DOC Version

advertisement
Final Thesis
Routing overheads vs. End To End
Connectivity on mobile ad hoc networks
By. Christopher Totani
Supervisor: Dr. Grant Wigley
Declaration
I declare that this thesis does not contain, without acknowledgement, any material submitted for a
degree or diploma within the university. To the best of my knowledge, all material previously
published or written by the author has been cited by the correct author(s).
Christopher Ian Totani
November 2014
Acknowledgements
I would like to thank all of my friends and family for support through this experience and giving
motivation to continue on. I would also like to thank my supervisor, Dr. Grant Wigley, for providing
me with the opportunity to do such a thesis with the Defence Force and for providing me with great
guidance upon construction of this thesis. I would finally like to thank the DSTO for providing me
with the opportunity to work with them and expand my knowledge in a field I would have never been
exposed to.
Table of Contents
Table of Figures ...................................................................................................................................... 6
1. Introduction ......................................................................................................................................... 7
1.1 Statement of Research................................................................................................................... 7
1.2 Field of Thesis............................................................................................................................... 8
1.3 Significance and Contributions ..................................................................................................... 8
2. Literature Review................................................................................................................................ 9
2.1 Network Fundamentals ................................................................................................................. 9
2.1.1 7 Layer OSI model ................................................................................................................. 9
2.1.2 What is a protocol .................................................................................................................. 9
2.1.3 Level 2 - Data Link Layer .................................................................................................... 10
2.1.4 Level 3 - Network Layer ...................................................................................................... 10
2.1.5 Network Layer protocols ..................................................................................................... 10
2.2 Concept of Ad-hoc networking and Mobile Ad Hoc Networking ............................................. 11
2.2.1 Structured Networks ............................................................................................................ 11
2.2.2 Problems with Structured Networks .................................................................................... 11
2.2.3 What is Ad-Hoc Networking and why it's desirable ............................................................ 12
2.2.4 Mobile Ad-Hoc networking (MANETs) and Routing ......................................................... 12
2.3 Introduction to Routing and Routing Protocols .......................................................................... 19
2.3.1 What is RIP (And RIPv2) .................................................................................................... 19
2.3.2 What is OSPF ....................................................................................................................... 20
2.3.3 MANET Routing Protocols ................................................................................................. 22
2.3.4 Routing Overheads............................................................................................................... 27
2.4 Conclusion of literature review ................................................................................................... 28
3. Methodology ..................................................................................................................................... 30
3.1 Question and Explanation ........................................................................................................... 30
3.1.1 Sub questions ....................................................................................................................... 30
3.2 Steps towards research and answering the question ................................................................... 31
3.2.1 Review literature and determine overheads ......................................................................... 31
3.2.2 Specify metrics of routing overheads ................................................................................... 31
3.2.3 Finalise routing protocols to be simulated ........................................................................... 31
3.2.4 Simulate routing algorithm under different scenarios .......................................................... 32
3.2.5 Map data to meaningful diagrams/tables, draw conclusions from data gathered. ............... 32
3.3 Research method to use .............................................................................................................. 32
3.4 Sampling and population ............................................................................................................ 33
3.5 Data Collection ........................................................................................................................... 34
3.6 Data Analysis .............................................................................................................................. 35
3.7 Tools ........................................................................................................................................... 36
4. Parameters for simulations ................................................................................................................ 38
4.1 Selected metrics for Routing overheads and End to End Connectivity ...................................... 38
4.1.1 Routing overhead metrics .................................................................................................... 38
4.1.2 End to End Connectivity/Performance metrics .................................................................... 40
4.2 Mobility Models.......................................................................................................................... 43
4.2.1 What are Mobility Models ................................................................................................... 43
4.3 Traffic Generated .................................................................................................................... 45
4.4 NS-2 Parameters ......................................................................................................................... 46
4.5 Bonnmotion Parameters .............................................................................................................. 49
4.6 NS-2 traffic generation simulation .............................................................................................. 49
5. Simulations ....................................................................................................................................... 52
5.1.1 Ad-Hoc On Demand Distant Vector Protocol (AODV) ..................................................... 52
5.1.3 Destination Sequenced Distance-Vector routing protocol (DSDV) .................................... 55
5.1.4 Optimised Link State Routing protocol (OLSR) ................................................................. 56
5.2 Simulation Results ...................................................................................................................... 58
5.2.1 Normalised routing load....................................................................................................... 58
5.2.2 Packet Delivery Ratio .......................................................................................................... 59
5.2.3 End to End Delay ................................................................................................................. 60
5.2.4 Throughput ........................................................................................................................... 61
5.2.5 Average Hop Count ............................................................................................................. 62
5.3 Summary of simulation results ................................................................................................... 62
6. Conclusion, Future Work .................................................................................................................. 64
6.1 Conclusion .................................................................................................................................. 64
6.2 Future Work ................................................................................................................................ 65
7. References ......................................................................................................................................... 67
Table of Figures
Figure 1 Example of MANET connections (black lines) vs. direct connection (red line) .................... 14
Figure 2 Structure of the HOLSR protocol ........................................................................................... 26
Figure 3 NS-2 Parameters ..................................................................................................................... 47
Figure 4 A Route Request Packet [a] .................................................................................................... 53
Figure 5 A Route Reply Packet [a] ....................................................................................................... 53
Figure 6 A DSDV "Message" packet [a] .............................................................................................. 55
Figure 7 Normalise Routing Load Results ............................................................................................ 58
Figure 8 Packet Delivery Ratio Results ................................................................................................ 59
Figure 9 End to End Delay Results ....................................................................................................... 60
Figure 10 Throughput Results .............................................................................................................. 61
Figure 11 Average Hop Count Results ................................................................................................. 62
1. Introduction
This introduction will give a synopsis on the purpose of the research on MANETS, what aspects of
research this paper will address and what will be contributed within the research.
1.1 Statement of Research
The research proposed will focus on Mobile Ad Hoc Networking (MANET) and Routing overheads
for MANETs. MANETs are an infrastructure-less based networking paradigm, which intends to run a
network purely with end user devices, rather than utilising any form of fixed infrastructure. In other
words, networking is in the hands of end user devices, rather than networking devices, such as routers
and switches. Routing overheads refers to the amount of networking resources dedicated towards up
keeping routing. Routing overheads are of major concern with MANETs, due to limited routing
resources, low bandwidth links, battery power reliance and the dynamic nature of a MANET, routing
can consume these resources. Routing overheads also become of major concern when the size and
distribution of the network is of significant size.
Applications for MANETs are optimal when infrastructure is either impossible, too costly or
impractical to use. MANETs also remove a single point of failure within the network, since it is self
configuring in maintaining routing of the network. The application of MANETs intended for this
research is for military use, requiring military needs to be met within the MANET. This thesis is
intended towards aiding within the research of military MANETs for the Defence Science and
Technology Organisation (DSTO).
As network size scales up, routing overheads become larger and more of a burden. Furthermore,
MANETs utilise wireless to connect between nodes, which tends to be lower in bandwidth in
comparison with wired network. These concerns are what motivates with research upon utilising
lower overheads with routing protocols.
The core influence of routing overheads is the routing protocols themselves, since protocols hold the
policies and algorithms required to function the network. The research shall evaluate several routing
protocols used within MANET structures and determine which theoretically provides lower routing
overheads. If possible, simulations will be done on such protocol to determine if routing overheads
would be significant, with full end to end connectivity being maintained.
The alternative towards utilising a MANET is a semi structured infrastructure, which contains
multiple gateways, utilising smaller MANETs to gain the benefits of both infrastructure and ad hoc
paradigms. However, such design will still incur costs with equipment and maintenance of the
network. Furthermore, it introduces a central point of failure within each sub-network, due to
dependency on the network gateway.
1.2 Field of Thesis
Ad Hoc Networking, Networking, Routing Protocol, Military Networks, Mobile Ad Hoc Networking,
1.3 Significance and Contributions
- Contribution for the defence force use of MANETS
- Expand understanding on hierarchical MANETs within context of military use.
- Expand understanding on comparisons between cylindrical and hierarchical MANETs
2. Literature Review
The literature review will start from this section, building up information on networking, ad-hoc
networking, mobile ad hoc networking and routing overheads. All information is based on research
found on research papers, journals, conference papers,
2.1 Network Fundamentals
This section will introduce the basics of networking and routing, focusing on the 7 layer OSI model,
protocols, routing and popular routing protocols. A network is a group of devices connected to share
data along connection lines. Data shared through a network is usually shared through wireless or
through physical lines. To maintain network connections, network devices are used to maintain
networking. Network devices can include router, switches, wireless gateways, network interface cards
and wireless cards. Devices typically used within a network are computers, PDAs, servers, phones
and networking equipment.
2.1.1 7 Layer OSI model
In networking, the whole basic scheme of protocols and packets can is expressed through the Open
System Connection (OSI) model. The OSI model includes the Application, Presentation, Session,
Transport, Network, Data link and Physical layer. Each layer is grouped by a protocol data unit
(PDU). This includes Data (Application, Presentation, Session), Segments (Transport), Datagram
(Network), Frame (Data Link) and Bit (Physical). For the sakes of relevance, we shall only focus on
Level 3 (Network) and level 2 (Data link) layer.
2.1.2 What is a protocol
A protocol in, networking terms, is a series of methods and rules used to regulate how data is
transmitted through a network. Protocols are generally expressed through packets, which are protocols
which can handle data. For example, the Transmission Control Protocol (TCP) is used for sending
reliable data over the internet. It regulates it by using a checksum to ensure the data contained is
within the packet, to ensure the data is accurate. the User Datagram Protocol on the other hand, use in
a similar fashion, but for more speedy and efficient data transfer, due to its lesser complexity.
Protocols are usually used within packets, which are a unit containing overheads/relevant packet
information and encapsulated data.
2.1.3 Level 2 - Data Link Layer
The data link layer functions in a similar fashion from the network layer, except it is responsible with
Media Access Control between nodes. Such includes error corrections and acknowledgements within
the physical layer, delivery of frames within a local network. The Data link layer is typically sub
layered between the Logical Link Control (LLC) layer and Media Access Control (MAC) layer.
2.1.4 Level 3 - Network Layer
The network layer is used mainly for routing, normally using routers. The layer is critical towards end
to end connectivity, since it maintains source and destination between hosts within a network. It also
maintains segmentation of networks and subnets.
2.1.5 Network Layer protocols
Due to routing occurring within the network layer, this section will focus on protocols which operate
within the Network layer. Such network layer protocols are commonly used within structured,
enterprise networks and are uncommon within military use. However, for the sakes of research, it is
important to acknowledge such protocols and reason with their suitability/unsuitability within mobile
ad-hoc networks.
2.2 Concept of Ad-hoc networking and Mobile Ad Hoc Networking
The previous sections have outlined the basics of networking and purposes of the network layer
protocol. This section shall expand upon an ad-hoc infrastructure and why a structured network may
not be suitable within a military.
2.2.1 Structured Networks
A structured networking infrastructure is a fixed, predictable network topology which uses specialised
devices to maintain communications. The networks explained within the previous section are
examples of structured networks. Network topologies within an structured network change less
frequently in comparison with an unstructured network. Network topology changes within a
structured network are usually due to faults (i.e. power outage), erroneous change of configuration
and addition of new hardware. Usual topology changes can be very costly, since information of
topology change must be between routing nodes.
Structured networks also allow special devices to do specific networking tasks. Typically within
enterprise networks, utilises the core, distribution and access layer. The core layer utilises layer 3
networking/routing and acts as the backbone for the whole network. The distribution layer utilises
both layer 3 networking and layer 2 switching, depending on the size of the distribution layer. The
access layer is purely for switching and brings connectivity for end devices.
2.2.2 Problems with Structured Networks
The disadvantages of Structured networks for this section will focus on general, with further sections
exposing why structured networks are not the best solution in the context of military ad hoc networks.
One disadvantage of Structured networks is the lack of flexibility for nodes to connect to any part of
the network. As outlined before, the typical core, distribution, access layer only allows access to a
network through an access layer. In certain situations, it is important for access to be given through
the whole network. In [2], the article shows that a military ad hoc network used allows flexibility for
team radios, as end devices, to communicate freely within the network and even allows the end
devices to act as repeaters within the network.
One major disadvantage of structured networks is since they have central network administration
which are configured by network administrators, they also require on-going maintenance/repairs to
keep uptime high.
Structured networks also require the usage of highly expensive routing, switching and access point
devices within a network. Such networks, especially with many nodes, will require more powerful and
more devices to support the infrastructure
2.2.3 What is Ad-Hoc Networking and why it's desirable
Ad-hoc networking is an unstructured network, which contains end devices which self manages the
whole network, typically over a wireless medium. Each node within the network acts as a link and
entry point within the network, allowing participants to connect freely anywhere within the network.
Ad-hoc networking also provides a natural self healing, self configurable network ecosystem which
requires little maintenance. This is due to Ad-Hoc networking protocols being structured around each
node being its own router and is aware of every other node.
2.2.3.1 Problems with ad hoc networking
The problems with ad hoc networking is due distinct rapid nature in comparison towards structured, it
is difficult to use protocols made for structured networks. For that reason, it must be taken into
consideration for us to For example, the OSPF routing protocol relies upon flooding packets within
the network to maintain a consistent topology. While this flooding is a very low cost within a wired
network, due to the reliability of wired links, wireless links would have a higher rate of error. Since
OSPF is very rapid with topology updates, we may see nodes drop out, taken off the current topology
and have to be re-added to the topology. Topology changes within OSPF also rely on neighbourhood
adjacencies (i.e. nearby nodes), which within an ad hoc network, can change if nodes move around a
lot. More information on OSPF for Mobile Ad Hoc Networks shall be explained through the MANET
section.
2.2.4 Mobile Ad-Hoc networking (MANETs) and Routing
What has been introduced was the concept of ad hoc networking. This section will be dedicated
towards an implementation of ad-hoc networking, known as Mobile Ad Hoc Networking (MANET).
A MANET is a type of ad-hoc infrastructure, which solely uses mobile devices to structure network.
2.2.4.1 Common uses of a MANET
Typically, MANETs are not used when Infrastructure services are more practical and cost effective to
use. MANETs are used when infrastructure cannot be utilised, due to the environment the network
must be used in. Besides the scope of the research, one prime example would be emergency services
using an ad-hoc network to communicate data in the event of a disaster. Hobbyist may also utilise
MANETs for educational purposes or personal uses for a Personal Area Network.
MANETs may also be beneficial to be used in infrastructure- less scenarios, which may contain a
simple task and not need the overheads of a structured network. One example is a vehicular ad-hoc
network used within a car's safety features. Such MANET would be used for cars to establish
communication to share data which may be useful in the event of a possible collision (e.g. car's speed
and distance to determine automatic braking pressure).
2.2.4.2 Military usages of MANETs
Military usage of MANETs is a major use, due to the self configuring properties and mobility of such
network design. Combatants can use MANETs to communicate over terrain which may have
restricted direct communication, since combatants in between equip with a radio link can act as a
intermediate repeater to help establish connections[3]. The image below depicts what has been
explained, with the black lines depicting a MANET structure and the red line depicting a direct
connection between the two end nodes. The 3 nodes in between act like a repeater and passes routing
through itself. It's important to note that if the two end nodes would not only have a poorer connection
due to distance, but also due to noise and attenuation from the nodes interfering inbetwen.
Figure 1 Example of MANET connections (black lines) vs. direct connection (red line)
One news article [28] shows an example of how the US military deployed a MANET called "Force 21
Battle Command Brigade and Below". This MANET was used to keep track of fast moving combat
vehicles, using GPS in conjunction to detect where vehicles were. The MANET was also used to
provide info on detected landmines, directions of where vehicles must go, location of friendly forces
and location of enemy forces. The MANET has also arguably helped with reducing incidents of
friendly fire, due to awareness of combatants.
2.2.4.3 Advantages of MANETs
MANETs provide many advantages than towards the alternative of using an infrastructure based
network. MANETs do not require networking devices, such as routers and switches, to maintain a
network, or a central medium to ensure full connectivity, such as a satellite network, therefore
reducing costs required to structure the network. MANETs also provide a self-configuring network,
which is required due to the high mobility nature of a MANET.
MANETs can also provide high mobility structure less network, since the MANET is dependent on
mobile devices. This allows full mobility of the network, without requiring fixed infrastructure to
piece the network together. Certain military situations may require soldiers to move forward rapidly
between areas. Infrastructure set up (i.e. antenna tower) within an area may not be portable enough to
move rapidly. Such infrastructure may be far away, to the point where connectivity is either unreliable
or no connectivity is present. Furthermore, the infrastructure could be limited due to terrain (i.e. in
hilly terrain, signals could be blocked between a rugged CB radio and a antenna tower due to a hill in
the way) if nodes are positioned in an awkward position.
MANETs are also highly scalable, due to the low cost and low barriers to entry for a node to connect
to the MANET. MANETs do not require additional equipment (e.g. Ethernet cables, wireless access
points, switches) to allow access into the network. Generally, a MANET would only require a
compatible wireless connection, the node is a supported device and required authentication to access
the MANET. However, this scalability is not unlimited and can be hampered due to other factors. This
can be hampered due to routing protocol, number of nodes (can fill routing tables), type of data
commonly utilised within the network.
2.2.4.4 Disadvantages of MANETs
Whilst the high mobility, scalability and cost are the major advantages within a MANET, it is not
without its downfalls.
One major downfall within a MANET is the reduced security in comparison with Infrastructure
networks. Typically, routing within a structured network only contains wired connections, which is
difficult tamper into without breaking into the system either through physical means or through the
network. Within a MANET, there are many entry points into the network at every node. This can
leave the network open to malicious attackers which may try to break into the network, or take down
the network. [7] sees an issue with DoS attacks being implemented within a MANET and could
provide serious effects upon the MANET if a DoS attack was successful. Whilst it is not within the
scope of this paper, it is important to outline this major downfall when implementing a MANET.
Another disadvantage of MANETs is the reduced bandwidth and speed of such structure in
comparison towards an infrastructure network. The MANET is dependent upon the capacity of each
nodes components and processing power. Such MANET links may not have the same bandwidth
capacity as a wireless access point on an infrastructure network. Furthermore, speed could also be
bottleneck through the node's processing power, since each node would have to process their own
data, let alone the routing updates/requests/hops. It is why the routing protocol of the MANET must
be chosen carefully that it produces less routing overheads within the MANET.
A further disadvantage is an extension to the argument previously mentioned in the ad hoc networking
section, which disputed the usability of common protocols used within infrastructure network. Due to
its relevance to the topic, the next section shall be dedicated towards the problems with TCP and
UDP within MANETs.
2.2.4.5 Problems with TCP in MANETs
Most networks usage revolves around internet usage, which requires the usage of the Transmission
Control Protocol (TCP). The TCP protocol is well used, due to its high accuracy of data transfer over
the internet. This is due to TCP's error correction methods of requiring a retransmission of data which
has been deemed erroneous. TCP is however not favoured with real time applications, such as Voice
over IP (VoIP) and real time video streaming. This is because the latency caused by TCP's
retransmission of erroneous data. TCP's retransmission of data has actually been proven through
several papers as to why TCP may not be the best solution for usage within MANETs.
[8] did testings with comparisons between TCP and UDP to look at their overheads and energy
consumed respectively. [9] has done tests on TCP, UDP and ICMP ( Internet Control Message
Protocol) and has compared the throughput of each protocols. The test was done over the ns-2
simulator, with the OSLR protocol as the routing protocol. [9] shows TCP fluctuating in throughput,
where as UDP and ICMP, which do no retransmissions, actually kept a constant throughput.
However, TCP should not be fully disregarded within the usage of MANETs. Research shows very
little data on TCP transmissions over a hierarchical MANET, with flat topology MANETs used within
tests. Furthermore, there are test [10] looks at the energy usage of TCP over MANETS, tracking the
joules expended to do a TCP transmission over different scenarios, under different MANET routing
protocols (AODV, DSDV, DSR, OLSR). The paper also looked at different TCP variants, looking at
New Reno, SACK, Vegas and Westwood. The Westwood TCP model saw a great balance between
error correction and computational power used.
An alternative towards the TCP protocol is the UDP protocol. Such protocol will be inevitable to be
used within voice and real time video use within a MANET. However, for data, TCP is commonly
used within structured networks. The UDP protocol, as shown in [8], gives more throughput than TCP
consistently. The UDP protocol also produces less overheads, due to its simplicity and lack of
retransmission erroneous data. However, UDP is far from a perfect solution for data transfer usage
within a MANET.
Since UDP does not retransmit data, there are chances of inaccurate data sent to the other nodes, such
as garbled text sent between two soldiers or a picture containing foreign artefacts caused by the poor
retransmission. Furthermore, UDP does not support encryption through the protocol itself, since it's
overhead contains no room for encryption. This can cause serious issues with attacks upon the
network, or the integrity of highly confidential data.
To ensure accurate data is processed through the network, a Quality of Service (QoS) framework
would be implemented. QoS looks at certain metrics, such as error rate, bandwidth and throughput to
ensure data is accurate and of high quality[13]. Quality of Service is usually used within services,
such as Voice over IP (VoIP) and real time video transfers, since such type of applications are real
time and cannot afford to have retransmission of packets through a network. QoS is generally seen as
a mechanism of quality control within a network, rather than an actual protocol implementation (i.e.
real time video would be transported through a UDP packet in layer 4 (Transport layer), but the
application layer protocol (i.e. RSVP protocol) may implement QoS metrics.
Quality of Service provides quality by priotising packets delivered through the MANET, typically
allowing real time packets to traverse through the network first (i.e VoIP, real time video) than ones
which are not produced in real time (i.e. HTTPS, FTP). The utlisation of QoS can be a quite critical
aspect, since the routing protocol and application layer protocol needs to support QoS metrics.
QoS is a significant factor within MANETs, since QoS not only could help provide accuracy with
data between nodes without the usage of TCP, but QoS needs to be optimised for MANETs [25]. The
limitations for QoS within in MANETs are due to the dynamic topology of a MANET, with limited
resources [36].
This section is not core to research, but to outline and demonstrate through literature that MANETs do
not behave or react like structured networks. This is to highlight the specialist attention MANETs
need, in order to function within reason and ensure protocols to be implement will be stable within a
MANET. However, a MANET routing protocol with QoS support should be considered of favorable
choice, especially if MANETs will be using a range of data (especially VoIP). A support QoS would
make better use of its limited bandwidth.
2.3 Introduction to Routing and Routing Protocols
Routing, within the context of packet data networks, focuses with utilising the most efficient path of
forwarding a packet within a network. Routing devices typically form the backbone of a wide area
network (WAN), where as switches are usually utilised within Local Area Networks (LANs). Due to
the complexity of many WANs with multiple LANs and the number of routers within a network,
usually for purposes of redundancy, uptime and performance, it is important that the protocols used
within network are suitable for the network. Routing is typically functioned within the layer 3 of a
network, with the utilisation of a routing protocol. Typical routing protocols are RIP, OSPF, EIGRP
and BGP. Routing does not rely on a routing protocol, since routing can be static by setting up what is
known as a static route. Static routes are generally the most efficient way of routing, however have
higher maintenance overheads and are typically not reserved for dynamic networks.
Routing typically relies on what is known as a Routing table. A routing table is a database stored
within the NVRAM of a routing device, which contains vital information for a routing infrastructure,
such information includes participating networks, routing protocol, source of networks, next hop and
next hop metric. Different routing protocols may have their own related tables, usually for other
metrics, network topology or supporting features.
Due to their relevance, the remainder of this section will focus on OSPF and RIP
2.3.1 What is RIP (And RIPv2)
Routing Information Protocol (RIP) is an older routing protocol, which utilises a very basic routing
algorithm. RIP only takes into consideration of hop count as a metric to determine a routing path. The
hop count is the number of routers a packet must route through, before it is transported to its final
destination. This hop count is limited to 15 hops, in which a 16th or more hop is considered an infinite
hop. Due to the limitations of RIP, it is not a commonly used protocol for larger scale networks. A
newer version, RIPv2, supports a higher hop count (96), however, lacks the extra metrics more
sophisticated protocols contain to implement more efficient routing paths [1].
2.3.2 What is OSPF
The Open Shortest Path First (OSPF) protocol is a link state, non proprietary routing protocol. It
utilises Shortest Path First to route to the most efficient destination. On top of that, OSPF has other
mechanisms which allows the network to run more efficiently. OSPF allows the network to be divided
into "areas", which allows grouping of certain routers within a network. OSPF introduces "Stub
routing", which allows configuration of certain areas to have smaller, more efficient routing tables. [1]
2.3.2.1 How OSPF Works
Each participating router within an OSPF network contains what is known as a Link State database
(LSDB). This Link State database contains the every participating router's ID (unique identification
for each router), Link State Information (maintained through Link State Advertisements(LSA)),
adjacencies, neighbourhoods and neighbouring router costs[33]. Link State Advertisements (LSA) are
used to maintain the LSDB and contains different types of LSA. Typical LSA's are LSA type 1
(Router LSA used for maintaining Inter area routing) and LSA type 2 (Network LSA used to
maintain lists of routers)[1]. These LSA advertisements are typically flooded through the network or
within the area, depending on the LSA type (LSA type 1 and two are flooded within an area only).
To maintain connectivity within the network, the OSPF protocol forms adjacencies with other
connected routers known as "neighbours". These neighbour adjacency help with determining with a
router's presence and links within a network. If a neighbour adjacency was to be lost, the link between
the two routers would be null and may cause a router to be removed from the network's topology.
These adjacencies are maintained by flooding hello packets out the interfaces of the router, to keep
neighbours alive and discover new links.
Furthermore. OSPF contains a unique feature which the network is divided into areas. The different
areas can be categorised as the backbone area (area 0, all areas connect to this), Regular Area
(contains internal/external routes) and stub area (only contains internal and default routes). Every
OSPF instance requires an area 0 and every area should be connected to this area. An area does not
have to be directly connected to area 0 and can be connected through a virtual link, which tunnels a
connection through a different area. However, it is commonly seen as bad practice and a sign of poor
network design[1].
2.3.2.2 OSPF on Ad-Hoc Networking
There is research on implementing extensions for OSPF to work on ad hoc networks, since OSPF is
an already established and well supported protocol. Some papers have introduced implementations on
MANET Extensions for OSPF[11] [12] [34]. Some implementations are either newly developed or
untested. Others have provided simulations and well documented information on the algorithm on
how their version of OSPF works
[27] notes the impossibility of implementing OSPF areas within an ad hoc network, due to how OSPF
areas form and moving nodes. [28] argues for mobile nodes to join other areas, however this would
break the fine tuning aspects of OSPF areas (e.g. allow certain areas to reduce routing tables).
Furthermore, nodes would have to be preconfigured to join OSPF areas, otherwise they will not
receive connectivity.
However, there is compelling research done on MANET extensions for OSPF, which may seem as a
plausible alternative to MANET routing protocol. This protocol is known as OSPF-MDR (OSPFMANET Designated Routers). It looks at utilising the designated router/backup designated router
feature of OSPF and treating the designated routers as a default gateway (similar to how a cluster head
is within a hierarchical MANET) [34]. The inter-area node is referred to as the MANET Designated
Router (MDR), with a back up node known as the Backup MDR (BMDR). The MDR is responsible
for flooding LSA packets, through gathering adjacencies formed, to form the topology. To help
reduce overheads, MDRs can be declared as a non flooding MDR to reduce routing overheads, by
reducing LSA updates[35].
It also tries to reduce routing overheads by declaring different types of adjacencies to reduce
overheads with hello packet flooding. When forming adjacencies, the node differentiates neighbour
relationships with adjacent nodes. This influences whether the node is routable or if even full
connectivity between the nodes are plausible [35]. These neighbourhood relationships are Down,
One-Way, Dependent, Selected Advertised (included in LSA) and Bidirectional [34]. Such extension
shows potential, however lacks any tests within the scope of military networks. However, such
protocol is a possible candidate for comparison with other MANET protocols[39].
2.3.3 MANET Routing Protocols
For this section, we shall produce the different routing protocol used within MANETs. The first part
of the section will contain Flat topology MANETs, where as the rest will cover hierarchical
MANETs. Please note that any references of "Flat", "Cylindrical" or "Full connectivity" will reference
to Flat topology MANETS, where as "Clustered" and "Hierarchical" will refer to Hierarchical
Topology MANETs.
2.3.3.1 Flat Topology MANETs
Flat topology MANETs are usually referred to as standard MANETs in academia and are widely used
within research purposes of MANETs, such as [9] comparing different flat topology MANETs within
their energy usage. Flat topology MANETs typically contain a routing table and routing hops. They
also do typical route requests, route replies and routing updates. Routing updates, however, can differ
between different Flat topology MANET protocols. Typically, different routing protocols function
differently by how they process routing updates. The two main types are known as Reactive and
Proactive, with a combination of both seen as hybrid.
Reactive MANET protocols
Reactive MANET protocols are seen as protocols which every node maintains their own tables and
only do routing updates sporadically [9]. Dependant on the protocol, routing tables are recalculated
during an event or when a node is not found. Reactive protocols are regarded in [9] and [10] as being
the most efficient protocols with little overheads.
Proactive MANET protocols
Proactive MANET protocols differ in that each node is required to update their own tables regularly.
[9] has suggested that Proactive MANET protocols function similarly towards protocol used within
infrastructure networks.
Hybrid MANET protocols
Hybrid MANET protocol try to bridge the best of both Reactive and MANET protocols, by being
universally reactive but locally proactive. Hierarchical MANET protocols could be considered a
Hybrid MANET protocol, since they tend to cluster the network and keep routing updates as local
within an area. However, hierarchical MANETs tend to also different on how each nodes are
considered within the structure of the route. More details will be provided within the Hierarchical
Topology MANETs section.
2.3.3.2 Flat topology MANET protocol Examples
This part will provide examples of a standard flat topology MANET protocol, how it functions, what
is its advantages and its downfall.
AODV (Ad hoc On-Demand Distance Vector)
The AODV protocol is a reactive MANET protocol, which utilises Route Requests (RREQs), Route
Replies (RREPs) and Route Errors (RERRs) [37] to maintain routing tables and topologies. RREQs
are broadcasted through the network, when the destination IP address of a routable packet is not in the
node's topology. An RREP will be sent to the requesting node if another node contains the destination
IP on its routing table, or if it is the destination. To avoid infinite loops upon the network, RREQ
contains a recent sequence number, which if a node receiving the RREQ has a sequence number
greater than or equal to the sequence number, it will send a RREP[38].
The AODV protocol tends to have lower overheads in comparison to proactive protocols, due to the
lack of flooding the network regularly for the sakes of updating routing tables. However, since routing
tables are not periodically synchronized, inconsistent routes may occur if sequence numbers are
further apart. Furthermore, the protocol does not optimise hop count (i.e. not bandwidth aware) to
reduce the number of hops.
DSR (Dynamic Source Routing)
The DSR protocol is another reactive MANET protocol, which uses a source route, route cache, route
discovery and route maintenance. When a node sends a packet, it sends a source route which contains
the next hops to the destination route. The route cache is a list of routes which the node has learned,
however if the node does not have a route, it sends out a route discovery to other nodes, which checks
it's cache for a route. Route maintenance is done to see if nearby nodes are within range, if not, it
removes the routes to the nearby nodes and invokes a route discovery to calculate newer routes.
A simulation conducted on this protocol has shown on a small scale network of 24 mobile hosts with
frequent movement of hosts, shows a 1% routing overhead in comparison to data packets
transmitted[39]. This paper shows DSR to scale well with highly mobile nodes and rapid topology
changes, however does show routing overheads increasing as routes become larger[39].
OLSR (Optimized Link State Routing)
The OLSR protocol is a proactive MANET protocol which utilises a link state algorithm and frequent
topology updates. The OLSR protocol uses hello packets and topology control to create, manage and
update the network topology. Hello messages are utilised to discover up to 2 hop neighborhood
information and help with election of multipoint relays. Multipoint relays function like OSPF
designated routers, in which they relay routing information between nodes and control traffic flow.
Multipoint relays help in reducing flooding within the network, hence reducing overheads[40].
Topology Control is used to efficiently calculate and update the topology of the network.
2.3.3.3 Hierarchical Topology MANETs
The hierarchical MANET Topology differs vastly from a Cylindrical type MANET, and in some
respects, a hybrid flat topology MANET. The hierarchical MANET topology is radically different
from the routing topologies introduced in the previous section. Hierarchical MANETs are based on
the realization of clustering, by placing specific nodes within groups [14]. Other routing protocols,
such as OLSR, may have features which designates a node to do specific tasks (i.e. multipoint relay),
but does not radically divide the network. Many of these protocols do not scale as efficiently as a
hierarchical topology MANET, especially when the network becomes more dense, due to increased
protocol overheads and increased interferences with broadcasts [15]. Furthermore, it is argued that
due to the clustered structure and lower routing overheads, bandwidth is increased through the
network [22].
Hierarchical MANETs try to localise routing updates, routing tables and routes as an attempt to
reduce routing overheads. As mentioned, a hierarchical MANET divides the network into
groups/clusters. From then, it allocates a node as a cluster head which does inter-cluster routing. The
protocols to be proposed will further cluster the cluster heads and form more clusters, depending on
the size of the network.
2.3.3.4 Hierarchical MANET protocols
This section will brief on the different kinds of hierarchical MANET protocols found within literature.
Hierarchical Optimized Link State Routing (HOLSR)
The HOLSR protocol is an extension to the OLSR protocol, except it break the structure of the
MANET into clusters and levels. Cluster heads would assume roles for multipoint relay; With
Topology Control and hello packet flooding is reduced down to the clusters, hence reducing flooding
overheads [18]. The levels of clusters are broken down to 3 levels and follow a criteria based upon
several metrics. Some of these metrics include power level, node capacity, number of wireless
interfaces and transmission range. A level 1 node typically is low powered, very low capacity and
may only contain one wireless receiver. A level 2 node would have more wireless receivers, but may
still be lacking in capacity and power. A level 3 node would be the highest capacity, most powerful
and multiple wireless receivers [21]. The image below depicts of a typical HOLSR network [21]. Note
the nodes upon the edges of clusters which overlap; Such nodes would contain the nodes of both
clusters within routing table, thus able to locally route within each cluster.
[21]
Figure 2 Structure of the HOLSR protocol
2.3.4 Routing Overheads
To provide scope within the research, this section shall be dedicated towards defining what routing
overheads are, what routing overheads to expect (i.e. metrics) based upon what other research has
defined routing overheads and why lower routing overheads may not be important.
Routing overheads are the processing requirement for a node within a network to maintain
connectivity. Many of these routing overheads can be seen as constraints within the network and can
increase bandwidth consumption, and energy usage.
The following are commonly discussed overheads, which will be used within the research of MANET
overheads.
2.3.4.1 Number of Neighbours
[30] suggests that the number of neighbours can have an effect on how routing overheads are shaped
within a MANET. [30] argues the more neighbours contained within a nodes network, the higher it
becomes an overhead due to the amount of route_request packets flooded.
2.3.4.2 Number of Hops
The number of hops can have an influence upon the routing overhead of a MANET Protocol. [30]
shows that as a network grows, the number of hops can influence the overhead of a route. This is
mainly affected towards nodes which are further apart from each other, due to high routing path
calculations. Many of the Hierarchical/clustered MANETs look at towards reducing routing tables and
hops for nodes in larger scaled MANETs, since a node would have less neighbouring nodes within
their cluster and only have to route to the clusterhead for intra-cluster routing. The H-LANMAR
protocol could be used reduce hops within a network, however [18] has determined it's inefficiencies
with its certain structure.
2.3.4.3 Route Discovery and positioning
Papers such as [23] [31] argue upon that discovering routing paths and position of nodes can influence
the overheads of routing. [23] uses GPS co-ordinate of nodes to help reduce routing path algorithm
calculations. [31] demonstrates a modified version of the AODV protocol called the Adaptive Request
Zone for Ad Hoc On-Demand Distance Vector (ARZAODV) protocol, which utilises an algorithm to
determine distance and positioning of a node.
2.3.4.4 Frequency of Updates
The frequency of updates refers to the amount of updates a network requires to maintain connectivity
within a network. [2] lists it as a critical feature upon an ad-hoc network.
2.3.4.5 Topology changes
[2] Also questions how the frequency of topology changes can affect the performance of a network.
This should be seen as a major overhead within a network, since topology changes requires each node
to recalculate (i.e. use resources) the topology of the network and send updates towards neighbours.
2.3.4.6 Misbehaving nodes
Misbehaving nodes are nodes which do not process routing updates or route packets as expected
within the network. This can be caused due to an overloaded node, a malicious node or a node
presented with a fault. As shown in [7], misbehaving nodes within a MANET can cause a reduction in
throughput of up to 32%. Misbehaving nodes would also require retransmission of packets, which
contributes to the overhead of a network.
2.3.4.7 Packet flooding
The number of packets required to flood a MANET can produce serious routing overheads within a
MANET. This is due to each node required to process the packets flooded through the network.
Packet flooding may be required, depending on the MANET routing protocol used within the
network. Packet flooding may be required to maintain routes, such as topology updates and route
requests. [32] has argued however that packet flooding may be more efficient for maintaining the
network, if the MANET has a lot of moving nodes.
2.4 Conclusion of literature review
In conclusion to the literature review, the concepts of networking and structured networks has been
outlined on how networks work. Structured networks, whilst desirable for perfect situations, tends to
be costly and high maintenance. For those reasons, the ad-hoc structure is best suited for
environments which produce unpredictable situations. For military use, MANETs are ideal due to the
high mobility and self configuring nature of MANETs. Routing is an important aspect of networks,
since it forms the backbone of end to end connectivity for all devices. Routing protocols chosen must
be suitable for the expect dynamic environment of MANETs. As found in literature, routing
overheads are an important consideration, since it causes constraints with bandwidth. For that, the
routing protocol for a MANET must be tested and chosen carefully to ensure routing overheads do not
cause too many constraints.
The question chosen for this research is What are the tradeoffs between end to end connectivity vs.
routing overheads with cylindrical and hierarchal mobile ad hoc networks, within the context of
urban military use?; The methodology section will clarify upon the question.
3. Methodology
This section will detail through how research will be conducted to answer the thesis question,
What are the tradeoffs between routing overheads and end to end connectivity, with cylindrical and
hierarchal mobile ad hoc networks, within the context of urban military use?
3.1 Question and Explanation
What are the tradeoffs between end to end connectivity vs. routing overheads with cylindrical and
hierarchal mobile ad hoc networks, within the context of urban military use?
The research conducted on Mobile Ad Hoc Networks will have a focus on routing overhead,
compared against full end to end connectivity of MANETs. The research will not define which
MANET structure would suit for military usage, but outline with detail on how each structure copes
with full connectivity and the overheads required to maintain connectivity. Military use is the main
focus upon where such MANETs will be utilised, so research must take into consideration of the type
of movement, terrain, integrity and structure of military activity. End to End Connectivity is defined
as connectivity between hosts (or nodes) within a network.
3.1.1 Sub questions
The following sub questions are to help answer critical aspects of the full questions. Many of these
sub questions are based upon DSTO's expectations of the thesis.
What size should the heirarchy start?
- Use figures based on military division sizes
- Reccomended figure should be a balance between a smaller MANET and a cost effective
- Take into consideration of next size up network (i.e. if 128 node network shows poor
performance, but 64 node network does, 64 may not be optimal, just in case networks have to
merge under emergency situations)
What performance drops are shown when network is scaled up?
- Related to first sub question
- Should help provide answers whether a flat topology MANET is even feasable
Any correlations between routing overheads and connectivity?
- What performance metrics are affected by routing overheads?
What protocol is suited for military use?
- i.e. protocol should perform consistent
3.2 Steps towards research and answering the question
3.2.1 Review literature and determine overheads
The literature must be reviewed in depth to identify what overheads will impact a MANET within a
military setting. The overheads introduced within this proposal are examples of major overheads
which may impact routing performance within a MANET. This process has already been started
through the proposal, introducing concepts of MANETs, routing, routing protocols and specific
MANET routing protocols.
3.2.2 Specify metrics of routing overheads
A criteria must laid out and specify what routing overheads are appropriate for determining its impact
upon the performance of a MANET. Critical routing overheads have been specified within the
proposal, however more will be specified within the thesis.
3.2.3 Finalise routing protocols to be simulated
The routing protocols displayed within the proposal are examples of many MANET routing protocols
and existing structured network protocols. Due to the timeframe of research, very few protocols will
be chosen for simulation. The protocols chosen for simulation will be based on past performance on
previous simulations from other research, which takes into consideration of similar metrics. A
cylindrical MANET protocol will be used to simulate a MANET without full end to end connectivity,
such as the MANET would have a central point, acting as a default gateway. A hierarchical MANET
will be assumed to have full connectivity, due to the way it arranges itself as clusters.
3.2.4 Simulate routing algorithm under different scenarios
Once protocols and metrics have been finalized, routing protocols will be tested and data will be
gathered from such simulations.
3.2.5 Map data to meaningful diagrams/tables, draw conclusions from data gathered.
Once simulations have been finalised, data gathered will be put into meaningful graphs/tables, results
explained and a conclusion will be decided to answer the question given.
3.3 Research method to use
Research method to use: System (Build a system, do simulations of routing protocols)
The research method to use is to build a system which could generate a functioning traffic, similar to
one for use in a military scenario. A system could come under as physical devices, like ones used
within military use or a virtual simulator, simulating the different MANET structures. The systems
should be able to produce relevant data for determining routing overheads and end to end
connectivity. The system must also produce reproducible results in which similar outcomes with the
same configuration proposed within the context of this research.
Utilising a virtual simulator will be the choice for this paper's research, since a virtual simulator will
allow a more desirable scenario. Whilst a realistic system, using radio wave devices, would be able to
produce more realistic results, it also has a few downfalls. Firstly, a sufficient sample size to produce
a military MANET, especially a hierarchical, would not be feasible without a significant budget for
the radio devices, enough participants for a decent sample size or terrain close to a military scenario.
Other research methods are not appropriate for this research, due to either lack of literature utilising
such method or impossibility of gaining relevant data. A case study is not relevant within research on
routing since real life scenarios of MANET routing has not been implemented by the defence force.
Furthermore, even if such MANET routing was implemented, it would be breach of confidentiality,
since such military data would be confidential. Another method is utilising interviews/questionnaires.
Again, such research method would require confidential data leaked. It would also assume the
respondents of the questionnaire/interview would have sufficient knowledge of networking.
Most qualitative research methods do not work, since Quantitive data can easily be reproduced
through a simulation or mock up of the network. Network metrics tend to not be subjective and
conclusions can be drawn clearer if a set of metric results were generated.
3.4 Sampling and population
The sample size of the simulation may need to be compromise between what the DSTO demands and
what is feasible to work within the duration of the thesis. Many papers, such as [9], [14] and [15]
outline tests with flat terrain and smaller sample sizes. No papers found so far have done simulations
with as many nodes as mentioned by the DSTO. This should be taken into consideration, as to
whether the larger amount of nodes would be manageable within the time frame, or are there issues
with the complexity of setting up such simulation.
[4] and [5], whilst conducted with real devices with a small sample size, shows examples of possible
terrain (e.g. hilly) which could affect the operation of wireless within the field.
The simulation of nodes should take into consideration on what the population of end users of such
MANETS and how they manipulate the network. The population of the network is expected to be
mobile, unpredictable and usually utilising low powered.
A Guideline of different sample sizes will be based upon the structure of the Australian defence force.
Information will only be based on what is available on the public domain, thus not leaking any
confidential information.
Army - Two or more corps
Corps - 30,000
Division - 10,000 - 20,000
Brigade - 2,500 - 5000
Batallion - 550 -1000
Company - 100 - 255
Platoon - 30 - 60
Section - 9 - 16
If a hierarchical MANET protocol is feasible to be simulated, then the size of a hierarchical MANET
will be the size of a batallion, with lower level MANETs based on sub units. For example, utilising HOLSR's 3 tier MANET levels, a structure could be of the following...
Size
- Nodes per MANET
- no. of MANETS
Level 3 - 4 nodes
-
1 MANET/Cluster head per MANET
Level 2 - 64 nodes
-
4 MANETS/Cluster head per MANET
Level 1 - 1024 nodes
-
16 MANETS / Cluster head per MANET
If a hierarchal MANET protocol is not feasible then the lower levels of a heirarchy will be assumed
for simulation.
3.5 Data Collection
Collection of data coming from a simulation will require every event from network activity to be
logged and recorded as the simulations take place. The majority of data collected from the network
simulations will be objective (i.e. Quantitive) and should not require subjective explanation of results.
Data will also be collected with 2 - 4 nominated routing protocols (1-2 Hierarchical and 1 - 2
cylindrical MANET protocol), depending on whether it is feasible within the time frame of the
honours thesis. The RIP and OSPF protocol (including the OSPF-BDR protocol) will also be included
within the simulation, to check their plausibility within a MANET.
Data collection should also take into consideration of not only the different MANET structures, but
also different scenarios per MANET structure. Such scenario's should include different terrains the
MANETs would be placed, different combat scenarios (i.e. how the MANET reacts towards nodes
moving rapidly and connecting/disconnecting, how nodes react towards large no. of nodes dropping
out) and how movement of nodes affect the MANET). Scenarios can be provided by utilising
different mobility models.
The MANET should also simulate typical data utilised by combatants. Actual data does not have to be
produced, however metrics used to measure the data will need to be collected. For example, a file
transfer of classified data would need metrics such as accuracy, time taken to deliver, packet losses
and average bandwidth . This may result in the utilisation of sending simple UDP broadcasts between
two nodes.
Alternatively, if hierarchical MANET routing protocols are without reach (i.e. not available freely on
NS-2) then the thesis will focus on aspects of hierarchical routing protocol.
3.6 Data Analysis
The analysis of the data produced by the simulation will require extra study upon metrics for Routing
Overheads and End to end connectivity. The routing overheads section of this proposal shows an idea
of how some typical routing procedures produce overheads within routing. Such information is
critical when analysing data produced from simulation, to allow a clearer conclusion as to what the
data produces. Furthermore, the metrics used will also be well defined with how expected results
influence the conclusion of the research.
The behaviour of packets within the MANET should also be looked at different levels. Layer 3 will be
one of the main focuses in determining routing overheads and end to end connectivity. However,
Layer 8 (Application Layer) will also be important to look at, especially with end to end connectivity.
Layer 8 packets may reveal further information for how end to end connectivity is affected, since we
shall be able to analyse the quality of application used within a MANET.
The mobility of nodes shall also be taken into consideration on how it affect the MANETs. The
mobility of nodes will be analysed on how they work within the MANET, how they influence the
structure of the MANET, how certain nodes (i.e. flapping nodes) influence the routing tables.
Heirarchial MANETs would need extra attention on how inter and intra cluster routing is affected by
topology changes due to moving nodes, especially with higher level cluster heads moving between
clusters. One aspect also of Hierarchical MANETs is to see how the network reacts to rapid changes
of cluster heads, especially if past or present clusterheads had more power to maintain routing
overheads (e.g. Handheld CB Radio vs. Vehicle mounted CB Radio).
Another aspect to analyse is how routing tables are maintained through each MANET structure. It's
important to see how frequent updates occur, how large the updates are (i.e. are the updates unicast or
multicast), how significant the updates are and how large the routing tables of each node. It will also
be in interest to see how routing occurs between nodes within different clusters within a hierarchical
MANET.
The data will also have to be checked to see if they are accurately processed through the simulation, if
the simulation itself has provided an accurate depiction of the different MANET routing protocols, is
wireless signal between nodes simulated properly especially within different terrains and if the
metrics and data collected relevant towards military use?
3.7 Tools
The various items used within the research are as needed: Software to provide network simulation,
source code for the different MANET structures, source code to simulate unreliable/unpredictable
radio links, software to graph the data and a machine capable of producing the simulation. The
network simulation to be used within the research will either be ns-2 or OPNET. ns-2 is a free
network simulation tool, which uses the TCL scripting language to program protocols within the
simulator. It is a commonly used simulator within industry and was the main network simulator used
within the research papers found for this proposal. The alternative is OPNET, which is the simulator
which the DSTO uses for simulation.
Microsoft excel will also be used to table the data and depict it in a more meaningful graphical form,
such as graphs, charts, bar graphs, etc.
To generate a mobility model, the java based tool BonnMotion will be utilsed[41]. This tool is able to
generate many different mobility models for a simulation scenario. BonnMotion supports generating
mobility models in many different platforms (including NS-2).
Traffic for each scenario will be generated using NS-2's built in traffic generation tool. This tool
allows the generation of both UDP and TCP traffic. Both UDP and TCP type traffic will be attempted
through the simulation.
Operating system used for this simulation is Lubuntu 12.04 32-bit. 32-bit operating systems tend to
produce more accurate results for NS-2. NS-2 in 64 bit operating systems tend to produce errors and
inaccurate results.
4. Parameters for simulations
4.1 Selected metrics for Routing overheads and End to End Connectivity
This section will detail on the metrics chosen to help answer the thesis question/sub-questions. The
metrics will be divided between routing overhead metrics and end to end connectivity metrics.
Routing overheads will measure the amount of traffic constraints and routing load of the specific
routing protocol, whilst end to end connectivity metrics will focus on the performance of the routing
protocol and connectivity between nodes.
4.1.1 Routing overhead metrics
Routing overhead metrics will be utilsed to calculate the efficiency of the routing protocol. There will
be no comparisons between routing protocol specific metrics (i.e. number of beaconing packets, such
as OLSR hello packets). This is because the algorithms of selected protocols tend to drastically differ,
thus making a very difficult comparision. For example, AODV utilises Route Request (RREQ) and
Route Replay (RREP) to calculate paths. AODV uses multipoint relays/elected nodes to calculate
routing paths, both produce completely different metrics, thus difficult to compare.
Papers typically usually look at raw routing/control packets sent through the network. This is feasable,
since NS-2 labels the packets type. For example a packet of type "CBR" would indicate a data packet
being traversed through the network, where as one labeled "AODV" or "HELLO" would indicate a
control packet.
The metrics should be an indicator of bandwidth and energy consumed by the protocol. Higher
routing overheads, as mentioned in previous sections, can lead to higher bandwidth consumption, thus
can affect the performance of the network. Energy consumed by the routing protocol would be a
byproduct of the routing protocol utilsing more computation power or utilising more network
hardware interfaces.
While a lower routing overhead figure is desirable, the performance of the routing protocol should not
be hampered for the sakes of lower routing overheads. A MANET can have lower routing overheads,
however, may also have a very low utilisation of bandwidth, low packet delivery ratio, high average
hop count and large end to end delays.
Normalised Routing overhead
Normalised Routing overhead (Also known as Normalised Control overhead or Control Overhead) is
a ratio of routing packets sent vs. data packets sent. Normalised Routing overhead gives a metric
which is compariable with the efficiency of the routing protocol in use. If a routing protocol requires
lesser overheads, it would thus use less bandwidth and require less energy to buffer and process
routing packets. Normalised routing overhead alone does not predict the efficiency of the routing
protocol, thus should be compared with other metrics.
A packet is determined as a Routing packet within NS-2 by reading what type of packet it is.
Ususually, it is a keyword such as "message" (indicating DSDV), "AODV", "DSR", "OLSR" or
"HELLO" (Hello packets sent for OLSR).
Note: ROUTING_PACKET, could substitute for a function which takes the keyword as a parmeter
and returns a boolean value if the packet is a routing packet.
Pseudo code:
set routingPacket to 0
set dataPacket to 0
For Each packet in packetSent
If packetType Equals ROUTING_PACKET Or packetType equals HELLO
Increment routingPacket
Else If packetType Equals DATA_PACKET
Increment dataPacket
normalisedRoutingOverhead Equals routingPacket/dataPacket
4.1.2 End to End Connectivity/Performance metrics
This section will detail on performance/End to End Connectivity metrics. The metrics listed are not all
possible End to End/performance metrics, but available and most commonly used metrics used within
other research. End to End Connectivity metrics will determine the overall performance of the
simulated MANET and how well a connection is defined. Such metrics will be compared toward the
routing overhead metric and look for several factors. One factor is if there is a correlation between
higher (or lower overheads) and performance increases/decreases. Another factor is looking whether
the routing overhead is significant with the performance of the MANET.
If a routing protocol produces high overheads and produces poor end to end connectivity, the routing
protocol will be deemed inefficient.
Average Hop Count
Average hop count is the average number of nodes a data packet must traverse before it reaches to its
destination. Average hop count can be an indicator of how efficient the routing protocol produces
routes between source and destination. Average hop count tends to correlate between average end to
end delay. If a routing protocol produces high average hop count and high delays, then the delay of a
data packet could be of the routing protocol not producing efficient enough routes. Such symptoms
can magnified within high mobile situations.
However, alone it is not an indicator of performance, mainly since calculation of Average hop count
does not take into consideration of bandwidth, latency. Furthermore, if a routing protocols produces
very little hop count, or with zero average hop counts, with a significantly larger network, this could
be a possible indicator of a network with poor convergence (typically in conjunction with low packet
delivery ratios, low delays and low throughput).
Whilst average hop count was introduced within the literature review as a routing overhead, it is
merely a small indicator. However, it does not provide enough data to assume the routing overhead of
a protocol. However, it is a good performance indicator of the network produced by the routing
protocol itself.
Pseudo code:
set packetFowarded to 0
set totalHopCount to 0
For each packet in packetSent
If packet is in FORWARD_STATE
For each node in Nodes
If nodeId is currentNode
Increment packetForwarded
If packet is DATA_PACKET
totalHopCount Equals totalHopCount Plus packetsForwarded
set averageHopCount to 0
averageHopCount Equals totalHopCount / totalPacketsRecieved
Throughput
Throughput is the amount of bandwidth utilised for data packets, which is usually measured in Kbps.
Generally speaking, higher values are better with values closer to the maximum amount of bandwidth
most optimal.
Pseudo code:
throughput = ((totalPacketsRecieved /100 ) * packetSizeBytes) / numberOfNodes
Average end to end delay
Average end to end delay is the time taken for a packet to traverse through the network in
milliseconds (ms). It looks at the time from the source of the packet to delivery toward the destination.
Lower delay is optimal, since it indicates a faster converging network and can expose the efficiency of
the routing protocol's routing capabilities.
Pseudo code:
set startTime to 0
set endTime to 0
set duration to 0
For each packet in packetSent
if packet Equals DATA_PACKET And packet is SENT_PACKET
startTime Equals Packet.Time
Else If packet Equals DATA_PACKET and packet is RECIEVED_PACKET
endTime Equals packet.time
duration Equals endTime Minus startTime
Packet Delivery Ratio
Packet Delivery ratio looks at the number of packets sent through the network and the number of
packets received, then converted to a percentage figure. Packet Delivery ratio can be affected by
routing overheads, mainly due to congestion of routing control packets which may lead to less data
packets sent. This can also be worsened by the MANET increasing in size, which typically requires
larger routing overheads.
The PDR can also be affected by the type of data being sent through the MANET. MANETs which
utilise more TCP data could have a lower PDR ratio, mainly due to extra packet sent out for
retransmission, hence more congestion.
Higher figures are better, however should be compared with the size of the network increasing to
determine where the PDR is greatly affect.
Pseudo code:
packetDeliveryRatio = (packetRecieved/packetSent) * 100
4.2 Mobility Models
This section will focus upon the theory behind mobility models, why they are important and what
purpose they produce for this simulation. The section will also give examples of mobility models
which are commonly used in past simulations and what could be used in future simulations.
4.2.1 What are Mobility Models
A mobility model is an algorithm which is used to generate the movement of a set of nodes within a
Mobile Ad-hoc network simulation. The mobility model is an important aspect of MANET
simulations, since it can provide a near realistic experience of a real life scenario without the need to
specify custom mobility. Mobility models typically look at three certain parameters to define mobility
(velocity, pause time and direction).
Mobility models are common in determining the performance of a MANET routing protocol or
research in algorithms for a MANET protocol. One of the critical aspects of MANET routing
protocols is not how efficient it is, but how it is able to respond to different environments. This puts
the aspect of routing overheads as less of a burden, since a protocol which may have high routing
overheads may also be able to react to different scenarios than other routing protocols. This can allow
one to aspect different algorithms or settings of a protocol to see what tradeoffs there are certain
situations.
Mobility models can also give a near emulation of terrain, with the exception of the ability to change
the effects attenuation or noise between wireless signals. However mobility models can alter the
placement of a node based on constraints from a terrain. For example, the manhattan model is a
VANET specific movement model which co-ordinates nodes into a grid-like street fashion. Such
mobility model would provide the constraints of vehicles speed in traffic, where the vehicle can be
placed (i.e. the road), but cannot duplicate the wireless signal interference caused by tall buildings and
rain.
The rest of the section will briefly go through main different types of mobility models, which will
then lead to the next section, which will provide examples of mobility models
Random models
Random mobility models are ones which have purely random movement with no restrictions. All
parameters such as speed, destination and direction are randomly generated with no influence upon
other nodes. An example is the Random Waypoint model. The Random waypoint takes into
consideration or random velocity, direction and pause time of each node. Random waypoint typically
produces nodes
Due to its wide usage when simulating MANETS on NS-2, Random Waypoint has been chosen for
simulation.
Models with temporal dependency
Mobility models with temporal dependency tend to produce influence upon how a node moves based
on its previous performance. An example of a Model with temporal dependency is the Guass-Markov
movement model.
Models with spacial dependency
Models with spacial dependency tend to have other nodes can affect the mobility of other nodes
4.3 Traffic Generated
This section will go over details on traffic generated for simulations. Since traffic is not the core focus
of the simulation, simplified traffic was generated for the simulation. Traffic does not indicate any
intuitive algorithm or application layer traffic.
Traffic generated will be of a Constant Bit rate (CBR) on the application level over UDP (transport
layer). It is a simplistic packet which simply sends over a MANET and has no timeout clause (i.e.
once sent, it will not be retransmitted if packet times out). Unlike TCP, if there is not data
retransmission of the packet if data from the packet is deemed in accurate (i.e. has artifacts when
presented on the application layer). FTP over TCP was attempted during trial simulations using NS2's traffic generation script. However, issues occurred with traffic occurring on simulations, which
lead to no connectivity. There were known issues with NS-2 and generating TCP traffic upon this
research, thus TCP will not be tested.
The size of the packet will be 512 bytes, the default size of a packet when traffic is generated using
NS-2's traffic generation script.
4.4 NS-2 Parameters
The parameters defined in the table will be universal for all simulations, except for ones which list
multiple variables. Such variables will also be defined in line with the corresponding variable listed.
For example 1500 for x and y would also mean number of nodes would be set to 32 and of 16
connections.
Number of Simulations per protocol = 5.
Physical/Datalink Layer
Channel Channel
Channel/WirelessChannel
Prop
Propagation
Propagation/TwoRayGround
Netif
Network Interface
Phy/WirelessPhy
Mac
Mac Layer
Mac/802_11
Ifq
Interface Priority Queue
Queue/DropTail/PriQueue (CMUPriQueue for DSR)
ll
Link Layer
LL
Ant
Antenna
Antenna/OmniAntenna
Scenario
x
x-co-ordinate size (pixels?)
1000, 1500, 2000, 4000, 8000
y
y-co-ordinate size (pixels?)
1000, 1500, 2000, 4000, 8000
ifqlen
Size of interface queue
100
seed
Random seed no.
0.0
nn
Number of nodes
16,32, 64, 128,256
stop
Time of simulation
100
(seconds)
Mc
Maximum Connections
8, 16, 32, 64, 128
Figure 3 NS-2 Parameters
Channel - Channel
This parameter sets the type of channel. For obvious reasons, this is set to WirelessChannel to denote
the nodes utilise wireless channel.
Prop - Propagation
This parameter defines the path position of the radio signal. Two-ray ground reflection model (as
defined) calculates a line of sight distance (line between transmitted and reciever) and reflected path
distance, which calculates a path which bounces from transmitter, to the ground then to the reciever.
Netif - Network Interface
This parameter refers to the type of physical interface each nodes use.
Mac - Mac Layer
This parameter refers to the protocol used in layer 2/data link layer. The protocol used is 802.11a, a
primitive form of wireless communication.
ifq - Interface Queue
This parameter determines what type if queue is used to store packets with packet buffer of each node.
ll - Link Layer [46] [47]
This parameter defines the built in link layer class to be invoked within the simulation. Keep in mind
the simulation file is a object oriented TCL script, hence would require such to be defined.
Ant - Antenna
This parameter defines the type of antenna which will be simulated by all nodes. OmniAntenna would
therefore define a multi-directional antenna.
x - x-co-ordinate size
Parameter defining the x co-ordinate boundaries of the simulation. Errors will be produced if a node
attempts to enter the simulation outside of the boundary.
y - y-co-ordinate size
Parameter defining the y co-ordinate boundaries of the simulation. Errors will be produced if a node
attempts to enter the simulation outside of the boundary.
ifqlen - Interface queue size
This parameter defines how many packets can be stored in the queue at a time.
Seed - Random seed number
Seed defines the starting point of the randomness of the simulation. Two simulations, provided they
contain same parameters, same generated mobility scenario and same generated traffic file will have
similar results if the seed number is the same. The same simulations with different seed will produce
different results.
nn - number of nodes
This parameter indicates the maximum number of nodes which will be simulated. If a node is
referenced exceeds the number of nodes defined, an exception will occur.
stop - time of simulation
This parameter indicates the duration (seconds) of the simulation
mc - maximum connections
This parameter indicates the maximum number of connections between two nodes which can occur
during a simulation.
4.5 Bonnmotion Parameters
bm -RandomWay [Mobility Model] - d [Time of Simulation] -n [Number of Nodes] -x [Width of
Simulation] -y [Height of Simulation]
4.6 NS-2 traffic generation simulation
Parameter: ns cbrgen.tcl - type cbr [Generates UDP traffic] -nn [Number of Nodes] -mc [Maximum
connections]-seed [Random number generation] - rate 4.0 [interval between packet transmission]
4.7 Simulation constraints
Due to the timeline of this research and the limitations of the NS-2 simulator, this section will outlne
the different constraints brought upon. These constraints could be used by the DSTO as possibilities
for future references, especially for simulations of routing protocols which were proposed within the
literature review, but cannot be simulated.
Within the literature review, there was possibilities of structured network protocols being utilised in
substitution for a designated MANET protocols. Whilst there were factors within literature which
outlined the downfalls of utilising structured routing protocols within MANETS (i.e. RIP, OSPF),
these cannot be confirmed within the simulations for this research. It is quite uncommon for
MANETS to be simulated with standard routing protocols, especially with NS-2. This could be
achieved with other network simulation tools, such as OPNET.
There is also the possibilities of programming the standard routing protocols to work under NS2,.however this would require in depth technical knowledge of each protocol to write code for the
program. This is achieved by writing a C program which contains the algorithm of the protocol. The
algorithm will also have to interface with the simulator itself. Producing the code and interfacing it
with NS-2 would take a significant amount of time, which therefore cannot be achieved within the
timeframe of this thesis.
Further constraints includes the complications of using multiple simulators to compare metrics
between simulators. Different simulators tend to have different outcomes upon the output of results,
primarily based upon the difference within their algorithms. OSPF-MDR was a proposed candidate to
simulate to provide data on how a modified structured network protocol would work within a
MANET. However, it uses a completely different simulator (Quagga). Research at The Ecole
Polytechnique Fed´ er´ ale de Lausanne have tested NS-2, GloMoSim and OPNET by flooding
packets through the network with tweaks towards the MAC layer and comparing metric data
produced. The study showed disparities with results produced, especially at the MAC layer[46].
Furthermore, Quagga is less of a simulator, but a router emulator and lacks features core to the
simulations (i.e. support for Mobility Models).
Further constraints are also with the availability of MANET routing protocols. NS-2 2.35 is bundled
with AODV, DSR, DSDV and TORA. The OLSR algorithm is an implementation as UM-OLSR
(Released under GNU General Public Licence), with no modifications done to the protocol[48].
There currently no publically available implementations of H-OLSR for NS-2. There are several
papers with claims of utilising H-OLSR on NS-2, however are noted to be customised
implementations[47]. This alone has changed the structure of the thesis drastically, which was
originally going to compare hierarchical and non hierarchical MANET implementations.
However, there is evidence within literature which suggests flat topology MANET protocols are not
suitable for larger MANETs. Simulation data on flat topology MANETs can also put forward the
plausibility of flat topology MANETS suitability within larger networks by simulating such larger
networks. By doing so, this could rule out the usage of flat topology networks used within military
MANETs. However, this will require future simulations on hierarchical MANETs to gather more
technical on hierarchical MANETs.
Another constraint within the simulations is varied terrain too complex to set up within the timeframe,
difficult to reproduce and not common in MANET simulations. Varied terrain within the simulation
could be manipulated by generating code which would change bandwidth/connection parameters
within the simulation when a node is at a specific position of the simulation (i.e. hilly terrain). This
would require a customised mobility model and changes to the simulation with geographical
knowledge on how radio waves are attenuated/boosted in certain terrains. This would require more
focus upon the MAC layer, which would push the thesis closer out of scope of the thesis.
Furthermore, many simulations referenced within the literature review do not utilise anything more
than flat terrain. Many either use a random positioned nodes with random movement, or use
something more methodic or academic, such as a mobility model.
5. Simulations
5.1 Protocols for simulation
The following protocols have been chosen for the simulations. Note that none of these protocols are
heirarchy based, however will be used a substitution due to resource constraints. This section will also
go into detail of how each protocol works and what are the expected outcomes for each protocol. The
literature review introduced examples of such routing protocols, this section will go into more detail.
5.1.1 Ad-Hoc On Demand Distant Vector Protocol (AODV)
AODV is one of the distance vector, on demand routing protocols chosen for simulations. It is one of
the most commonly used protocol for research on MANETs. AODV uses route request (RREQ) and
route reply (RREP) packets to generate a routing path. When a data packet is transmitted and routing
path is needed (based on a route with a lower sequence number used to indicate age of route path), A
RREP is sent through the MANET with a Destination IP Address and Originator IP Address (itself),
with a sequence number. Nodes in between the source and destination will send a RREQ, with the
source's IP address within the packet until it has reached the destination. The destination node will
then send a RREP with its IP address as destination recursively to each node until it has reached the
source. This is done without the destination requiring the routing tables or topology information of the
router[46].
0
1
2
3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
Type
J R G D U
Reserved
RREQ ID
Destination IP Address
Destination Sequence Number
Originator IP Address
Originator Sequence Number
Hop Count
Figure 4 A Route Request Packet [49]
0
1
2
3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
Type
R A
Reserved
Prefix Size
Hop count
Destination IP Address
Destination Sequence Number
Source IP address
Time to Live
Figure 5 A Route Reply Packet [49]
If a route cannot be produced, mainly due to an unreachable destination, A Route Error (RERR) is
produced by the node and is marked to be sent back to the source. Because the routing algorithm with
discovering and updating routes is only produced when needed, AODV tends to have reduced routing
overheads. This is more pronounced in low mobility situations, or where connection dropouts are
very minimal. However, whenever mobility increases, so does the routing overhead and performance
of AODV can be hampered. This can also cause delays in packet delivery, due to the process of
AODV's on demand route discovery.
5.1.2 Dynamic Source Routing (DSR)
DSR is similar to AODV in how it is a distance vector, on demand routing protocol. DSR also utilises
RREQ and RREP packets to gather routing path information. The difference is DSR tends to have two
paradigms, route discovery and route maintenance. Route Discovery includes the process similar to
AODV in discovering routes, except each node has a "route source list". whenever a node recieves a
non duplicate RREQ, but is not the destination, it adds the source address to its source route list. This
creates a routing cache of routing paths in which if the node is not the destination, but recieves a route
request, it can send a RREP if it has a know n path to the destination.
This technique is designed to reduce routing overheads, making DSR with one of the least amount of
routing overheads produced. Furthermore, this provides more possible paths, thus should help with
calculation of paths with the least amount of hops. One major difference between DSR and AODV is
DSR uses its route cache to store information, where as AODV utilises a routing table.
Route Maintenance is the aspect of keeping routes up to date and ensuring the topology is up to date
and all routes within the routing cache are routable. Route Maintenance has three condition upon the
inspection of RREQ in which if few are met, a route may be considered outdated. If no conditions are
met, then the route is considered impossible and a Route Error is sent back to the destination.. [50]
DSR has been argued to provide better routing than AODV in larger MANETs and more consistent
routing when mobility increases. However, DSR tends to have delays in propagating data packets
through the network.
5.1.3 Destination Sequenced Distance-Vector routing protocol (DSDV)
DSDV is drastically different in comparison towards the two other protocols chosen (AODV and
DSR). Whilst a distance vector protocol, DSDV is a proactive based protocol. To recap from the
literature review, a proactive protocol is one which periodically updates routing tables, rather than
when a route is needed. The DSDV protocol shows similar characteristics towards the RIP algorithm,
in which it uses the same Bellman-Ford Algorithm.
DSDV sends out a "Message" packet containing the Destination, hop count and sequence number to
help find routing paths to certain. DSDV does this periodically every 30 seconds to ensure routes are
up to date, whether the topology has changed or not. Which differentiates DSDV from RIP is it uses
a sequence number to version routing updates. If a sequence number of a route in the routing table is
lower than one received from a message packet, then the route is considered obsolete and a new
routing path is calculated. This also helps with reducing infinite looping routing paths and broadcast
storms.
0
1
2
3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
Destination Address
Hop Count
Sequence Number
Figure 6 A DSDV "Message" packet [51]
Due to its nature of periodically updating, DSDV would tend to have higher routing overheads than
DSR and AODV, especially in low mobility situations in comparison. However, DSDV has been
shown to have high routing overheads in high mobility situations, even more so than OLSR,
according to a previous comparison between the two protocols [52].
5.1.4 Optimised Link State Routing protocol (OLSR)
OLSR is one of two, proactive routing protocol, but is the only link state protocol chosen within
simulations. OLSR floods it's link state through the network, sharing all routing information to every
node. Whilst link state flooding can produce a better performing network with more accurate given to
every nodes to calculate more efficient paths, it can produce higher routing overheads.
To resolve this issue with higher routing overheads, OLSR uses Multipoint relays (MPR) to take
topology information. A Multipoint Relay (MPR) is a designated node which takes in topology
information and uses it to calculate routes. MPR's also broadcasts packets to be flooded through the
network.
HELLO packets are flooded sprayed out by nodes to neighbouring nodes to find and maintain node
adjacencies. Hello packets also contain link status and neighborhood information of other nodes
within the MANET of up to two hops. HELLO packets are also sent to MPRs to help calculate
topology information, mainly due to information of neighborhood relationships.
Nodes also send out Topology Control messages (usually every 2 seconds) to share information about
neighborhood relationship between other nodes. Topology Control messages are also received by the
MPR who also forwards to each other to help build up routing paths based on neighbourhood
information from other nodes.
OLSR tends to work well with TCP traffic as shown in a previous simulation of a VANET with 50
nodes [54]. Furthermore, OLSR is the only routing protocol tested which has an well researched
hierarchical implementation (H-OLSR). Moreover, OLSR is the only routing protocol which supports
QoS, which was deemed in previous sections as a feature which could potentially help with better
utilisation of low bandwidth, especially if different types of traffic are to be sued.
OLSR however tends to have higher routing overheads, especially at moderate to larger networks,
mainly due to its link state routing and proactive nature.
5.2 Simulation Results
This section will detail upon what simulations have produced and interpretation of such simulations.
5.2.1 Normalised routing load
Normalised Routing Load
Normalised Routing Load
50
45
40
35
30
25
20
15
10
5
0
AODV
DSR
DSDV
OLSR
16
0.0794
0.0794
0.1852
0.8508
32
1.1187
0.6836
0.6491
2.897
64
1.3004
1.2064
0.7782
3.884
128
5.578
4.0186
7.6448
12.3089
256
4.4232
4.8402
7.2245
43.9886
Figure 7 Normalise Routing Load Results
All protocols, show fairly low routing overheads between 16 and 64 nodes, though OLSR's routing
overhead is higher than all protocols at all sample sizes. DSDV, whilst considered to be of high
routing overheads, does not appear to have high routing overheads at larger network sizes. As a matter
of fact, DSDV has lower routing overheads than DSR and AODV (Though, marginally) at 32 and 64
nodes. This could be due to the fact that the mobility model used within the simulation has adequate
amount of movement which makes AODV and DSR having to re-calculate routes more constantly.
One networks reaches at around 128 nodes, routing overheads rapidly increases, especially for OLSR,
which worsens for OLSR at 256 nodes, showing that routing overheads could possibly be
unsustainable for OLSR.
5.2.2 Packet Delivery Ratio
Packet Delivery Ratio
Packet Delivery Ratio
120.00%
100.00%
80.00%
60.00%
40.00%
20.00%
0.00%
AODV
DSR
DSDV
OLSR
16
95.62%
97.35%
69.35%
87.26%
32
42.94%
44.14%
30.74%
37.85%
64
45.35%
47.69%
34.46%
41.15%
128
18.00%
17.66%
4.15%
15.89%
256
2.61%
2.55%
2.60%
2.67%
Figure 8 Packet Delivery Ratio Results
Packet Delivery Ratio for AODV and DSR are nearly perfect at 16 nodes (95.62% and 97.35%
respectively) with OLSR showing decent PDR at 16 nodes. DSDV however shows significantly lower
PDR at 69% at 16 nodes. PDR performance drastically drops at 32 nodes but increases at 64 nodes.
Note this increase is possible due to the different densities of each simulations, since at 32 nodes, the
size of the MANETS were 1500 x 1500 and 2000 x 2000 at 64 nodes, meaning the possibilities of
direct connections skewing results.
PDR also drastically drops at 128 nodes and higher, indicating poor PDR, possibly due to the size of
the network producing bigger routing overheads. This can be explained when looking at hop count, in
which the MANET needs to calculate longer paths. This with mobility put into place has possibly
causes longer routing paths or infinite routing paths (especially with DSDV).
Routing overhead may also have influence upon the PDR figures, since both AODV and DSR have
lower routing overheads in comparison with OLSR. DSDV however skews this claim, since it shows
lower routing overheads at 32 and 64 nodes.
5.2.3 End to End Delay
End to End Delay (ms)
End to End Delay
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
AODV
DSR
DSDV
OLSR
16
0.2406
0.4505
0.0193
0.0174
32
0.2002
0.2876
0.0182
0.0133
64
0.2546
1.1283
0.0143
0.0189
128
0.3779
1.5715
0.0129
0.0492
256
0.0065
0.0068
0.0066
0.0065
Figure 9 End to End Delay Results
End to End delay shows both proactive protocols (DSDV and OLSR) to have less packet delay than
AODV and DSR consistently. Both AODV and DSR show relatively larger packet delays, however
DSR's End to End delay worsens at larger MANETs ,especially at 128 nodes. The decrease of end to
end delay for DSR could be explained by the density of the network increasing, with 1500 x 1500 at
32 nodes vs 1000 x 1000 size at 16 nodes, therefore more direct connections are occurring or more
efficient routes produced (More so towards more direct connections, mainly due to similar hop counts
at 16 and 32 nodes).
At 256 nodes, End to End delay drops dramatically, below even figures given at 16 nodes. This is a
good indicator that only direct connections are occuring, basing on average hop count figures show no
hops produced, low PDR and low utilisation of bandwidth (throughput).
5.2.4 Throughput
Throughput
Throughput (Kbps)
60
50
40
30
20
10
0
AODV
DSR
DSDV
OLSR
16
55.232
56.416
39.904
50.176
32
20.48
21.088
14.592
18.016
64
16.992
17.864
12.84
15.384
128
5.024
4.936
1.16
4.48
256
0.69
0.676
0.686
0.704
Figure 10 Throughput Results
Throughput figures shows correlations with changes in comparison with the PDR, with AODV and
DSR performing better at 16 nodes, OLSR performing quite well and DSDV not so well. Throughput
halves at 32 nodes with performance periodically dropping at larger sizes. MANETS at 128 and 256
nodes shows very little throughput.
5.2.5 Average Hop Count
Average Hop Count
Normalised Routing Load
40
35
30
25
20
15
10
5
0
AODV
DSR
DSDV
OLSR
16
3
5
1
2
32
5
5
2
2
64
7
9
2
3
128
30
37
3
25
256
0
1
0
0
Figure 11 Average Hop Count Results
Hop counts appear to be fair at 16 and 32 nodes with the reactive protocols (AODV and DSR) with
slightly higher hop counts than the proactive protocols (DSR, DSDV). At 64 nodes, DSDV and OLSR
remain consistent, where as AODV and DSR show slight increases. at 128 nodes AODV and DSR
hop count figures rises, with DSDV's figures staying consistent. However, DSDV's figures may has
less to do with efficiency of the routing protocol and more to do with lack of convergence, based on
its throughput and Packet Delivery Ratio. 256 nodes shows all MANETs with little to no routing
activity, since no hops are generated.
5.3 Summary of simulation results
Simulations have provided both reasonable and contradictory results based on assumptions from past
simulations. DSDV appears to have lower routing overheads than expected, despite being a proactive
protocol, in comparison with reactive protocols. This could be due to the mobility introduced into
simulations, which has caused an increase of routing overheads for the reactive protocols.
Furthermore, DSR is claimed to have lower overheads than AODV, however the figures presented in
simulations appear to be negligible. When looking at end to end connectivity metrics, performance of
both protocols are close, with DSR (Except for End to End Delay and Average Hop Count)
performing slightly better.
DSDV could also possibly suffering from routing loops due to outdated routing paths. DSDV only
updates routing paths every 30 seconds, in comparison to OLSR periodically sending out topology
control messages every 2 seconds and DSR/AODV whenever a route is requested. With old routing
paths used, this can cause packets to loop through the network, time out and be dropped.
When looking at what size the lower level MANETs should be when implementing a hierarchical
network, it appears that 16 nodes would perform the most adequate. Majority of protocols and
performance metrics showed performance drops at 32 and 64 nodes, with major performance drops at
over 100 nodes.
With tradeoffs between routing overheads and end to end connectivity, it is important to look at
routing overheads from how they are caused. On the basis of routing overheads caused by the increase
of complexity of the network, then simulations have shown that with larger networks, and with more
complex routing required, routing overheads can affect the performance of the network. However, on
the basis of routing overheads as in the structure of how the protocol works, routing overheads appear
to have less of an effect on performance of the routing protocol.
OLSR, with the highest amount of routing overhead appears to perform more consistent through all
metrics, whereas the other routing protocols perform worse on certain metrics but perform better on
others. It appears performance is more so influenced through the how the routing protocol works,
rather than how much work the routing protocol.
Due to its known support for QoS and well known hierarchical implementation, along with consistent
performance amongst all performance metrics, OLSR appears to be best suited for military MANET
usage.
6. Conclusion, Future Work
6.1 Conclusion
Starting from literature review, MANETS appear to be completely different from a typical structured
network and require more attention on certain aspects. MANETs generally do not performance as well
structured networks, but are desirable due to its self configuration and low maintenance aspects. With
that in perspective, MANETs will also require special attention to what type of routing protocol to be
chosen when looking at the networking aspect. Whilst already established routing protocols, such as
RIP and OSPF are possible candidates, mainly due to wide support and knowledge, they do tend to
fail when put on MANETs. Whilst that was not to proven through simulations due to lack of
availability of such algorithms on NS-2, it was demonstrated through MANET extensions such
protocols which showed the adjacency problem.
When looking at MANETs, one aspect which has popped is the routing overhead, which is the
amount of resources used to maintain routing. What was the main job of the router is now the job of
end devices. Since end devices typically do not have the capabilities of a router, it is important for
such overheads to be reduced. One major influence on routing overheads, as shown in simulations,
was the size of the network, in which the complexity of the size of the topology causes more
overheads and a reduction of performance.
One method of reducing routing overheads is to structure the MANET in a hierarchical fashion, in
which the overall MANET is divided into smaller MANETs. Whilst the theory of hierarchical
MANETs producing less routing overheads and better end to end connectivity is not tested through
simulations, the performance of flat topology MANETs is proven to be hampered as the network is
scaled up.
Simulations on flat topology MANETs showed that the protocol had little effect on routing overheads
in comparison to the size. Sizes were based on different levels of military structure. The thesis also
looked at aspects of simulation and how they can influence results. One aspect was the mobility
model, which was explained to have a similar effect to terrain, only without discrepancies in
connection due to climate. Simulations also showed that the protocol's routing overhead had little
effect on the protocol of the MANET.
6.2 Future Work
The state of the thesis is incomplete, mainly due to lack of resources and time to depict an adequate
answer in regards to hierarchical MANETS. Hierarchical MANETs were not directly compared
within this research, therefore for future research, Hierarchical MANETs should be simulated. This
could be produced through the customization of one of the routing protocols (such as OLSR) on NS-2.
However, there is an implementation of H-OLSR known to be available on OPNET which could be
used to simulate to compare between a hierarchical and flat topology MANET.
It is also shown for optimal performance, MANETs should have as little nodes as possible, thus
requiring the Hierarchical model to support a large number of MANETS. It is current unknown as to
how a node in a higher level MANET could support the number of smaller MANETS. More research,
especially simulations, should look at this aspect.
As for getting a better representation of a military mobility, a better mobility model could be used
based on known movements of combatants. Different mobility models could be customised for
different military scenarios. Minus geographical and climate constraints, a good mobility model can
produce something closer to harsher terrain.
Structured routing protocols are typically not suitable in their current form, unless the protocol is
modified to suite typical MANET conditions. There are several key aspects which could be used to
identify whether the protocol is suitable for MANETS.
Adjacencies
How the routing protocol produces adjacencies to work out optimal routing paths can affect the
performance. Do they try to form adjacencies with every node within a network? Many structured
network protocols try to do that, mainly due to expectations of limited physical ports. In wireless, than
can be dangerous, especially when there are many nearby nodes within the same network.
Flooding of Packets
Many structured MANET protocols flood packets to maintain adjacencies and to discover/maintain
routing. If too many packets are flooded, higher routing overheads will be produced. However, too
little can hamper the efficiency of a routing protocol in the form of in efficient routes, delays, and
poor packet delivery.
Method of reducing Routing Overhead
It is important to see if the protocol attempts to reduce routing overheads, since this can be critical
upon the performance of the overall network. Does it attempt such reduction of overheads? If so, does
it produce any performance gains, or where does the performance decrease? One aspect to look at if it
divides/clusters or forms hierarchies of the network. Furthermore, a simpler network protocol does not
always mean a better performing one. DSDV was one of the first implementations of a MANET
protocol and is the simplest of the 4 tests, however it performed significantly poor with certain aspects
of simulation.
7. References
[1]Teare, D. (2012). Implementing Cisco IP Routing (ROUTE) Foundation Learning Guide. Indiapolis, IN,
USA, Cisco Press.
[2]Subbarao, M. W. (1999). Ad Hoc Networking Critical Features and Performance Metrics. National
Institute of Standards and Technology, Wireless Communications Technology Group: 11.
[3]Schechter, E. (2013). Mobile ad hoc networks end reliance on infrastructure: 1. Accessed 20/03/14.
URL: http://www.c4isrnet.com/article/M5/20131113/C4ISRNET04/311130028/Mobile-ad-hoc-networksend-reliance-infrastructure
[4]Plesse, T., et al. (2004). OLSR Performance Measurement in a Military Mobile Ad-hoc Network.
Proceedings of the 24th International Conference on Distributed Computing Systems Workshops
(ICDCSW’04) - IEEE: 6.
[5]Fischer, J. o., et al. (2012 ). A measurement-based path loss model for wireless links in mobile ad-hoc
networks (MANET) operating in the VHF and UHF band Topical Conference on Antennas and Propagation
in Wireless Communications (APWC). Cape Town, South Africa: 349 - 352.
[6]Wang, H., et al. (2007). IMPLEMENTING MOBILE AD HOC NETWORKING (MANET) OVER LEGACY
TACTICAL RADIO LINKS Military Communications Conference. Orlando, FL, USA 1 - 7
[7]Marti, S., et al. (2000). Mitigating Routing Misbehavior in Mobile Ad Hoc Networks. MobiCom '00
Proceedings of the 6th annual international conference on Mobile computing and networking 11.
[8]Kamal, J. M. M., et al. (2010). Lessons Learned from Real MANET Experiments and Simulation-based
Evaluation of UDP and TCP Proceedings of 13th International Conference on Computer and Information
Technology, Dhaka, Bangladesh.
[9]Spiewla, J. K. Dynamic Routing Protocols and Energy Efficient Communication in Mobile Ad Hoc
Networks. EE 4272 - Computer Networks,Michigan Technological University
[10]T.S.Asha and Dr.N.J.R.Muniraj Network Connectivity based Energy Efficient Topology Control Scheme
for MANET Proceedings of International Conference on Optical Imaging Sensor and Security, Coimbatore,
Tamil Nadu, India.
[11]James, S. (2008, 2008). "OSPF Extensions for Mobile Ad-hoc Networks." from
http://www.cse.wustl.edu/~jain/cse574-08/ftp/ospf/.
[12]Spagnolo, P. A. and T. R. Henderson (2006). Comparison of Proposed OSPF MANET Extensions.
Military Communications Conference. MILCOM 2006. Washington, DC IEEE: 1-7.
[13]Chaparro, P. A., et al. (2010). Supporting scalable video transmission in MANETs through distributed
admission control mechanisms. Euromicro Conference on Parallel, Distributed and Network-based
Processing. Pisa, Italy, IEEE. 18th: 238 - 245
[14]Belding-Royer, E. M. Hierarchical Routing in Ad hoc Mobile Networks, University of California: 22.
[15]Lee, B., et al. (2004). "Issues in Scalable Clustered Network Architecture for Mobile Ad Hoc
Networks." Hhandbook of Mobile Computing.
[16]Chunhua, Z. and T. Cheng (2009). A Multi-Hop Cluster Based Routing Protocol for MANET. 1st
International Conference on Information Science and Engineering (ICISE). Nanjing, China: 2465 - 2468
[17]Park, V. D. and M. S. Corson (1997). A Highly Adaptive Distributed Routing Algorithm for Mobile
Wireless Networks. Proceedings of IEEE InfoCom', Kobe, Japan.
[18]Wang, M., et al. (2007). Comparison of Two Hierarchical Routing Protocols for Heterogeneous
MANET Defence R&D Canada -- Ottawa: 44.
[19]Broch, J., et al. (1998). A Performance Comparison of Multi-Hop Wireless Ad Hoc Network Routing
Protocols. Proceedings of the Fourth Annual ACM/IEEE International Conference on Mobile Computing
and Networking (MobiCom’98),. ACM. Dallas, Texas, USA.
[20]Shiflet, C. F., et al. (2004). Address Aggregation in Mobile Ad hoc Networks. Communications, 2004
IEEE International Conference IEEE. 6.
[21]Villasenor-Gonzalez, L., et al. (2005). HOLSR: A Hierarchical Proactive Routing Mechanism for Mobile
Ad Hoc Networks. IEEE Communications Magazine, IEEE: 8.
[22]Sheltami, T. and H. Mouftah (2003). A Comparative Study of On-demand and Cluster-based Routing
Protocols in MANETs Conference Proceedings of the Performance, Computing, and Communications
Conference, Canada, IEEE International.
[23]Joa-Ng, M. and I.-T. Lu (2000). A GPS-based Peer-to-Peer Hierarchical Link State Routing For Mobile
Ad Hoc Networks IEEE 51st Vehicular Technology Conference Proceedings, Tokyo, Japan, IEEE.
[24]SEDDIK-GHALEB, A., et al. (2009). TCP Computational Energy Cost within Wireless Mobile Ad Hoc
Network Rabat, Morocco: 955-962.
[25]Morgan, Y. L. and T. Kunz (2005). A Design Framework for Wireless MANET QoS Gateway Wireless
Conference 2005 - Next Generation Wireless and Mobile Communications and Services (European
Wireless). 11th European: 1 - 7.
[26]Evans, S. C., et al. (2005). Route Based QoS and the Biased Early Drop Algorithm (BED) Second
International Conference on Quality of Service in Heterogeneous Wired/Wireless Networks. Lake Vista, FL
8 - 17.
[27] Holter, K., et al. (2007). Design and Implementation of Wireless OSPF for Mobile Ad Hoc Networks.
University Graduate Center: 6.
[28] The Associated Press (2003). Battlefield Internet Helps Forces in Iraq ABC News.
[29] Peacock, B. A. (2007). CONNECTING THE EDGE: Mobile Ad-Hoc Networks (MANETs) for Network
Centric Warfare. Blue Horizons Paper, Air War College: 51.
[30] Naserian, M., et al. (2005). Routing Overhead Analysis for Reactive Routing Protocols in Wireless Ad
Hoc Networks, IEEE.
[31] Thipchaksurat, S. and P. Kirdpipat (2011). Position-based Routing Protocol by Reducing Routing
Overhead with Adaptive Request Zone for Mobile Ad Hoc Networks The 8th Electrical Engineering!
Electronics, Computer, Telecommunications and Information Technology (ECTI) Association of Thailand.
Thailand, IEEE. 8: 389 - 392.
[32] Jacquet, P. (2000). Overhead in Mobile Ad-hoc Network Protocols. L. Viennot. LE CHESNAY Cedex
(France), Unit´e de recherche INRIA Rocquencourt. 1: 23.
[33] Guide, C. (2014). "CCNP ROUTE 642-902 :: OSPF | CCNP Guide:." Retrieved 5/5/2014, 2014, from
http://www.ccnpguide.com/ccnp-route-642-902-ospf/.
[34] Ogier, R. (2008). OSPF-MDR: Extension of OSPF for Mobile Ad Hoc Networks.
[35] Ogier, R. G. and (2009). Mobile Ad Hoc Network (MANET) Extension of OSPF Using Connected
Dominating Set (CDS) Flooding. MANET Extension of OSPF. P. Spagnolo, IETF - RFC: 71.
[36] Wu, K. QoS Support in Mobile Ad Hoc Networks. J. Harms. University of Alberta - Computing Science
Department.
[37] Perkins, C., et al. (2003). Ad hoc On-Demand Distance Vector (AODV) Routing. University of
California, Santa Barbara, The Internet Society: 37.
[38] (2007). "Ad hoc On-Demand Distance Vector Routing." Retrieved 20/5/14, 2014, from
http://moment.cs.ucsb.edu/AODV/aodv.html.
[39] Johnson, D. B. Dynamic Source Routing in Ad Hoc Wireless Networks. D. A. Maltz, Computer Science
Department Carnegie Mellon University.
[40] Jacquet, P., et al. Optimized Link State Routing Protocol for Ad Hoc Networks. Le Chesnay Cedex,
France, INRIA Rocquencourt.
[41] Aschenbruck, R. Ernst, E. Gerhards-Padilla, and M. Schwamborn: "BonnMotion - a Mobility Scenario
Generation and Analysis Tool," in Proc. of the 3rd International ICST Conference on Simulation Tools and
Techniques (SIMUTools '10), Torremolinos, Malaga, Spain, 2010.
[42] F. Bai, N. Sadagopan, and A. Helmy, Important: a framework to systematically analyze
the impact of mobility on performance of routing protocols for ad hoc networks, in Proceedings of
IEEE Information Communications Conference (INFOCOM 2003), San
Francisco, Apr. 2003
[43] Bai, F. and A. Helmy A SURVEY OF MOBILITY MODELS in Wireless Adhoc Networks University of
Southern California,U.S.A Southern California.
[44] Henderson, T. (2005). "Network Components in a mobilenode." from
http://www.isi.edu/nsnam/ns/doc/node173.html.
[45] Henderson, T. (2005). "LL(link-layer) Class" from http://www.isi.edu/nsnam/ns/doc/node155.html.
[46] Sasson, D. C. Y. and A. Schiper (2002). "On the Accuracy of MANET Simulators ∗.", Distributed Systems
Laboratory Ecole Polytechnique Fed´ er´ ale de Lausanne (EPFL)
[47] Fasolo, E., et al. (2007). VoIP Communications in Wireless Ad-hoc Network with Gateways. 12th IEEE
Symposium on Computers and Communications,.
[48] Ros, F. J. (2014). "UM-OLSR." from http://masimum.inf.um.es/fjrm/development/um-olsr/.
[49] Perkins, C. (2003). "RFC-3561 Ad Hoc On-Demand Distance Vector (AODV) Routing." from
https://www.ietf.org/rfc/rfc3561.txt.
[50] Grepet, A. C. C. and S. Maag "A Validation Model for the DSR protocol." GET / Institut National des
Telecommunications.
[51] Sterbenz, H. N. Y. C. E. K. C. J. P. R. J. P. G. (2011). "Destination-Sequenced Distance Vector (DSDV) Routing
Protocol Implementation in ns-3." University of Kansas.
[52] Kumawat, R. and V. Somani (2011). "Comparative Analysis of DSDV and OLSR Routing
Protocols in MANET at Different Traffic Load." International Journal of Computer Applications® (IJCA)
International Conference on Computer Communication and Networks.
[53] Ade, S. A. and P.A.Tijare (2010). "Performance Comparison of AODV, DSDV, OLSR and DSR
Routing Protocols in Mobile Ad Hoc Networks." International Journal of Information Technology and Knowledge
Management 2(2).
[54] Kumar, S., et al. (2911). "Traffic Pattern Based Performance Comparison of AODV, DSDV & OLSR
MANET Routing Protocols using Freeway Mobility Model." / (IJCSIT) International Journal of Computer Science
and Information Technologies 2(4): 1606 - 1611.
Download