anycast communication
cast remains steady as it select one optimal (nearer) back-up machine among the list
of all available back-up machines, if it fails, than it chooses the next best option.
In contrast if we use multicast than it has to send request to all backup machines
available in that multicast group. So the communication cost using multicast scheme
is much higher than the anycast scheme. Furthermore it is logically concluded that if
the back-up machines are pole apart than the use of multicast scheme will be almost
93
Table 6.1: Nodes participating in the computational grid with their specificationNodes Specification Platform No of hops away from
the manager
Manager P IV, 512 RAM, 3.2GHz Windows XP –
Executer 1 P IV, 512 RAM, 1.73GHz Windows XP 3
Executer 2 P IV, 512 RAM, 2.0GHz Windows XP 4
Executer 3 P IV, 1G RAM, 3.0GHz Windows XP 9
Executer 4 P IV, 1G RAM, 3.0GHz Windows XP 4
Executer 5 P IV, 512 RAM, 2.0GHz Windows XP 8
Executer 6 P IV, 512 RAM, 3.0GHz Windows XP 10
impractical.
Figure 6.6: Cost Analysis using Multicast and Anycast
6.6 Conclusion
This chapter focused on fault tolerant techniques. The main contribution of this
research is the implementation of fault tolerant techniques using anycasting with
modified forwarding mechanism and its analytical analysis for the computational grid
against the multicast technique. The study analyses the communication cost involved
in terms of delay time and its comparison with the previous techniques. It has been
observed that using anycast technique for the back up selection is better than the other
methods due the less communication cost involved. Communication cost analysis
using cross platform and mixed toolkits might be an interesting future work.
94
Chapter 7
CONCLUSIONS AND
FUTURE WORK
7.1 Conclusions
The research presented in this thesis is focused on anycast routing. Other techniques
like multicast and unicast have been studied frequently in literature but anycast ser-
vice did not receive much attention from the research community. Anycast routing
is an important service which is used for various interesting applications in different
networks. In this dissertation anycast routing protocol has been modeled, imple-
mented and analyzed in various networks, namely, Delay/Disruption Tolerant Net-
works (DTNs), Wireless Mesh Networks (WMNs), Computational Grid environment
and Wireless Sensor Networks (WSNs).
The classification of DTNs into three subcategories, namely: Message Ferries Net-
works (MFN), Interplanetary Networks (IPN) and Intermittently Connected Mobile
Ad hoc Networks (ICMAN) have been proposed. A mobile model for anycast routing
in DTNs has been presented. Furthermore, a novel forwarding scheme for ICMANs
called Receivers Based Forwarding (RBF) has been proposed, which considers the
number of anycast receivers available through a link as well as the path length to
95
the nearest receiver through that link in deciding the next hop while forwarding an
anycast bundle. Comparative performance has been analyzed with shortest path for-
warding scheme through simulation.
An anycast routing model for WMNs has been presented. It considers the traffic
from the gateway to the mesh clients having multiple anycast groups. Geocast traf-
fic in which the packets reach to the group head via unicast traffic and then are
broadcasted inside the group has also been studied in this dissertation. Moreover,
intergroup communication between different anycast groups has been modeled, im-
plemented and analyzed. The network is modeled, simulated and analyzed for its
various parameters using OmNet++ simulator. In addition to above contribution
for WMNs, the presented protocol has been analyzed for various security attacks
including external intruder attacks and selfish node attacks. Countermeasures for
eliminating these intruders have been proposed and analyzed.
Anycast routing has been surveyed for energy issues in WSNs. Previously, numerous
other techniques have been proposed to lessen the consumption of energy in WSNs.
Problems in previous techniques have been identified and hence the dissertation pro-
posed an energy aware anycast routing protocol (EAA) for WSNs. EAA smartly
distributes the sensed data among the existing nodes in a cost efficient manner to
gain maximum network lifetime. Extensive simulation and comparison with state of
the art technique proposed by Hoe et al. [69] has been carried out using tiny operat-
ing system simulator (TOSSIM). It has been observed that EAA is a better solution
for minimizing the energy consumption which leads to maximum network lifetime.
According to the average results EAA save almost 9.1% energy at every node of the
sensor network. It minimizes the header to payload ration which leads to maximum
network lifetime and 40% improvement has been recorded.
Regarding the computational grid, detailed implementation of fault tolerant tech-
niques using anycast has been carried out in this dissertation. With reference to
computational grid, the main contribution of this dissertation is the implementation
96
of fault tolerant techniques using anycasting with modified forwarding mechanism
and its analytical analysis for the computational grid. The study has investigated
the communication cost involved in the proposed scheme and its comparison with the
previous techniques. Analytical analysis of proposed and previous technique which
was based upon multicast has also been presented. The implementation of the com-
putational grid has been carried out in a real testbed based on Alchemi toolkit.
7.2 Future Directions
Effort has been made for modeling and analyzing the anycast routing protocol in
various networks but still there is a room for more investigations. The thesis therefore
recommends the following future directions:
• Regarding the DTNs, performance of the presented anycast model can be ana-
lyzed with more realistic scenarios. There are still security issues which are not
covered in this dissertation; secure anycast routing in DTNs might be a good
future direction.
• Concerning the WMNs, although an effort has been made to secure the pro-
posed technique yet it can be further investigated with standard authentication
mechanisms. The anycast routing may further be analyzed for more realistic
set-up with random moving patterns.
• About the WSNs, an effort has been made to make the anycast routing more
energy efficient by presenting the EAA, yet it can be further optimized by
adopting the energy saving techniques at the critical nodes. EAA can be coupled
with other energy efficient techniques at the intermediate hops and specifically
at the nodes which are nearer to the base stations.
• Relating to the computational grid, communication cost analysis using cross
platform and mixed toolkits might be an interesting future work. The cost
analysis has been done in this study with one manger and six executers. It can
97
be further extended with more mangers and executers spread over poles apart
geographical locations.
98
Bibliography
[1] X. Wang F. Akyildiz and W. Wang. Wireless mesh networks: a survey. vol-
ume 47, pages 445–487. Computer Networks, 15 March, 2005.
[2] Judith Masthoff Wendy Moncur, Ehud Reiter and Alex Carmichael. Model-
ing the socially intelligent communication of health information to a patients
personal social network. IEEE Transactions on Information Technology in
Biomedicine, 14(2):319–325, March 2010.
[3] Vikram Kaul Sunil Samtani Jaewon Kang, John Sucec and Mariusz A. Fecko.
Robust pim-sm multicasting using anycast routing in wireless ad hoc networks.
IEEE International Conference on Communications, ICC ’09., 2009.
[4] Yuan zheiig Jianxin Wang and Weijia Jia. An aodv based anycast protocol in
mobile ad hoc network. PACRIM IEEE, 2003.
[5] Ederson R. Silva and Paulo R. Guardieiro. An efficient genetic algorithm for
anycast routing in delay/disruption tolerant networks. IEEE Communication
Letters, 14(4):315–317, April 2010.
[6] R. Fujimoto H. Wu and G. Riley. Analytical models for data dissemination in
vehicle-to- vehicle networks. IEEE VTC, 2004.
[7] Ness B. Shroff IEEE Joohwan Kim, Xiaojun Lin and Prasun Sinha. Minimiz-
ing delay and maximizing lifetime for wireless sensor networks with anycast.
IEEE/ACM Transactions on Networking, 18(2):515–528, April 2010.
99
[8] Shuhui Yang Mihaela Cardei and Senior Jie Wu. Algorithms for fault-tolerant
topology in heterogeneous wireless sensor networks. IEEE Transactions on
Parallel and Distributed Systems, 19(4):545–558, January 2008.
[9] Vijay Anand Archana Bharathidasan and Sai Ponduru. Sensor networks: An
overview. University of California, Davis, CA 95616, 2007.
[10] Ian Foster and Carl Kesselman(eds). The grid: Blueprint for a new computing
infrastructure. Morgan Kaufman, 1999.
[11] Ian Foster, Carl Kesselman, and Steven Tuecke. The anatomy of the grid:
Enabling scalable virtual organizations. International J. Supercomputer Appli-
cations, 2001.
[12] L. Liu X. Hong J. Wu J. Lin. Optical grid synergy with peer-to-peer. IET
Communications, 3(3):487–499, January 2009.
[13] Muhammad Imran, Iftikhar Azim Niaz, Sajjad Haider, Naveed Hussain, and
M. A. Ansari. Towards optimal fault tolerant scheduling in computational grid.
IEEE-ICET, 2007.
[14] Dujeong Lee Sangsu Jung, Malaz Kserawi and June-Koo Kevin Rhee. Dis-
tributed potential field based routing and autonomous load balancing for wire-
less mesh networks. IEEE Communication Letters, 13(6):429–431, June 2009.
[15] T. Mendez C. Partridge and W. Milliken. Host anycasting service. RFC 1546,
1993.
[16] K.Fall. A delay-tolerant network architecture for challenged internets. IRB-
TR-03-003, SIGCOMM, 2003.
[17] L. Pelusi, A. Passarella, and M. Conti. Opportunistic networking: data forward-
ing in disconnected mobile ad hoc networks. IEEE Communications Magazine,
44(11), 2006.
100
[18] Q. Ye, L. Cheng, M. C. Chuah, and B. D. Davison. Os-multicast: On-demand
situation-aware multicasting in disruption tolerant networks. IEEE VTC, Aus-
tralia, 2006.
[19] A.Vahdat and D. Becker. Epidemic routing for partially- connected ad hoc
networks. Technical report, Duke University, 2000.
[20] J. Davis, A. Fagg, and B. Levine. Wearable computers and packet transport
mechanisms in highly partitioned ad hoc networks. Intl, Symposium on Wear-
able Computers, 2001.
[21] K. M. Hanna, B. Levine, and R. Manmatha. Mobile distributed information
retrieval for highly-partitioned networks. IEEE ICNP, 2003.
[22] W. Zhao, M. Ammar, and E. Zegura. A message ferrying approach for data
delivery in sparse mobile ad hoc networks. ACM Mobihoc, 2004.
[23] T. Spyropoulos, K. Psounis, and C.S. Raghavendra. Spray and wait: An efficient
routing scheme for intermittently connected mobile networks. ACM SIGCOMM,
2005.
[24] K. Harras and K. Almeroth. Transport layer issues in delay tolerant mobile
networks. IFIP Networking, 2006.
[25] K. Harras, M. Wittie, K. Almeroth, and E. Belding. Paranets: A parallel
network architecture for challenged networks. HotMobile, 2007.
[26] B. Burns, O. Brock, and B. Neil Levine. Mora routing and capacity building in
disruption-tolerant networks. Elsevier Ad hoc Networks, 2008.
[27] Z. Chen, H. Kung, and D. Vlah. Ad hoc relay wireless networks over. moving
vehicles on highways. ACM Mobihoc, 2001.
[28] S. Bhattachargee, M. Ammar, E. Zegura, N. Shah, and Z. Fei. Application layer
anycasting. IEEE ISPFOCOM, 1997.
101
[29] D. Katabi and J. Wroclawski. A framework for scalable global ipanycast (gia).
ACM SIGCOMM, 2000.
[30] Park V. and J. Macker. Anycast routing for mobile services. IEEE CISS, 1999.
[31] Park V. and J. Macker. Anycast routing for mobile networking. IEEE MILCOM,
1999.
[32] Wang J., Zheng Y., and Jia W. An aodv-based anycast protocol in mobile ad
hoc network. IEEE PIMRC, 2003.
[33] Peng G., Yang J., and Gao C. Ardsr: an anycast routing protocol for mobile
ad hoc network. IEEE ISCAC, 2004.
[34] Z. Xie, J. Wang, Y. Zheng, and S. Chen. A novel anycast routing algorithm in
manet. IEEE PACRIM, 2003.
[35] Romit Roy Choudhury and Nitin H.Vaidya. Maclayer anycasting in ad hoc
networks. ACM SIGCOMM Computer Communication Review, 34(1), 2004.
[36] Y. Gong, Y. Xiong, Q. Zhang, Z. Zhang, W. Wang, and Z. Xu. Anycast routing
in delay tolerant networks. IEEE Globecom, 2006.
[37] C. Perkins. Ad-hoc on-demand distance vector routing. MILCOM’97 panel on
Ad Hoc Networks, 1997.
[38] T. Clausen and P. Jacquet. Optimized link state routing protocol. IETF Internet
Draft, IEEE, 2003.
[39] Raouf Hamzaoui Shakeel Ahmad and Marwan Al-Akaidi. Adaptive unicast
video streaming with rateless codes and feedback. IEEE Transactions on circuits
and system for video technology, 20(2):275–285, February 2010.
[40] D.B. Johnson and D.A. Maltz. Dynamic source routing in ad hoc wireless
networks. Mobile Computing, Kluwer Academic, 1996.
102
[41] C. Perkins and P. Bhagwat. Highly dynamic destination-sequenced distance-
vector routing (dsdv) for mobile computers. ACM SIGCOMM’ 94 Conference
on Communications Architectures, Protocols and Applications, ACM, 1994.
[42] M. May V. Lenders and B. Plattner. Density-based vs. proximity-based anycast
routing for mobile networks. IEEE INFOCOM, 2006.
[43] V. Park and S. Corson. Temporally-ordered routing algorithm (tora). IETF
Internet Draft, 2001.
[44] Simon Heimlicher Rainer Baumann, Vincent Lenders and Martin May. Routing
in large-scale wireless mesh networks using temperature fields. IEEE Networks,
22:25–31, January-February 2008.
[45] Vincent Lenders Rainer Baumann, Simon Heimlicher and Martin May. Routing
packets into wireless mesh networks. IEEE WoWMoM, 2007.
[46] Syed Muhammad Reza Muhammad Ali Khan and Hamed Moradi. A brief
overview of wireless mesh networks with focus on routing. IEEE-ICC, 2007.
[47] A. Lin A. Basu and S. Ramanathan. Routing using potentials: A dynamic
traffic-aware routing algorithm. ACM-SIGCOMM’03, 2003.
[48] J. Faruque and A. Helmy. Rugged: Routing on fingerprint gradients in sensor
networks. IEEE ICPS, 2004.
[49] S. Toumpis and L. Tassiulas. Packetostatics: Deployement of massively dense
sensor networks as an electrostatic problem. IEEE INFOCOM, 2005.
[50] J. Gao Q. Fang and L. J. Guibas. Locating and bypassing routing holes in
sensor networks. IEEE INFOCOM, 2004.
[51] M. Wattenhofer M. O’Dell, R. O’Dell and R. Wattenhofer. Lost in space or
positioning in sensor networks. REALWSN, 2005.
103
[52] Mihail L. Sichitiu. Wireless mesh networks chal-
lenges and opportunities. 2005-2006 [Online] Available:
http://www.ece.ncsu.edu/wireless/MadeInWALAN/wmnTutorial.ppt [Last
Accessed: April. 20, 2010].
[53] Abid Ali Minhas. Power Aware Routing Protocols for Wireless ad hoc Sensor
Networks. PhD thesis, Graz University of Technology, Graz, Austria, March
2007.
[54] H.N. Gabow T.X. Brown and Q. Zhang. flow-life curve for a wireless ad hoc
network. pages 128–136. ACM MobiHoc, Long Beach, CA, 2001.
[55] M. Bhardwaj and A.P. Chandrakasan. Bounding the lifetime of sensor networks
via optimal role assignments. pages 1587–1596, New York, NY, 2002. IEEE
Infocom.
[56] A. Chandrakasan W. Heinzelman and H. Balakrishnan. An application-specific
protocol architecture for wireless microsensor networks. IEEE Trans. on Wire-
less Commun, 1(4):660–670, October 2002.
[57] K. Dasgupta K. Kalpakis and P. Namjoshi. Efficient algorithms for maximum
lifetime data gathering and aggregation in wireless sensor networks. Computer
Networks, 42(6):697–716, August 2003.
[58] D. Simplot J. Cartigny and I. Stojmenovic. Localized minimum energy broad-
casting in ad-hoc networks. pages 2210–2217, San Francisco, CA, 2003. IEEE
Infocom, IEEE.
[59] F. Hadi, N. Shah, A.H. Syed, and M. Yasin. Adaptive anycast: A new anycast
protocol for performance improvement in delay tolerant networks. IEEE ICIT,
2007.
[60] F. Hadi, N. Shah, A.H. Syed, and M. Yasin. Effect of group size on anycasting
with receiver based forwarding in delay tolerant networks. IEEE ICEE, 2007.
104
[61] Thomas Staub Markus Walchli, Reto Zurbuchen and Torsten Braun. Back-
bone mac for energy-constrained wireless sensor networks. 34th Annual IEEE
Conference on Local Computer Networks (LCN), IEEE, 2009.
[62] Wang Wenbo Wang Xiangli, Li Layuan. An energy-efficiency multicast rout-
ing algorithm in wireless sensor networks. IEEE International Colloquium on
Computing, Communication, Control, and Management, IEEE, 2008.
[63] H. Tian C. Qing and T. Abdelzaher. ucast: Unified connectionless multicast
for energy efficient content distribution in sensor networks. IEEE Transactions
on Parallel and Distributed Systems, 18(2):240 – 250, Feb 2007.
[64] S. B. Wu and K.S. Candan. Gmp: Distributed geographic multicast routing in
wireless sensor networks. pages 49 – 49. 26th IEEE International Conference
Distributed on Computing Systems, IEEE, 2006.
[65] Wen-Zhan Song Xiang-Yang Li and Weizhao Wang. A unified energy efficient
topology for unicast and broadcast. MobiCom’05, ACM, 2005.
[66] Gregory M. P. O’hare Raja Jurdak, Antonio G. Ruzzelli and russell higgs. Di-
rected broadcast with overhearing for sensor networks. ACM Transactions on
Sensor Networks (TOSN), 6(1), December 2009.
[67] C. Intanagonwiwat and D. De Lucia. The sink-based anycast routing
protocol for ad hoc wireless sensor networks. Technical report, USC
Computer Science Department, HRL Research Laboratories available at
http://www.usc.edu/dept/cs/technical reports.html, 1999.
[68] N. Bulusu W. Hu and S. Jha. A communication paradigm for hybrid sen-
sor/actuator networks. pages 201–205, Barcelona, Spain, 2004. 15th IEEE In-
ternational Symposium on Personal, Indoor and Mobile Radio Communications
(PIMRC), IEEE.
105
[69] Hou Shi and Hanif D. Sherali. Optimal base station selection for anycast routing
in wireless sensor networks. IEEE transactions on vehicular technology, 55(3),
May 2006.
[70] R. Buyya S. Venugopal and K. Ramamohanarao. A taxonomy of data grids
for distributed data sharing, management and processing. ACM Computing
Surveys (CSUR), 2006.
[71] I. Foster and C. Kesselman. The grid: Blueprint for a new computing infras-
tructure. Morgan Kaufman, 1999.
[72] C. Kesselman I. Foster and S. Tuecke. The anatomy of the grid: Enabling
scalable virtual organizations. International J. Supercomputer Applications,
2001.
[73] Y. Tanimura, T. Ikegami, H. Nakada, Y. Tanaka, and S. Sekiguchi. Imple-
mentation of fault-tolerant gridrpc applications. Journal of Grid Computing,
4(2):145–157, June 2006.
[74] A. Wahheed, W. Smith, J. George, and J. Yan. An infrastructure far monitoring
and management in compuwtionvl grids. Proceedings of the 5th Workshop on
Languages, Compilers, and Run-time System for Scalable Computer, March
2000.
[75] Anh Nguyen-Tuong. Integrating Fault-Tolerance Techniques in Grid Applica-
tions. PhD thesis, August 2000.
[76] R. Medeiros, W. Cirne, F. Brasileiro, and J. Sauve. Faults in grids: Why are
they so bad and what can be done about it? pages 18–24, Proc. of the 4th
International Workshop on Grid Computing, Phoenix, 2003. Arizona, USA.
[77] B. Nazir and T. Khan. Fault tolerant job scheduling in computational grid.
pages 708–713. ICET-IEEE, Peshawar, Pakistan, 2006.
106
[78] E. Roman. A survey of checkpoint/restart implementations. Technical report,
Lawrence Berkeley National Laboratory, LBNL-54942, 2003.
[79] Sriram Krishnan. An Architecture for Checkpointing and Migration of Dis-
tributed Components on the Grid. PhD thesis, Department of Computer Sci-
ence, Indiana University, November 2004.
[80] Muhammad Affan and M. A. Ansari. Distributed fault management for com-
putational grids. In Proc. of Fifth International Conference on Grid and Co-
operative Computing (GCC 2006), pages 363–368, Changsha, Hunan, China,
2006.
[81] Forrest Warthman. Delay-tolerant networks (DTNs) A tutorial. [Online]
http://www.ipnsig.org/reports/DTN Tutorial11.pdf [Last accessed May 10,
2010].
[82] A. Lindgren. Routing and Quality of Service in Wireless and Disruption Toler-
ant Networks. PhD thesis, Lulea University of Technology, 2006.
[83] Daniel E. Crocker Phaedra Green Todd O’Brien Syd Levitus George
W. Boehlert, Daniel P. Costa and Burney J. Le Boeuf. Autonomous pinniped
environmental samplers; using instrumented animals as oceanographic data col-
lectors. Journal of Atmospheric and Oceanic Technology, 18(11):1882–1893,
May 2001.
[84] Vincent D. Park and M. Scott Corson. A highly adaptive distributed routing
algorithm for mobile wireless networks. Kobe, Japan, 1997. INFOCOM ’97.
[85] S. Burleigh. Delay-tolerant networking: An approach to interplanetary internet.
IEEE Communications Magazine, 2003.
[86] Richard G. Ogier, Fred L. Templin, and Mark G. Lewis. Topology dissemination
based on reverse-path forwarding. Mobile Computing, February 2004.
107
[87] Network simulator. [Online] http://www.isi.edu/nsnam/ns.
[88] B. S. Manoj C. Siva Ram Murthy Tamma Bheemarjuna Reddy, I. Karthigeyan.
Quality of service provisioning in ad hoc wireless networks: a survey of issues
and solutions. Ad Hoc Networks, 4(1):83–124, 2006.
[89] Aguayo D. A.Chambers B. Morris Couto, D.S.J.D. ’performance of multihop
wireless networks: Shortest path is not enough. ACM SIG COMM Computer
Communications Review, 33(1):83– 86, June 2003.
[90] Bernhard Plattner Vincent Lenders, Martin May. Service discovery in mobile
ad hoc networks. a field theoretic approach. Elsevier Journal on Pervasive and
Mobile Computing, 1(3):343–370, September 2005.
[91] Pavan Nuggehalli Mario Strasser Martin May Bernhard Plattner Praveen Ku-
mar, Joy Kuri. Connectivity-aware routing in sensor networks. Sensorcomm’
07, 2007.
[92] Omnet++. [Online] http://www.omnetpp.org.
[93] Daniel K. Lee Ben L. Titzer and Jens Palsberg. Avrora: scalable sensor network
simulation with precise timing. pages 477 – 482, Los Angeles, CA, USA, 2005.
Fourth International Symposium on Information Processing in Sensor Networks
(IPSN).
[94] Dionysys Blazakis Jonathan McGee Dan Rusk Manish Karir, Jonathan Polley
and John S. Baras. Atemu: a fine-grained sensor network simulator. pages 145
– 152. First Annual IEEE Communications Society Conference on Sensor and
Ad Hoc Communications and Networks, 2004.
[95] Lewis Girod Jeremy Elson and Deborah Estrin. Emstar: Development with
high system visibility. IEEE Wireless Communications, 11(6):70–77, December
2004.
108
[96] Rimon Barr. Swans: Scalable wireless ad hoc network simulation. In
Handbook on Theoretical and Algorithmic Aspects of Sensor, Ad hoc Wire-
less, and Peer-to-Peer Networks Ch. 19, CRC Press, pages 297–311, URL:
http://jist.ece.cornell.edu/docs.htm, accessed Feb 14, 2010], 2005.
[97] Qualnet. http://www.scalable-networks.com/ [Last accessed Feb 14, 2010].
[98] Opnet. http://www.opnet.com/ [Last accessed Feb 14, 2010]].
[99] Sense. http://www.ita.cs.rpi.edu/sense/index.html [Last accessed Feb 14, 2010].
[100] Matt Welsh Philip Levis, Nelson Lee and David Culler. Tossim: Ac-
curate and scalable simulation of entire tinyos applications. First ACM
Conference on Embedded Networked Sensor Systems (SenSys 2003). URL:
http://www.eecs.berkeley.edu/ pal/pubs/nido.pdf [Last accessed Feb 14, 2010],
2003.
[101] Tinyos. http://www.tinyos.net [Last accessed Feb 14, 2010].
[102] Tinyos applications. http://webs.cs.berkeley.edu/tos/tinyos-
1.x/doc/multihop/multihop routing.html, [Last accessed Feb 15, 2010].
[103] Tinyviz. URL: http://www.tinyos.net/tinyos-.x/doc/tutorial/lesson5.html,
[Last accessed Feb 15, 2010].
[104] S. Venugopal, R. Buyya, and K. Ramamohanarao. A taxonomy of data grids
for distributed data sharing. Management and Processing,ACM Computing
Surveys (CSUR), 2006.
[105] R. Medeiros, W. Cirn, F. Brasileiro, and J. Sauve. Faults in grids: Why are
they so bad and what can be done about it? In Proc. of the 4th International
Workshop on Grid Computing, pages 18–24, Phoenix, Arizona, USA, 2003.
[106] Ian Foster and Carl Kesselman. Globus: A metacomputing infrastructure
toolkit. Journal of Grid Computing, pages 115–128, 1997.
109
[107] Net Grid Computing Framework. GRIDS Lab, University of Melbourne.
[108] Akshay Luther, Rajkumar Buyya, Rajiv Ranjan, and Srikumar Venugopal. Al-
chemi: A .net-based grid computing framework and its integration into global
grids. Technical report, GRIDS-TR- 2003-8, Australia, 2004.
[109] A. Luther, R. Buyya, R. Ranjan, and S. Venugopal. Peer-to-peer grid computing
and a .net-based alchemi framework. New York, USA, 2005. Wiley Press.
[110] Riphah international university, islamabad, pakistan. available at:
http://riphah.edu.pk/ accessed May 18, 2010.
110