Quality of Service in mobile ad hoc networks: a

Transcription

Int. J. Ad Hoc and Ubiquitous Computing, Vol. x, No. x, xxxx
1
Quality of Service in mobile ad hoc networks:
a survey
Ash Mohammad Abbas* and Øivind Kure
Center for Quantifiable Quality of Service in Communication Systems,
Norwegian University of Science and Technology,
O.S. Bragstads Plass, 2E,
N-7491 Trondheim, Norway
E-mail: ash@q2s.ntnu.no
E-mail: okure@unik.no
*Corresponding author
Abstract: To support multimedia applications, it is desirable that an ad hoc network has a
provision of Quality of Service (QoS). However, the provision of QoS in a mobile ad hoc
network is a challenging task. In this paper, we present a review of the current research
related to the provision of QoS in an ad hoc environment. We examine issues and challenges
involved in providing QoS in an ad hoc network. We discuss methods of QoS provisioning
at different levels including those at the levels of routing, Medium Access Control (MAC),
and cross layer. Also, we discuss schemes for admission control and scheduling that are
proposed in the literature for the provision of QoS. We compare salient features of various
solutions and approaches and point out directions for future work.
Keywords: ad hoc networks; QoS; quality of service; methodologies; admission control;
scheduling; fairness.
Reference to this paper should be made as follows: Abbas, A.M. and Kure, Ø. (xxxx)
‘Quality of Service in mobile ad hoc networks: a survey’, Int. J. Ad Hoc and Ubiquitous
Computing, Vol. x, No. x, pp.xxx–xxx.
Biographical notes: Ash Mohammad Abbas is a Post-Doctoral Research Fellow under
European Research and Consortium on Informatics and Mathematics (ERCIM) “Alain
Bensoussan Fellowship” at Centre for Quantifiable Quality of Service in Communication
Systems (Q2S), A Center for Excellence, Norwegian University of Science and Technology
(NTNU). He is a Reader at the Department of Computer Engineering, Aligarh Muslim
University, Aligarh, India and is currently on leave for academic pursuit. He obtained PhD
from Department of Computer Science and Engineering, Indian Institute of Technology
Delhi, India. His current research interests include mobile ad hoc networks, quality of service,
and routing.
Øivind Kure is a Professor at Centre for Quantifiable Quality of Service in Communication
Systems (Q2S), A Center for Excellence, Norwegian University of Science and Technology
(NTNU). He received his PhD from the University of California, Berkeley, USA. His current
research interests include wireless ad hoc and sensor networks, Quality of Service, and
routing.
1 Introduction
An ad hoc network can be formed on-the-fly and
spontaneously without the required intervention of a
centralised access point or an existing infrastructure.
An ad hoc network provides a cost effective means of
communication among many mobile hosts. Applications
of an ad hoc network include battlefield communications
where soldier need to decide for a defend or offend, riot
control and law enforcement where only law enforcing
personnel need to communicate while others are not
allowed to do so to prevent spreading of rumours,
emergency rescue missions and disaster recovery where
Copyright © 2008 Inderscience Enterprises Ltd.
the communication infrastructure is abolished. Further,
people may communicate forming an ad hoc network
in convention centres and online conferences and
classrooms without routing their calls to the available
infrastructure. Thus, an ad hoc network may provide
a cost-effective and cheaper way to share information
among many mobile hosts.
The unique characteristics of an ad hoc network
differentiate it from other classes of networks. The
mobile devices are connected through wireless links that
may have several effects such as fading, environmental,
obstacles, etc. The devices used to form an ad hoc
network possess limited transmission range, therefore,
2
A.M. Abbas and Ø. Kure
the routes between a source and a destination are often
multihop. As there are no separate routers, nodes that
are part of the network need to cooperate with each
other for relaying packets of one another towards their
ultimate destinations. The devices are often operated
through batteries, the depletion of battery power may
cause failure of nodes and associated links. The devices
may move about randomly, and therefore, the topology
of the network varies dynamically.
Quality of Service (QoS) means that the network
should provide some kind of guarantee or assurance
about the level or grade of service provided to an
application. The actual form of QoS and the QoS
parameter to be considered depends upon specific
requirements of an application. For example, an
application that is delay sensitive may require the QoS
in terms of delay guarantees. Some applications may
require that the packets should flow at certain minimum
bandwidth. In that case, the bandwidth will be a QoS
parameter. Certain application may require a guarantee
that the packets are delivered from a given source to
destination reliably, then, reliability will be a parameter
for QoS.
As more and more applications are added on top
of an ad hoc network, it is desirable that the network
should provide QoS in some form or the other. However,
the characteristics of an ad hoc network pose several
challenges in the provision of QoS. Some of these
challenges are as follows.
•
The topology of the network varies dynamically.
Therefore, it is difficult to design a scheme or a
protocol that is able to provide hard guarantees
about the QoS desired by an application.
•
The resources of the devices used are limited,
therefore, any such scheme or a protocol should be
a light-weight scheme. In other words, the protocol
should not consume a significant amount of energy
or should not incur a large amount of
computational or communication overheads.
Due to the above mentioned challenges, one may not
expect hard guarantees about the QoS. However, one
would rather be interested in QoS with soft guarantees.
Owing to its importance, a lot of research is directed
to the provision of QoS in ad hoc networks. There are a
few surveys related to the QoS provisioning in a mobile
ad hoc network including an early survey presented in
Perkins and Hughes (2002). A survey of QoS support in
Time Division Multiple Access (TDMA) based ad hoc
networks is presented in Jawhar and Wu (2005). Another
survey about the issues and solutions pertaining to QoS
in a mobile ad hoc network is presented in Reddy et al.
(2006). However, a large part of the survey contains the
description of routing protocols and QoS architectural
frameworks. Surveys of QoS routing solutions for
mobile ad hoc networks are presented in Zhang and
Mouftah (2005) and Hanzo and Tafazolli (2007).
The survey presented in Zhang and Mouftah (2005)
contains discussion on basic problems encountered in
QoS provisioning together with solutions reported in the
literature. The survey presented in Hanzo and Tafazolli
(2007) brings out various factors related to the QoS
provisioning at the level of routing such as design
considerations and trade-offs. However, there is need
for a comprehensive survey on the provision of QoS
in mobile ad hoc networks. Further, we believe that a
lot of research has carried out since the times of the
surveys that appeared in Perkins and Hughes (2002),
Reddy et al. (2006), Zhang and Mouftah (2005) and
Hanzo and Tafazolli (2007), and a lot more issues
have emerged. Therefore, there is a need to revisit the
methodologies presented in the literature with different
issues and perspectives.
In this paper, we describe a comprehensive survey
of the research carried out in this area. We start from
the theoretical background about provision of QoS and
discuss state-of-the-art research carried out in different
directions that address different issues. In this survey, we
have selected only those research papers that appeared
either in high quality refereed journals or in proceedings
of reputed refereed conferences. We have tried to select
examples that represent a group of strategies and help in
understanding a concept.
The rest of this paper is organised as follows.
In Section 2, we describe preliminary concepts related to
the provision of QoS in ad hoc networks. In Section 3, we
present a classification of methods of QoS provisioning.
Section 4 contains a brief overview of architectures for
QoS provisioning in ad hoc networks. In Section 5, we
describe the provision of QoS at the level of routing
and the schemes that address issues involved therein.
In Section 6, we discuss the provision of QoS at the
MAC layer. In Section 7, we discuss cross-layer strategies
for providing QoS in ad hoc networks. In Section 8, we
discuss the research carried out in the area of admission
control. In Section 9, we discuss different scheduling
algorithms used in QoS provisioning. Section 10 contains
the description of research related to the fairness
while providing QoS. Section 11 contains miscellaneous
methods of QoS provisioning. The last section is for
conclusion.
2 Preliminaries
In this section, we describe preliminary concepts related
to the provision of QoS in mobile ad hoc networks.
We first discuss notion of QoS and then categories of QoS
constraints.
2.1 Notion of QoS
There is no universally agreed upon definition of QoS.
In general, QoS means that the network is able to
provide some form of guarantees about the level or
the grade of service. Generally, the level of service is
based on some parameters or constraints, often known
3
Quality of Service in mobile ad hoc networks: a survey
as QoS parameters or QoS constraints. Examples of
QoS constraints include available bandwidth, end-to-end
delay, delay variations or jitter, probability of packet loss,
etc. In what follows, we briefly describe general categories
of QoS constraints.
2.2 Categories of QoS constraints
A constraint makes the job of a protocol more stressful
as compared to the scenario when there is no constraint
specified. For example, in case of routing, a route will
be considered if it satisfies the specified constraints.
The overall value of a constraint from a source to a
destination may be expressed in terms of the values
of its constituents. Let there be a multihop path
between nodes u and v consisting of nodes u1 , u2 , . . . , uk .
Let c(i, j) denotes the value of constraint c between
nodes i and j or link (i, j). Alternatively, the value of
a constraint along a path depends upon the individual
values of the constraint along the links that form the
path. Based on how the values of path constraints
are related to the values of their corresponding link
constraints, QoS constraints are classified into the
following three broad categories (Chao and Guo, 2002).
•
Additive: A constraint whose overall value is
summation of the values of its constituents. In other
words,
Figure 1 shows a pictorial representation of QoS
constraints and their combinations that fall in NP
and P classes of problems. The symbols used are as
follows: A for additive, M for multiplicative, and C
for concave. Note that finding an optimal path subject
to a combination of two or more additive and/or
multiplicative1 constraints is an NP complete problem.
As a result, the problem of selecting any combination
of two or more QoS parameters (such as delay, jitter,
hopcount, and loss probability) and optimising them is
an NP-complete problem. The only combinations of QoS
parameters or metrics that are computationally feasible
are bandwidth (which is a concave constraint) and any
one of the additive/multiplicative constraints mentioned
above. However, approximation algorithms exist for a
combination of two or more constraints. Further, a
commonly used method for multiple constraints to be
satisfied is sequential filtering. In sequential filtering, one
first examines whether paths between a given source and
a destination satisfy a QoS constraint. The subset of
the paths that satisfy the first QoS constraint is further
examined whether it satisfies the second QoS constraint,
and so on.
Figure 1 Constraints and their combination that lie
in P or NP
c(u, v) = c(u, u1 ) + c(u1 , u2 ) + · · · + c(uk , v).
For example delay, jitter, hopcount are additive
constraints.
•
Multiplicative: A constraint whose resulting value is
a product of the values of its constituents. In other
words,
c(u, v) = c(u, u1 ).c(u1 , u2 ). . . . .c(uk , v).
For example, reliability, and the probability of
packet loss are multiplicative constraints.
•
Concave: A constraint is concave if
c(u, v) = min{c(u, u1 ) + c(u1 , u2 ) + · · · + c(uk , v)}.
For example, bandwidth along a path is minimum
of the bandwidths of the links that constitute the
path. Therefore, bandwidth is a concave QoS
constraint.
Note that sometimes one may not be able to devise
methods to identify paths that satisfy a combination
of QoS constraints. In other words, there may exist a
combination of QoS constraints that may not be satisfied
by an underlying algorithm in a reasonable amount of
time. Such a problem falls into a specific class of problems
called NP-complete problems. If a combination of QoS
constraints can be satisfied within a reasonable amount
of time, then such a problem is said to be a polynomial
(P) class of problem. There is a specific procedure to
determine whether a problem is in P or NP.
In addition to the usual network operations, the
functionalities that are to be incorporated for QoS
provisioning are (as described in Kurose and Ross (2006))
as follows:
•
traffic classifier
•
resource reservation
•
scheduling
•
admission control.
As mentioned earlier, these are general principles that
may apply to any network for the provision of QoS
irrespective of whether the network is wired or wireless,
and whether the network possess an infrastructure or
does not possess any infrastructure. In case of mobile
ad hoc networks, there are a lot of issues and challenges
in providing QoS.2 Some of them have already been
mentioned in the previous section.
In what follows, we elaborate the issues and challenges
in QoS provisioning in mobile ad hoc networks.
4
2.3 Issues and challenges
As mentioned earlier, in an ad hoc network the
mobile devices are connected through wireless links
that are more prone to errors as compared to their
wired counterparts. There are problems such as hidden
terminal, multipath fading, etc. As opposed to a wired
network, there are no separate routers, therefore, the
mobile devices need to route packets of one another
towards their ultimate destination. Generally mobile
devices are equipped with omni-directional antennas, and
therefore, transmission of a node are heard by nodes in
its vicinity. This causes one more issue – nodes need to
coordinate among themselves for transmissions through
a shared channel. In other words, a node cannot decide
on its alone about the time of the beginning of a
transmission because the channel might be occupied by
another node in its vicinity. Therefore, the time taken
in waiting for the transmission depends upon who are
the other neighbouring nodes contending for the channel.
However, their is no such issue in case of wired networks
as the channel, therein, is not shared. Note that there may
be multiple hops from a given source to a destination in
an ad hoc network and at each hop nodes may contend
for the channel. Due to channel contention, it is difficult
to provide any hard guarantees about the end-to-end
delays.
On the other hand, the topology of an ad hoc network
varies dynamically due to either movement of mobile
devices or depletion of battery power. It may affect QoS
guarantees provided by the network because a change in
the topology of the network may require to rediscover
the routes adding to the latencies and thus affecting the
QoS. It may also happen that the newly discovered routes
are longer than the routes available before the topological
change which will affect the QoS more stringently, as
the resources that were reserved for a flow before the
topological change are no longer reserved, they have
to be reserved along newer routes. It may also happen
that the amount of resources required by the flow or
application is no longer available, adding further latencies
and affecting the QoS. Therefore, another issue involved
in the provision of QoS in mobile ad hoc networks is how
to tackle changes in the topology of the network.
Another issue in case of mobile ad hoc networks
is that the resources of participating nodes are limited.
Therefore, a protocol (that is considered to be efficient
for wired or networks with some kind of existing
infrastructure) that requires extensive computations and
communications may not be a good option in such
networks. Therefore, a protocol for providing QoS in
ad hoc networks should be light-weight as far as possible
and should be able to utilise resources in an efficient and
effective manner.
Note that QoS can be provided, in some form or the
other, at different layers of the protocol stack. At what
layer and in what form the QoS is provided depends upon
the requirements of an application. Depending upon the
actual QoS requirements of an application, the issues
involved in providing QoS are different. Also, the actual
issues involved can be specific to the method used for QoS
provisioning.
In what follows, we discuss methods of classification
of different methodologies for providing QoS in mobile
ad hoc networks.
3 Classification of methodologies
As we pointed out earlier, the issues involved in QoS
provisioning are different depending upon the type of
and the requirements of an application, the methodology
of QoS provisioning, and the layer of protocol stack at
which the QoS is to be provided. In this section, we
wish to classify the methodologies of QoS provisioning
in mobile ad hoc networks and then we discuss different
methods reported in the literature in each class.
Figure 2 shows a classification of methods of
QoS provisioning in mobile ad hoc networks that we
call layered classification. In general, depending upon
at which layer of TCP/IP protocol suite the QoS
provisioning is implemented, one can classify methods of
QoS provisioning into the following categories:
•
MAC layer
•
network layer
•
cross layer.
Figure 2 A layered classification of methods of QoS provisioning in ad hoc networks
Based on the methodologies used, these categories can
be further divided into subcategories. For example,
MAC layer schemes are divided broadly into three
subcategories:
•
IEEE 802.11
•
TDMA
•
CDMA.
There can be several other MAC layer protocols as well.
Cross layer strategies are also divided into three sub
categories, namely, scheduling, multirate, and resource
allocation. The strategies at the network layer consists
mainly of routing protocols which can be divided into
unipath and multipath categories.
On the other hand, schemes can be classified based on
the major functionalities required for QoS provisioning.
We call it a functional classification, and is shown
in Figure 3. The major categories that are part of
such a classification are routing, MAC, differentiation,
admission control, resource reservation, and scheduling.
There can be different subcategories. For example,
scheduling can be divided into rate based scheduling,
prioritised scheduling, opportunistic scheduling, etc.
There can be other classifications as well. For
example, there can be a parametric classification
in which methods are classified based on the QoS
parameters such as delay based, bandwidth based,
throughput based, etc. Another classification can be the
behavioural classification in which methods are classified
based on the manner in which QoS is provided such as
per hop, per flow, per class, etc.
On the other hand, we would like to mention that
none of these classifications is an absolute classification.
In other words, there might be a method that is classified
to fall in a category, however, it may happen that the
method incorporates or uses other methods that belong
to other categories. However, we believe that different
works are carried out by different researchers with
different objectives and issues in mind, and therefore,
a classification simply helps in studying them in a bit
5
organised manner. Therefore, it is not essential for a
classification to classify methods into absolute levels or
categories.
In what follows, we review the research reported in the
literature pertaining to the major categories. We would
like to mention that in the ensuing discussion we do not
strictly follow either layered or functional classification
for the sake of simplicity.
4 QoS architectures
In general, an architectural framework consists of a
group of modules that are required for the QoS
provisioning in an ad hoc network e.g., routing,
MAC layer, admission control, resource reservation, etc.
We describe here some of the architectural frameworks
proposed in the literature for the provision of QoS in
mobile ad hoc networks.
An IP-based framework, called INSIGNIA,3 for
providing QoS in mobile ad hoc networks is presented
in Lee et al. (2000). It uses in-band signalling. The term
in-band signalling means that the control information
is carried with data, and there is no separate control
channel as opposed to another type of signalling called
out-of-band signalling where control information and
data are sent on separate channels. The framework
is aimed to provide adaptive QoS guarantees to an
application. By adaptive, we mean that there is a
minimum QoS that has to be provided to an application.
The level of QoS can be enhanced later if resources
required to support the enhanced QoS are available.
The architecture has several modules that are routing,
in-band signalling, admission control, packet forwarding
or scheduling, MAC protocol, etc. However, it is a
stateful architecture because it uses soft state resource
management scheme to utilise the resources.
An architecture for the provision of QoS in MANETs
is called Stateless Wireless Ad hoc Network (SWAN)
Ahn et al. (2002). It supports service differentiation for
realtime and best-effort traffic. The proposed architecture
Figure 3 A functional classification of methods of QoS provisioning in ad hoc networks
6
Table 1 A comparison of architectural frameworks for QoS provisioning
Architecture
Features
Comments
INSIGNIA (Lee et al., 2000)
SWAN (Ahn et al., 2002)
Adaptive QoS
Service differentiation
FQMM (Xiao et al., 2000)
Combination of differentiated
and integrated services
Soft state resource management
Does not require QoS support
from MAC layer
Scalable
handles both the realtime traffic and the best-effort
traffic. Local rate control is used for handling the
best-effort traffic of TCP, and a sender-based admission
control is used for the realtime traffic of UDP. A merit of
SWAN is that it is independent of the underlying MAC
layer. Alternatively, it does not require a QoS support
from the MAC layer. It can be used for QoS provisioning
even if the underlying MAC layer provides a best-effort
service.
Another architectural model called Flexible QoS
Model for Mobile ad hoc networks (FQMM) is presented
in Xiao et al. (2000). It provides a combination of
aggregation of traffic classes as that of differentiated
services model and at the same time it possesses the per
flow granularity of integrated services model. Specifically,
high priority flows are provided per flow QoS and lower
priority flows are aggregated into a set of service classes.
The hybrid treatment of the traffic is based on the
intuition that the fraction of flows with high priority and
that need per flow QoS is much less as compared to flows
with low priority. However, the service level of a flow
may change dynamically from per flow to per class and
vice versa depending upon the traffic load in the network.
The framework is scalable and provides a basis for future
models to come up with better QoS guarantees.
A comparison of these architectural frameworks for
QoS provisioning in ad hoc networks is given in Table 1.
Future work
There are several issues that can be addressed in future
such as how to decide about the classification of traffic
and what percentage of traffic can be treated on per flow
basis. Also, none of these architectures takes into account
how to sustain a given level of QoS in the presence of
mobile nodes or how a source node can predict what
level of QoS it may receive based on the current network
conditions. All these issues may be addressed in future.
In what follows, we review the research related to the
provision of QoS at the level of routing.
5 Routing
There is a lot of work reported in the literature in which
the provision of QoS in an ad hoc network is embedded
in the routing protocol itself. The provision of QoS at the
level of routing can be divided into different categories
for the sake of simplifying the discussion.
5.1 QoS routing protocols
Several protocols are proposed in the literature with a
provision of QoS at the level of routing. Their objectives
and aims differ based on their applications and strategies
used. In what follows, we discuss some of them.
5.1.1 Bandwidth estimation based routing
Generally, routing protocols find a feasible route from
a source to a destination without taking into account
current traffic in the network or specific requirements
of an application. As a result, the network may
become overloaded and the requirements of a realtime
application that requires a support for QoS may not be
met. Therefore, there seems a need to estimate the traffic
in the network so that a feedback can be provided to the
application in case when its requirements cannot be met.
A QoS aware routing that is based on the bandwidth
estimation for mobile ad hoc networks is proposed in
Chen and Heinzelman (2005). The protocol incorporates
an admission control scheme together with a feedback
scheme to meet the QoS requirements of realtime
applications. The QoS routing protocol is based on
Ad hoc On-demand Distance Vector (AODV) routing.
The protocol is based on the intuition that the end-to-end
throughput of a route depends upon the minimum
end-to-end residual bandwidth available along the route.
To estimate available residual bandwidth, two methods
are used. In the first method, hosts listen to the channel
and estimate the available bandwidth that is based on
the ratio of the time for which the channel is free and
the time for which the channel is busy. This estimate
is called ‘listen’ bandwidth estimation. In the second
method, every host disseminate the information about
the bandwidth used by it currently. This dissemination is
carried out through ‘hello’ messages. A host estimates its
available bandwidth indicated in the ‘hello’ messages that
it receives from its two hop neighbours. This is called the
‘hello’ bandwidth estimation. The performance of ‘hello’
bandwidth estimation comes out to be better as compared
to the ‘listen’ bandwidth estimation when releasing the
bandwidth is required immediately. However, ‘listen’
bandwidth estimation does not incur an additional
overhead as compared to ‘hello’ bandwidth estimation.
The reason is that in ‘hello’ bandwidth estimation, the
information about bandwidth consumed by neighbours is
attached in ‘hello’ messages, and these ‘hello’ messages
are sent periodically.
A QoS routing protocol called Bandwidth
Reservation in Ad hoc Wireless Networks (BRAWN) is
proposed in Guimaraes et al. (2009). BRAWN is also
based on bandwidth estimation, however, it is designed
for multirate networks. In a multirate network, nodes
can choose among several modulation schemes and can
communicate to neighbours using different transmission
rates depending on channel conditions. BRAWN
provides a network layer solution i.e., no modification
is required at the lower layers. The scheme employed
computes the available bandwidth at a node which is
then used to accept or reject a flow.
5.1.2 Interference aware routing
A QoS routing protocol that guarantees the bandwidth
for ad hoc networks with interference considerations is
presented in Jia et al. (2005). The protocol addresses
the problem of Ad hoc Shortest Widest Paths (ASWP)
routing. A routing protocol that is aware of the
interference called as Interference Aware QoS Routing
(IQRouting) is proposed in Gupta et al. (2005).
In IQRouting, several paths are probed using flow
packets in a distributed fashion for satisfying QoS. The
paths that satisfy the QoS are known as candidate paths.
The path that is the best in terms of the QoS amongst all
candidate paths is chosen by the destination node.
7
5.3 Backbone based QoS routing
A QoS routing protocol called Core-Extraction
Distributed Ad hoc Routing (CEDAR) is proposed in
Sivakumar et al. (1999). The protocol is aimed to satisfy
bandwidth requirements of a flow from a given source
to a destination. The protocol consists of three steps
namely core extraction, link-state propagation, and
route computation. In the core extraction step, a group
of nodes are elected to form the core of the network in a
dynamic and distributed fashion by using an approximate
algorithm for minimum dominating set. A core node is
responsible for maintaining the local topology of the
nodes in its domain and it also computes the routes for
these nodes.
A routing protocol called QoS Mobile Routing over
Ad hoc On-demand Distance Vector routing (QMRBAODV) is proposed in Ivascu et al. (2009). The
main motivation behind the protocol is that a realistic
MANET is likely to be heterogeneous. In other words,
there are some nodes that are rich in resources while
others may not be rich enough. The protocol constructs
a routing backbone consisting of nodes that are rich in
resources. These backbone nodes are responsible to route
packets to end nodes. Figure 4 shows a network with a
backbone.
Figure 4 Backbone or core based QoS routing
5.2 Position based QoS routing
A QoS routing protocol called Geographical Vehicular
Grid (GVGrid) for Vehicular Ad hoc Networks
(VANET) is presented in Sun et al. (2006). GVGrid is an
on-demand and a position based routing protocol that
identifies a route from a source node (which is a fixed
base station) to vehicles that lie in a destination region.
The geographical region is divided into squares of equal
size and is called a grid. An underlying assumption in
GVGrid is that every vehicle possesses a digital map and
knows its geographical position and the direction of its
movement through a Geographical Positioning System
(GPS). The intermediate nodes remember the route till it
is broken. The quality of route is determined based on:
•
the lifetime
•
the packet arrival ratio.
The lifetime of a route is the duration between the time
instances when a route was established till a link on
the route is failed. The packet arrival ratio is the ratio
of the packets that have arrived at the destination and
the packets that were sent by the source. It has been
shown through simulations that GVGrid performs better
in terms of lifetime as compared to the contemporary
protocols.
A position based QoS routing for Ultra-Wide Band
(UWB) based ad hoc networks is proposed in Abdrabu
and Zhuang (2006). However, this protocol uses cross
layer properties, therefore, we shall discuss it while
describing cross layer solutions.
It is worth pointing out that the overheads of movement
of an ordinary mobile node will be much less than the
overheads involved when a node that is part of the core or
the backbone of the network moves. Therefore, protocols
that are based on constructing a core or the backbone
may not allow the backbone nodes to move, otherwise,
the backbone is required to be constructed a fresh.
5.4 Multipath QoS routing
A single route from a given source to a destination may
not be able to provide enough resources required by
an application. Therefore, a protocol should be capable
8
of identifying multiple paths from a given source to a
destination.
A QoS routing protocol called Ticket Based Probing
(TBP) is proposed in Chen and Nahrstedt (1999). A ticket
is a permission for an intermediate node to search
exactly one path. The source sends probes towards the
destination to search for a low cost path that satisfy the
QoS constraint. Each probe is required to carry at least
one ticket and a probe with more than one tickets is
allowed to split at an intermediate node each searching
a different downstream subpath (see Figure 5). As a
result, there can be multiple paths from a given source
to the destination. However, multiple paths searched may
not satisfy any kind of disjointness. The protocol avoids
flooding by the ticket mechanism and is able to handle
sessions with either delay or bandwidth constraints.
There are multiple levels of path redundancies so as to
direct traffic through an alternate path in case of failure
of a link along an existing path. However, the protocol
needs a beacon mechanism to know the neighbours and
beacon packets are transmitted periodically, therefore,
the overhead may be significant.
Figure 5 Ticket based probing: probe p1 has only one ticket
while p2 has two tickets
A routing scheme, in which different connections from a
source to a destination may request different bandwidths
through different alternate paths, is presented in Jia
et al. (2004). The proposed multipath routing deals
with inaccurate link state problem. Specifically, two
categories of K-shortest routing algorithms, hop-based
and bandwidth-based are proposed, and five categories
of path selection algorithms are presented, which are:
Best-K-Widest (BKW), Random-K-Widest (RKW),
Shortest-K-Widest (SKW), Best-K-Shortest (BKS),
and Widest-K-Shortest (WKS). In order to reduce the
amount of flow of link-state update information, three
schemes are proposed that are based on the distance
of the destination from the source. These schemes are:
Deterministic Frequency Reduction (DFR), Probabilistic
Frequency Reduction (PFR), and Hop/Distance
Threshold (HDT).
5.5 Multicast QoS routing
The issues involved in multicast routing with a provision
of QoS are investigated in Wu and Jia (2007). Therein,
authors use a concept of ‘node bandwidth’. Note that
node bandwidth might be related to the transmission rate
by a node. As the bandwidths at the nodes of an ad hoc
network are limited, a multicast call can be blocked
despite the fact that enough bandwidth is available in the
system so as to support the call. This is because a single
multicast tree does not exist at the node to support the
call. Authors in Wu and Jia (2007) proposed a multicast
routing scheme by using either multiple paths or multiple
multicast trees so that the bandwidth requirement of a
call are met. Therein, three multicast routing schemes are
proposed named as Shortest Path Tree Based Multiple
Paths (SPTM), Least Cost Tree Based Multiple Paths
(LCTM), and Multiple Least Cost Trees (MLCT). The
routing trees obtained using these strategies can meet the
user’s requirements in terms of delays and bandwidth.
5.6 QoS routing with resource allocation
A framework for generalised QoS routing with resource
allocation is presented in Bashandy et al. (2005). The
framework combines routing with the allocation of
resources along the routes. It employs a dynamic
programming algorithm that tries to find an optimal
path between a given source and a destination and
computes the amount of resources required at each
intermediate node so as to satisfy the end-to-end QoS
requirements along the path. The QoS parameters
considered are end-to-end delay, jitter, reliability, and
bandwidth. These QoS parameters are represented as
functions of resources rather than a fixed value metrics.
Further, the applicability of the framework to rate
based service disciplines (such as GPS, packet by Packet
Generalised Processor Sharing (PGPS), and Self-Clocked
Fair Queueing (SCFQ)) is also investigated.
In what follows, we discuss multiple constraint based
QoS routing.
5.7 Constraint based QoS routing
A problem that is considered to be a part of the QoS
provisioning in a network is that how to identify paths
that satisfy two or more constraints. As mentioned
earlier, the problem of finding paths with two or
more additive/multiplicative QoS constraints is an NP
complete problem. However, the problem of finding
paths with one additive/multiplicative constraint and
a concave constraint is not an NP complete problem.
For example, one can find paths that satisfy two QoS
constraints – delay and bandwidth. Note that delay is
an additive constraint and bandwidth is a concave (or
min-max) constraint. In Wang and Crowcroft (1996),
an algorithm for bandwidth-delay based QoS routing is
presented. The algorithm first removes all links which
9
do not satisfy the bandwidth constraint and then finds a
shortest path (in terms of delays) in the resultant graph.
Due to the NP completeness of the QoS routing
with two or more constraints, often either approximate
algorithms or heuristics are used in place of exact
algorithms. However, there are papers Mieghem and
Kuipers (2004), Kuipers and Mieghem (2005), Mieghem
et al. (2001) which favour to use exact algorithms
instead of heuristics. The reason given is that the
NP-completeness of QoS routing problems with multiple
constraints occurs in specially constructed network
topologies and such topologies are unlikely to occur in
practice in a realistic network. For topologies that do not
show the NP completeness, the complexities of heuristics
and exact algorithms can be comparable.
A hop-by-hop QoS routing called Self-Adaptive
Multiple Constraint Routing Algorithm (SAMCRA) is
proposed in Mieghem et al. (2001). In Mieghem and
Kuipers (2004), some underlying concepts of an exact
QoS routing are explained and are incorporated in
SAMCRA. One of the concept is nonlinear definition of
the path length. In the usual linear definition, individual
path lengths are simply added to give the resultant path
length. In nonlinear definition, the resultant path length is
1/q power of the summation of the q powers of individual
path lengths. Other concepts incorporated in SAMCRA
are k-shortest path approach, non-dominance of paths,
and look ahead. These concepts may serve as basic
building blocks of a multiconstrained routing algorithm.
The authors of Mieghem and Kuipers (2004) extend
a similar theme in Kuipers and Mieghem (2005) in
which they have the view that exact QoS routing is
tractable in practice. Specifically, the NP completeness
of a multiconstrained routing problem hinges upon the
following conditions:
•
•
•
the topology
the granularity of link weights
the correlation among link weights
•
the constraints.
Therein, the impact of these conditions on the complexity
of QoS routing is studied through mathematical analysis
which is further validated using simulations.
In Xue et al. (2007), Fully Polynomial Time
Approximation Schemes (FPTAS) for finding a path
between a source to a destination subject to many
additive QoS constraints are presented, which are as
follows. The first FPTAS uses a single auxiliary edge
weight to compute a shortest path. The algorithm has
a time complexity of O(Km + n log n) where K is the
number of additive constraints, n is the number of nodes,
and m is the number of links. It is a K-approximation
algorithm and can be used in hop-by-hop routing
protocols. The second FPTAS is for an optimised
version of QoS routing problem and has the time
K−1 complexity of O n
. The third FPTAS is also for
an optimised version of QoS routing problem and has
K−1 a time complexity of O n log n + m H
, when
there exists an H-hop path satisfying K number of
QoS constraints. However, for implementing the FPTAS
proposed in Xue et al. (2007), some modifications are
required. For example, the routing tables are required
to store the next hop addresses for every source and
destination pair for some discrete values of , and the
values of can be used to determine the classes of traffic
in the network.
A framework for sufficient rate constraints for QoS
flows in ad hoc networks is presented in Gupta et al.
(2007). It is a theoretical model that predicts the capacity
of an arbitrary ad hoc network. The proposed model uses
a concept of a specialised graph called a conflict graph.
A conflict graph represents the interference relationships
between different links of the network. Specifically, a
link in the network graph is represented by a node in
the conflict graph. There is a link between two nodes in
the conflict graph iff the corresponding two links in the
network graph interfere with each other. Therein, authors
proposed two sets of constraints – row constraints, and
clique constraints. In a network where some nodes use
row constraints and others use clique constraints, the
union of constraints being satisfied across the network
constitutes sufficient conditions for a flow rate vector to be
feasible. The row and clique constraints can be computed
in a distributed fashion using localised information about
the topology of the network. Distributed algorithms
for capacity estimation that utilise these constraints are
proposed. Further, the capacity estimated may be used to
implement admission control and QoS routing schemes
in an ad hoc network. A comparison of QoS routing
protocols is shown in Table 2.
Table 2 A comparison of QoS routing protocols
Protocol
Features
QoS parameter
Comments
QoSR-L (Lin and Liu, 1999)
TBP (Chen and Nahrstedt, 1999)
BE (Chen and Heinzelman, 2005)
Bandwidth
Delay/Bandwidth
Bandwidth
GVGrid (Sun et al., 2006)
Bandwidth guarantees
Ticket based probing
Bandwidth estimation
Based on AODV
Position based routing
ASWP (Jia et al., 2005)
IQRouting (Gupta et al., 2005)
CEDAR (Sivakumar et al., 1999)
QMRB-AODV (Ivascu et al., 2009)
BRAWN (Guimaraes et al., 2009)
Interference considerations
Interference aware
Core extraction
Backbone construction
Computes available bandwidth
Time slot assignment
Multiple paths
Hello bandwidth estimation
Listen bandwidth estimation
Used in Vehicular
Ad hoc Network (VANET)
Ad hoc Shortest Widest Path
Use of candidate paths
No reliance on TDMA
Traffic distribution
Multirate ad hoc networks
Lifetime
Packet arrival ratio
Bandwidth
Bandwidth
Bandwidth
Throughput
Bandwidth
10
Future work
In future, one may address the issue of a predictable
QoS. By the term ‘predictable QoS’, we mean that the
source should be able to predict upto what extent the QoS
may be fulfilled along a route. Ofcourse, the predictable
QoS may depend upon the network conditions such as
node mobility, node and link failures, and interference.
One may also address upto what extent the QoS can
be sustained despite the changes in the topology of the
network.
6 QoS at MAC layer
In this section, we describe the research reported in the
literature related to the provision of QoS at the MAC
layer.
6.1 Types of MAC layers
We consider the following major MAC layers for ad hoc
networks– Carrier Sense Multiple Access with Collision
Avoidance (CSMA/CA), IEEE 802.11, Time Division
Multiple Access (TDMA), Code Division Multiple
Access (CDMA), and their variations that are reported in
the literature.
6.1.1 CSMA/CA based networks
CSMA is one of the most widely used MAC layer in
ad hoc wireless networks. A contention based MAC
scheme called Black-Burst (BB) is proposed in Sobrinho
and Krishnakumar (1999). The scheme BB is a
distributed MAC scheme for the provision of QoS to
realtime traffic for ad hoc networks that are based on
CSMA. In this scheme, nodes contend for the channel
until the channel becomes free. The packets that are
marked real time have priority over non-realtime packets
and a node that wishes to transmit realtime packet has
a priority for access to the channel over the nodes that
have non-realtime packets.
A medium access control protocol called Sticky
CSMA/CA that provides implicit synchronisation
and realtime QoS in wireless mesh networks is
proposed in Singh et al. (2007). The traffic considered
is a combination of Voice over IP (VoIP) and
delay-insensitive traffic. Monitoring of medium in Sticky
CSMA/CA is similar to standard CSMA/CA with the
difference that nodes in Sticky CSMA/CA remember
what activities are carried out in the recent past. The
information about the activities in the recent past is used
by a newly arriving VoIP flow to grab the medium as
soon as it is free, and thereafter, the VoIP flow sticks
to a periodic schedule. The gaps left by the realtime
traffic are filled by delay-insensitive traffic using the
leftover bandwidth. It is shown that the total voice
call carrying capacity of Sticky CSMA/CA increases
significantly as compared to the classical CSMA/CA
based techniques such as IEEE 802.11 b/e. Further,
authors therein, discuss that one can enhance QoS by
imposing periodicity at the application layer.
6.1.2 IEEE 802.11 based networks
Note that IEEE 802.11 is the most widely used MAC
layer for mobile ad hoc networks. There are two major
flavours of IEEE 802.11 pertaining to coordination
function – the first is with Distributed Coordination
Function (DCF) and the second is with Enhanced
Distributed Channel Access (EDCA). The version with
EDCA is also called IEEE 802.11e and has a support for
the provision of QoS.
It seems important to analyse how well the network
with IEEE 802.11 as MAC layer protocol can support
QoS. A unified model for the performance analysis
of IEEE 802.11e Enhanced Distributed Coordination
Function (EDCA) is proposed in Hui and Devetsikiotis
(2005). The parameters analysed are saturation
throughput and delay under the assumptions of finite
number of nodes and ideal channel conditions in a
single hop Wireless Local Area Network (WLAN). The
model is unified in the sense that it combines a Markov
model (Bianchi, 2000), p-persistent CSMA/CA model,
and average value model (Tay and Chua, 2001) for
the analysis of DCF. The model uses either multiple
bidimensional chains or multiple average value analysis
in separate back-off subperiods assuming that the
p-persistent behaviour is time-dependent. The proposed
model is validated through simulation results.
An analytical model to assess the performance of
IEEE 802.11 wireless LAN in terms of the provision
of QoS is presented in Zhai et al. (2005). The
parameters analysed are maximum throughput and
available bandwidth, delay and variations in delay,
and packet loss rate. The model is validated through
simulation results and it has been shown that by
controlling the total traffic rate, IEEE 802.11 can support
strict QoS requirements while achieving high channel
utilisation. Another study regarding the performance
assessment of IEEE 802.11 EDCF is presented in Hwang
and Wu (2008). A fixed point analysis of single cell IEEE
802.11 Wireless Local Area Networks (WLAN) is carried
out in Kumar et al. (2007). A mechanism to estimate
the available bandwidth of IEEE 802.11 based ad hoc
networks is presented in Sarr et al. (2008).
An analysis of priority schemes of IEEE 802.11 and
IEEE 802.11e based wireless LANs is carried out in Xiao
(2005). The proposed backoff based priority schemes
differentiate among the minimum backoff window
size, the backoff window increasing factor, and the
retransmission limit. The parameters considered in the
analytical model are saturation throughput, saturation
delays, and frame dropping probabilities of different
priority classes.
However, all these studies do not take into account
the transmission range and/or carrier-sense range while
analysing the throughput. Note that the transmission
range and carrier sensing range play a vital role in
the effective bandwidth and/or throughput in case of
ad hoc network. Analysis taking into considerations the
transmission and carrier sensing ranges of mobile devices
remains a challenging task and may be addressed in
future.
6.1.3 TDMA based networks
In an ad hoc network that is based on TDMA, the
notion of bandwidth is related to the free time slots
available along a link for transmission. The number
of free slots is decided by the neighbours and their
transmission/reception activities. Therefore, reservation
of bandwidth from a source to a destination implies
reservation of time slots that are free and are available
for sending packets. The methods of determining what
time slots are free and available for transmission can be
different and define the approach taken by a protocol.
In what follows, we discuss some of the approaches
proposed in the literature.
A protocol called Distributed Slot Reservation
Protocol (DSRP) that comprises of several strategies
for dynamic bandwidth allocation to be used in QoS
routing for TDMA based ad hoc networks is proposed
in Shih et al. (2006). In DSRP, QoS routing depends on
the information gathered only from one-hop neighbours
and thus making the protocol localised and distributed.
As mentioned earlier, the notion of bandwidth in ad hoc
networks where TDMA is used at the MAC layer is
related to the number of free slots. There are two types of
policies related to slot allocation to links in the network.
A policy that checks what are the valid and free slots for
a link is called a Slot Inhibited Policy (SIP) and a policy
that decides which slots can be used by a link is called a
Slot Decision Policy (SDP). Among the SDPs proposed
are:
•
Three-Hop Backward Decision Policy (3BDP)
•
Least Conflict First Policy (LCFP)
•
Most Reuse First Policy (MRFP).
DSRP is an on-demand slot reservation protocol for
QoS routing. When a source wants to communicate to
a destination, source initiates a QoS route discovery
by sending a Route Request (RREQ) packet to its
neighbours. During the route discovery phase, SIPs
decide what are the valid slots and SDPs decide which
slots out of the valid slots can be used by a link. The
actual reservation of slots is carried out when a Route
Reply (RREP), which is generated by the destination
after receiving an RREQ, travels upstream. However,
it has little more hassle. It may happen that the slots
that have been decided to reserve against an RREQ may
be reserved by the RREQs generated by other nodes.
To take care of that a Slot Adjustment Protocol (SAP) is
used to coordinate the reservation of slots for a link by
different RREQs. For that purpose, an algorithm for slot
adjustment is proposed.
11
6.1.4 CDMA based networks
A game-theoretic approach to energy efficient
modulation in networks that utilise CDMA with delay
as a QoS constraint is proposed in Meshkati et al.
(2007). The approach focuses on an M-ary Quadrature
Amplitude Modulation (MQAM). This is analogous to
a noncooperative game in which each user chooses its
strategy in terms of transmit power, transmit symbol
rate, and constellation size, in order to maximise its own
utility while satisfying its QoS constraints. The utility
function used is a measure of the number of reliable bits
transmitted per unit of energy consumed, and is suitable
for energy constrained networks. Using the proposed
game-theoretic framework, the tradeoffs among energy
efficiency, delay, throughput, and modulation order can
be quantified. Specifically, the effect of coding on energy
efficiency is studied and quantified using the proposed
game-theoretic approach. Further, the tradeoff between
energy efficiency and spectral efficiency has also been
illustrated.
Low complexity MAC protocols for QoS support in
third generation radio access networks are proposed
in O’Farrell and Omiyi (2005). The paper presents a
resource metric for networks where the users have
different signal-to-interference ratio requirements.
Further, a deadline-driven backoff procedure that can
be used in scheduling at the MAC layer so as to
provide a QoS support in terms of delay constraints,
is presented. Basically, the proposed control strategies
for distributed scheduling at the MAC layer are
spread spectrum techniques. The strategies are called
Single-Threshold Overload Signal Spread Spectrum
(ST-OSSS), Multiple-Threshold Overload Signal Spread
Spectrum (MT-OSSS), and Overload Signal Spread
Spectrum with Overload Detection (OSSS/OD).
The access procedure for ST-OSSS and MT-OSSS
is overload avoidance while that for OSSS/OD is
overload detection. Finally, a MAC scheduling algorithm
called Earlier Deadline First Parallel Link Scheduler
(EDF-PLS) is presented with the aim to optimise spectral
efficiency.
6.2 Recent trends in MAC design
A standard called IEEE 802.11e supports WLAN
applications with QoS requirements. In IEEE 802.11e,
a MAC access method called Hybrid Coordination
Function (HCF) is introduced. HCF includes Enhanced
Distributed Channel Access (EDCA) and HCF
Controlled Channel Access (HCCA). In HCCA,
a reference design that consists of a reference scheduling
and an admission control is provided. In the reference
design of IEEE 802.11e, the scheduler computes a
common scheduled Service Interval (SI) that is the
minimum of the delay bounds of all streams. It then
computes the duration called Transmission Opportunity
(TXOP) and Traffic Specifications (TSPEC). Admission
control is carried out, and TXOP are allocated to each
12
station in a round robin fashion. In fact, the reference
design is such that QoS is provided using the most
stringent (i.e., the minimum of all streams) delay bound.
The problem due to the use of the most stringent delay
bound is that there is an overallocation of bandwidth.
The reason is that shorter is the SI, the larger will be
the fluctuations of packet arrival rate, and this implies
that more bandwidth is needed to transmit all arrived
packets. Further, a short SI may lead to more number of
polling during a period, and there may be more polling
overhead.
Therefore, the issue is that how to avoid
overallocation of bandwidth. The answer is that different
SIs should be used for different streams with different
delay requirements. For that, an Equal-Spacing (ESP)
based design is presented in Zhao and Tsang (2006)
which generalises the reference scheduling of 802.11e. The
word ‘spacing’ means an interval between two successive
scheduling of a stream. The ESP schedules each stream
with equal spacing, however, it may schedule different
streams with different spacing.
A MAC layer protocol called an Adaptive MAC
(AMAC) that uses m-ary tree algorithms for QoS
support in single cell ad hoc networks is proposed in
Tsigkas and Pavlidou (2008). The proposed protocol
is compared with Dynamic Priority MAC Protocol for
Time Bounded services (DP-TB) (Tsigkas and Pavlidou,
2006) and Elimination-Yield Non-Preemptive Priority
Multiple Access (EY-NPMA). Note that EY-NPMA is
MAC protocol for HIPERLAN and is based on dynamic
priorities. The proposed protocol AMAC provides a per
flow service guarantees to flows with delay as the QoS
parameter. It incorporates the RTS/CTS mechanism to
protect itself from the hidden terminal problem. The
protocol adapts to traffic conditions i.e., it is sensitive
to traffic load variations. However, the protocol requires
channel synchronisation among all those stations that
intend to transmit.
A MAC scheme called Busy-Tone-Based (BTB) MAC
is proposed in Wang et al. (2008) for supporting voice
and data traffic in ad hoc wireless networks. The
issues addressed are unfairness and priority reversal
due to hidden terminal, exposed terminal, and location
dependent contention. BTB ensures guaranteed priority
access for delay sensitive voice traffic using transmitter
busy tones in the node backoff procedure. It resolves
hidden and exposed terminal problems. The priority
assigned is independent of user locations, and therefore,
priority reversal is resolved. However, it does not address
the issue of capture effect where a receiver may be able
to receive correctly the desired frame even if a collision
occurs.
A MAC protocol called multi-Channel MAC
(CMAC) for dynamic channel allocation in multichannel
wireless sensor networks4 is proposed in Chowdhury
et al. (2009). Note that the problem of channel
assignment with minimum interference in single
transceiver multichannel communication is in NP,
therefore, CMAC provides a heuristic and approximate
solution. The scheme employed in CMAC takes into
account energy constraints in sensor nodes by placing
them in default sleep mode as far as possible. Its
performance is compared with Sensor MAC (SMAC)
and it is observed that the performance of CMAC
comes out to be better than SMAC in terms of energy
consumption and throughput.
A mechanism for dynamically assigning priorities of
packets based on their deadlines and the traversed hops
is proposed in Reddy et al. (2007). The mechanism called
ReAllocative Priority (ReAP) is aimed to provide QoS at
the MAC layer to video traffic. The scheme introduces
adaptive TXOP by modifying the TXOP dynamically
based on the queue length. However, the scheme does not
take into account the mobility. A comparison of various
MAC layer schemes with QoS provisioning is shown in
Table 3.
Table 3 A comparison of MAC layer protocols used in QoS provisioning
Protocol
Features
QoS parameter
Comments
Sticky CSMA
Singh et al. (2007)
DSRP
Shih et al. (2006)
GTA
Meshkati et al. (2007)
LC
O’Farrell and Omiyi (2005)
ESP
Zhao and Tsang (2006)
AMAC
Tsigkas and Pavlidou (2008)
BTB MAC
Wang et al. (2008)
CMAC
Chowdhury et al. (2009)
Implicit synchronisation
Delay
Slot reservation,
TDMA based
Quadrature Amplitude Modulation,
CDMA based
Resource metric for SNR
requirements, CDMA based
Equal-Spacing based
reference design
Per flow service guarantees
Bandwidth
Best-effort traffic fills the
gaps left by realtime traffic
Slot Adjustment Protocol,
Dynamic bandwidth allocation
Game-Theoretic Approach
Priority is independent of
user locations
Dynamic Channel Allocation
Delay
Delay
Delay
Delay
Delay
Energy
Deadline-driven backoff procedure,
Optimise spectral efficiency
Generalises reference scheduling
of 802.11e
m-ary tree algorithms,
Dynamic priorities
Resolves hidden and
exposed terminal problems
Multichannel MAC
Future work
In future, a major problem that needs to be addressed is
the design of an efficient MAC for an ad hoc network
that supports QoS. Although, there are extensions of
CSMA/CA and IEEE 802.11 for supporting QoS,
however, a node that wishes to transmit may need
to wait for a random amount of time due to the
exponential backoff mechanism. Also, neighbours in the
carrier sensing range cannot transmit simultaneously due
to hidden and exposed terminal problems. Moreover,
neighbouring nodes that are lying along a path from the
source to the destination cannot transmit simultaneously
due to intra-path correlation, and thus the effective
bandwidth of the path is less than the raw bandwidth
of a link. Similarly, inter-path correlation may bite the
advantages of having multiple paths from a given source
to a destination. One may address how to mitigate
the effect of intra-path and inter-path correlations
while providing the QoS in mobile ad hoc networks.
Another direction for future work can be the analysis of
IEEE 802.11 and its variations taking into account the
transmission range and carrier sensing range.
In what follows, we discuss cross layer schemes.
7 Cross-layer QoS strategies
A cross-layer protocol may utilise the properties or
functionalities of more than one layers.
7.1 TDMA-based cross layer routing
A cross layer QoS Routing (QoSR-Z) is proposed
in Zhu and Corson (2002). QoSR-Z is based on Ad hoc
On-demand Distance Vector (AODV) routing and builds
QoS routes on-demand. Also, it uses TDMA. A network
layer function (i.e., routing) is coupled with the MAC
layer function (i.e., finding the time slots), therefore, we
call it a cross layer protocol. It is based on the assumption
that the application is session oriented and requires a
constant bandwidth. Specifically, a session specifies its
QoS requirement in terms of the number of transmission
time slots needed on its route from a given source S to
a destination D. For each session or flow, QoS routing
protocol not only finds the route, it also determines the
slots for each link on the route.
Another cross layer protocol called QoS Routing
(QoSR-L) protocol based on Destination Sequence
Distance Vector (DSDV) routing that tries to provide
bandwidth guarantees is proposed in Lin and Liu
(1999). Bandwidth is defined in terms of number of
time slots available for transmission. The protocol is
based on heuristics that may provide an approximate
solution. This is because finding a schedule of free
slots that maximises the available bandwidth is an NP
complete problem. The protocol is based on TDMA
and requires some sort of global synchronisation that
incurs an extra overhead. Another protocol that reserves
13
the bandwidth for QoS routing in networks that utilise
TDMA at the MAC layer is proposed in Liao et al.
(2002). We call this protocol as TDMA-Based Bandwidth
Reservation Protocol (TBRP) to refer it for the purpose
of comparison. The protocol takes into account the
hidden terminal problem and the exposed terminal
problem during the route establishment phase.
Note that protocols that require a scheme for time
slot assignments do not seem to be suitable for an ad hoc
network. The reason is that in an ad hoc network, nodes
are allowed to move from their positions, and any such
movement of nodes may require updates in the time slot
assignments incurring a significantly large overhead.
7.2 UWB-based cross layer routing
A cross layer protocol called QoS Greedy Perimeter
Stateless Routing (QoS-GPSR) is proposed in Abdrabu
and Zhuang (2006). In QoS-GPSR, a resource allocation
algorithm is coupled with a position based routing
algorithm. It is suited for a single channel UltraWide-Band (UWB) based ad hoc network where UWB
provides position information. The cross layer design of
QoS-GPSR utilises contention based MAC, e.g. IEEE
802.11. Admission control at each node along the route
is based on the bandwidth availability information
provided by the MAC for the node and its neighbours
in its carrier sensing range. Further, the cross layer
design enables to consider the possible simultaneous
transmissions of nodes belonging to the same route
when making bandwidth reservations, as these may affect
the effective throughput. The protocol QoS-GPSR can
also work with TDMA MAC protocols, e.g. 802.15.3,
provided that a proper packet scheduling algorithm is
employed.
7.3 Cross layer schemes for resource allocation
A cross layer scheme for resource allocation that operates
at the physical and data link layers for the provision of
QoS guarantees in wireless relay networks is presented
in Tang and Zhang (2007). The goal of the scheme
proposed, therein, is to maximise the relay network
throughput subject to a given delay QoS constraint. The
scheme uses the concepts from information theory and
the concepts of effective capacity in wireless networks.
The delay constraint is characterised by a parameter
called QoS exponent which represents the information
exchanged between the physical layer and the data link
layer of the proposed cross layer scheme. There are two
major classes of wireless relay networks – Amplify and
Forward (AF), and Decode and Forward (DF). Dynamic
resource allocation algorithms for both types of networks
are developed that can be used for QoS guarantees
for multimedia communications in wireless networks.
Further, analytical models are proposed for dynamic
resource allocation in AF and DF relay networks. A fixed
power allocation for DF relay networks has also been
analysed.
14
7.4 Cross layer schemes involving physical layer
A cross-layer optimisation for Low Density Parity
Check (LDPC) coded multirate multiuser systems with
QoS constraints is presented in Li and Wang (2006).
The proposed multirate multiple-access wireless system
uses variable spreading gain and chip level random
interleaving. The proposed framework uses some services
from the physical layer as well as some services from
the network layer, and therefore, it is termed as
cross-layer. At the physical layer, QoS requirements
are specified in terms of target bit error rate of
each user and the transmission powers of users are
adjusted according to the current system load and the
corresponding rate requirements of the users. The QoS
requirements specified at the network layer are call
blocking probabilities, call connection delays, packet
congestion probabilities and packet loss rates. A method
for call admission control that is based on Multicriterion
Reinforcement Learning (MCRL) is proposed so as to
maximise the average revenue of the network subject to
call level as well as packet level QoS. The method can
handle multiple QoS constraints.
A cross-layer scheduling algorithm with QoS support
in wireless networks is proposed in Liu et al. (2006).
The proposed algorithm is for MAC layer scheduling
and may handle multiple connections with diverse QoS
requirements with each connection employing Adaptive
Modulation and Coding (AMC) at the physical layer, and
therefore, it is called a cross-layer algorithm. A priority
is assigned to each connection and a connection with the
highest priority is scheduled first. The performance of
the proposed scheduling algorithm is validated through
simulations. A comparison of cross layer techniques is
presented in Table 4.
Note that a cross-layer technique combines the
functionalities of network layer with MAC and/or
physical layer. Ultimately, the packets have to pass
through different layer at the sender and receiver ends.
Separating the layers may simplify the design of the
protocol. On the other hand, the design of a crosslayer protocol seems to be complex. However, once a
cross-layer protocol is designed, it is expected to perform
better than a protocol that uses functionalities of only
a single layer. However, the performance depends upon
how judiciously the functionalities involving more than
one layers are incorporated.
Future work
In future, one may address how to utilise the cross layer
techniques to enhance the throughput when there are
multiple source destination pairs in an ad hoc network.
The throughput depends upon many factors such as raw
bandwidth, node density, interference. Moreover, one
may also address how to design a cross layer solution so
as to minimise the number of packets dropped as that
may decide packet retransmissions. Further, the design
of cross layer techniques that are energy efficient forms
another direction for future works.
In the previous sections, we described schemes based
on the layered classification. We now switch to schemes
that are part of the functional classifications. Note that
the schemes with routing and/or MAC functionalities are
already covered in the previous sections.
In what follows, we discuss the research reported in
the literature related to admission control.
8 Admission control
Admission control is an important function for the
provision of QoS as it determines which packet is allowed
to enter and which packet is not allowed to enter into
the network. The decision may be based on many factors
such as what may be the consequence of allowing a
packet or flow to enter into the network.
Table 4 A comparison of cross layer schemes used for QoS provisioning
Scheme
Features
QoS parameter
Layers
QoSR-Z
(Zhu and Corson, 2002)
Based on AODV,
Based on TDMA
Bandwidth
Network and MAC
QoSR-L
(Lin and Liu, 1999)
Based on DSDV,
Based on TDMA
Bandwidth
Network and MAC
TBRP
(Liao et al., 2002)
TDMA based,
Slot reservation
Bandwidth
Network and MAC
QoS-GPSR
(Abdrabu and Zhuang, 2006)
Based on UWB,
Position based routing
Delay
Network and MAC
AF-DF
(Tang and Zhang, 2007)
Resource allocation
Throughput and delay
Physical and data link
MMSQ
(Li and Wang, 2006)
Multirate multiuser
Multiple criteria
Physical and network
CLS
(Liu et al., 2006)
Adaptive modulation
and coding
Delay
Physical and MAC
8.1 Measurement based admission control
A scheme for admission control, which is based on
the local measurements carried out at a node, is
proposed in Xiao and Li (2004). We call it Measurement
Based Admission Control (MBAC). In MBAC, there
are different Access Categories (AC) for best-effort,
video probe, video, and voice traffic. The scheme is as
follows. Each station carries out measurements during
beacon intervals to estimate the transmission budget.
The transmission budget provides an indication about
the allowable transmission time and the time utilised
for transmissions for each AC. Each station determines
the transmission limit for each AC for the current
beacon interval. This is based on the transmission count
during the previous beacon interval and the transmission
budget calculated. The time taken in the transmissions
of voice and video packets should not exceed the local
transmission limit for each AC. If the transmission
budget for an AC is depleted or nearly depleted, new
streams are not allowed, and existing streams are not
allowed to increase the transmission time per beacon
interval that they were already using so that existing
voice and video streams are protected. However, the
proposed schemes are limited to single hop wireless
ad hoc networks.
8.2 Local admission control
Schemes for local data control and admission control are
proposed in Xiao and Li (2004). The schemes are aimed to
provide QoS while there is realtime traffic and data traffic
in the network. A major issue addressed therein is how to
adapt parameters such as Contention Window (CW) and
Arbitration Inter-Frame Spacing (AIFS) so as to provide
QoS to real time traffic. Specifically, two schemes for
data control are proposed: Dynamic Function Mapping
(DFM), and Dynamic Traffic (DT). In the DFM scheme,
a function is defined for mapping traffic load conditions
to parameters such as Contention Window (CW) and
Arbitration Inter-Frame Spacing (AIFS). In the DT
scheme, parameters are dynamically varied according
to the tendencies of traffic load conditions. The word
‘tendency’ means whether the traffic continues to increase
or continues to decrease. A Contention-aware Admission
Control Protocol (CACP) is proposed in Yang and
Kravets (2005). CACP provides admission control for
flows in a single channel ad hoc network. It is based on
the knowledge of local resources at a node and the effect
of admitting the new flow on the neighbouring nodes.
CACP focuses on QoS support in terms of bandwidth
allocation.
8.3 MAC based admission control
A call admission control scheme for QoS guarantees in
terms of bandwidth is proposed in Liu et al. (2005).
The notion of bandwidth is related to the number of
15
free time slots. Authors, therein, proposed an algorithm
for assignment of time slots in an ad hoc network that
either employs Time Division Multiple Access (TDMA)
or Code Division Multiple Access (CDMA) at the MAC
layer. It is proved that a time slot assignment for a unit
of bandwidth, if exists, can be found in O(|P |) time,
where P represents a path and |P | represents the number
of links in path P . The same is extended for two units
of bandwidth. Further, an algorithm for slot assignment
problem with k slots in a TDMA frame is proposed that
has a complexity of O(|P |2 k). Finally, an algorithm for
time slot assignment problem is integrated into a call
admission control scheme so as to provide QoS in terms
of bandwidth guarantees.
Schemes for call admission control for Wideband
Code Division Multiple Access (WCDMA) based mobile
wireless networks with multiple service classes that
negotiate QoS are presented in Paschos et al. (2005).
Therein, bandwidth reservation is used to improve the
QoS and packets are either blocked or dropped to prevent
the system to form an outage situation. However, the
probability of either blocking or dropping a packet is
kept fairly low. The traffic is divided into two classes:
realtime and non-realtime. In one of the call admission
control scheme that is used in a lightly loaded system,
calls are admitted and serviced with the maximum
resource requirements of the user. At a point, when no
further calls can be accommodated, a procedure that
negotiates with new and ongoing calls to free some of the
resources is applied. However, the drawback of such a
scheme is that non-realtime traffic is preempted to make
room for the realtime traffic. It has been shown that
this drawback can be addressed using a small proportion
of the available resources as a non-realtime reservation
bandwidth. The performance of the proposed algorithm
is analysed using a multistep Markov analysis.
8.4 Rate based admission control
A protocol that is a combination of admission control
and rate policing called Multi-Priority Admission and
Rate Control (MPARC) is proposed in Yang and
Kravets (2007). The protocol aims to provide throughput
guarantees for multi-priority traffic in ad hoc networks.
The core of MPARC is a model for allocating bandwidth
in a unsaturated, semi-saturated, and saturated network.
In MPARC, different priorities are assigned for realtime
and best-effort flows. The protocol guarantees that the
throughput of currently admitted realtime flows will
not decrease either due to realtime flows arriving later
with equal or lower priorities or due to best-effort
flows. In other words, throughput of currently admitted
realtime flows can decrease only when a realtime flow
with a high priority arrives. In MPARC, this is ensured
using a mechanism for admission control for newly
arriving realtime flows and applying a rate policing for
best-effort traffic.
16
Table 5 A comparison of methods for call admission control
Protocol
Features
Comments
CACB (Liu et al., 2005)
TDMA/CDMA at MAC layer,
Time slot assignment
MBAC (Xiao and Li, 2004)
Access categories,
Transmission budget
Delay guarantees
CAC-WCDMA (Paschos et al., 2005)
Bandwidth reservation
Traffic is divided into classes
MPARC (Yang and Kravets, 2007)
Combination of admission control
with rate policing
Throughput guarantees
CACP (Yang and Kravets, 2005)
Localised and distributed
PAC Chakeres et al. (2007)
Perceptive Admission Control
and Channel Monitoring
Ensures low packet loss
and delays
8.5 Perceptive Admission Control
A protocol for Perceptive Admission Control (PAC)
for providing QoS in a wireless network is proposed
in Chakeres et al. (2007). The protocol monitors the
wireless channel and adapts the decisions of admission
control so that the utilisation of the network is
maintained while preventing congestion. To make an
admission decision, sources consider the entire region
that a new flow’s transmissions will impact. It is shown
that, with a proper Carrier Sense (CS) range, the time
for which the channel is sensed busy is a good estimate
of the utilised and available bandwidth. The performance
of the proposed protocol is analysed through simulations
and it comes out that PAC ensures low packet losses and
delays for all packets that are admitted into the network.
However, the limitation of the protocol is that it focuses
on a single hop admission control.
A comparison of methods of call admission control
is shown in Table 5. Note that we refer the method
presented in Liu et al. (2005) as CACB and that proposed
in Paschos et al. (2005) as CAC-WCDMA.
Future work
The directions for future work pertaining to admission
control for QoS provisioning in mobile ad hoc networks
are as follows. Firstly, one may come up with a scheme
for measurement based admission control that can be
applied in a generalised mobile ad hoc network. By the
term ‘generalised’, we mean that there may be multiple
hops between a given source to a destination and nodes
are free to move about randomly. Secondly, one may
propose a scheme for perceptive admission control in a
generalised mobile ad hoc network.
In what follows, we describe the research related
to scheduling which is another functionality to be
considered for provision of QoS in an ad hoc network.
reported in the literature pertaining to the provision of
QoS that use scheduling.
9.1 Prioritised scheduling
A Distributed Priority Scheduling (DPS) for provision
of delay based QoS is proposed in Kanodia et al. (2002).
In DPS, each node locally constructs a scheduling
table based on the information that it overhears,
and incorporates estimate of its relative priority
into MAC. The DPS uses a coordinated multihop
scheduling mechanism in which at an intermediate node,
downstream priorities of a packet are modified based on
the service provided to the packet upstream. In other
words, the downstream priority index of a packet is
computed recursively based on its upstream priority
index. By doing so, downstream nodes can help packets
catch up if they are excessively delayed upstream, whereas
the priorities of the packets that have arrived earlier can
be reduced so that more urgent packets can pass through
quickly.
Schemes for optimising the transmission schedule
of packets using priorities in multihop latency aware
scheduling for delay constrained data are presented in
Liang and Dong (2007). The objective of schemes is to
determine the weights to be assigned to the distance that
the packet has yet to travel and the remaining lifetime
so as to rank the packets according to their urgency.
The transmission schemes presented are cross-layer in
nature and include schedules such as EDF and Largest
Distance First (LDF). Further, an analytical framework
using recursive nonhomogeneous Markovian analysis is
also presented. The aim of the analytical framework is to
study the effect of lifetime-distance factor on packet loss
probability. Therein, it is shown that a proper balance in
the lifetime-distance factor can significantly improve the
performance of the network.
9.2 Rate based scheduling
9 Scheduling
A mechanism for providing QoS in a network is
scheduling. In this section, we discuss the research
An analytical model for tradeoffs between delay bounds
and computational complexity of packet scheduling
algorithms for the provision of QoS in a network is
described in Xu and Lipton (2005). Therein, the authors
compute the delays incurred in a scheduling algorithm
that employs a service discipline with respect to the delays
incurred in GPS. These delays are called GPS-relative
delays. It is proved that, under a slightly restrictive
model, the lower bound computational complexity of any
scheduling algorithm that guarantees O(1) GPS-relative
delay is Ω(log n), where n is the number of sessions. It
has been found that the lower bound on the complexity
remains the same even if the delay bound is relaxed
to O(nα ), where 0 < α < 1 which implies that the
complexity remains constant during the delay interval
[O(1), O(n)]. Further, the analytical results obtained
for the above complexity are extended for a stronger
computational model called the linear decision tree
model. It has been shown that the same lower bounds on
the complexity, under certain conditions, can be applied
to guarantee tight bounds on end-to-end delays provided
that the delay bounds are obtained using the latency-rate
framework.
An analytical model for end-to-end delay bounds
for traffic aggregates under guaranteed rate scheduling
algorithms is presented in Sun and Shin (2005). In that,
end-to-end delay bounds for a single aggregation are
derived deterministically assuming that all incoming
flows at an aggregator conform to the token-bucket
model. Three types of Guaranteed Rate (GR) scheduling
algorithms are proposed:
•
stand-alone GR
•
two-level hierarchical
•
rate-controlled two-level hierarchical GR; and an
aggregator can use any of these algorithms.
The end-to-end delay bounds for multiple aggregation
fall within an aggregation region when aggregators use
the rate control for two-level hierarchical GR scheduling
algorithm. It is shown that using the proposed guaranteed
scheduling algorithms, there exists a delay bound for each
flow, and under certain conditions the bounds are smaller
than that of per flow scheduling. The analytical delay
bounds are validated through simulations.
A rate control scheme called RCS for adaptive real
time applications in IP networks with lossy links and long
round trip times is proposed in Akyilidz et al. (2007).
In RCS, all routers along a path use a priority discipline;
and special types of low priority packets called dummy
packets are used to probe the availability of network
resources. Moreover, to guard the RCS from temporal
signal losses, an algorithm for improving the robustness
of RCS is proposed. Specifically, the behaviour of RCS
is investigated in the presence of packet losses due to link
errors, network congestion, and temporal signal losses.
Further, delay bounds for realtime traffic sources are also
studied.
9.3 Scheduling with end-to-end delay bounds
Node delay assignment strategies to support end-to-end
delay requirements in heterogeneous networks are
17
proposed in Znati and Melhem (2004).5 In Znati and
Melhem (2004), a methodology to compute a set of
feasible per node delays for a given flow and for a
class of delay based servers is described. Consequently,
the problem of an optimal assignment of per node
delays that takes into account the workload across the
routing path is formalised. The resulting solution is
optimal, however, its implementation overhead is high.
To overcome this shortcoming, authors, therein, propose
two heuristics called Equi-Partition Heuristic (EPH)
and Load-Balancing Heuristic (LBH). EPH uses the
equi-partition concept to compute the initial delay values
and adjusts them to meet the end-to-end delay
requirements. LPH distributes the load proportionally
across all nodes on the routing path using the concept
of relaxation factor. The performance of these heuristics
comes out to be comparable to that of the optimal
scheme.
9.4 Proportional differentiation
Schemes for QoS assurance through class selection
and proportional differentiation in wireless ad hoc
networks are proposed in Wang and Ramanathan
(2005). The proposed schemes are based on a service
model called Neighbourhood Proportional Delay
Differentiation (NPDD) model Wang and Ramanathan
(2003). In NPDD, there are a number of service classes
of packets and these classes are ordered in terms of
per-hop queueing delays at each node. The NPDD model
not only holds for each node, it also holds across nodes
in a certain neighbourhood. In other words, the delay
proportionality is maintained at each node and among
nodes in the same contending set. Two nodes are said to
belong to the same contending set if there exists a route
between them. The NPDD scheduler is work-conserving
and an algorithm known as Waiting Time Priority (WTP)
(Dovrolis et al., 2002) is adopted in which each class
is serviced in a First-in-First-Out (FIFO) queue. The
scheduler always schedules the head-of-line packet with
highest priority of a service class. Scheduling algorithms
(Wang and Ramanathan, 2005) used in NPDD are called
Dynamic Class Selection (DCS) algorithms. Further,
applications of DCS are modelled using a game-theoretic
approach. For noncooperative games with selfish players,
analytical models for the existence of an equilibrium,
the feasibility of an equilibrium, and the guaranteed
convergence to a feasible equilibrium when it exists,
are proposed for both single hop as well as multihop
networks.
9.5 Opportunistic scheduling
An opportunistic scheduling discipline to handle a
combination of realtime and best-effort traffic at a
wireless access point that utilises multiuser diversity
is proposed in Patil and Veciana (2007). The aim
is to support probabilistic QoS in terms of service
rate guarantees to realtime traffic while still achieving
18
opportunistic throughput to all traffic classes and users.
The lower bounds on the service that is to be provided
to a realtime session are obtained under the assumption
that variations in the capacities demanded by users are
either known or can be estimated. These bounds may
enable one in predicting the QoS required by the real time
traffic and the same can be incorporated using resource
management and admission control schemes. Further,
authors show that there can be a tradeoff between how
tightly one can satisfy the QoS requirements and the
overall system throughput. In other words, if the QoS
requirements for realtime traffic are very tight, then
realtime traffic will be given a higher priority again
and again. This adversely affects the traffic other than
realtime traffic and thus decreasing the opportunistic
throughput gains.
Future work
In future, one may propose scheduling schemes that are
capable of providing per hop and per packet proportional
differentiation. The type of service to be received may
depend upon the urgency level of the packet and the
service received by the packet upstream. Also, one
should consider whether putting a packet to a higher
differentiation class may lead to catching the deadline at
the final destination or not. Another direction for future
research can be designing a scheduling scheme that is fair
to realtime as well as non-realtime packets.
In what follows, we describe research related to the
fairness in QoS provisioning.
10 Fairness
Note that fairness is not a functionality to be
incorporated for the provision of QoS, however, it is
among the goals or objectives of QoS provisioning that
should not be overlooked. The meaning of fairness may
depend on the context and it should be considered to
be associated with the requirements of the application.
In other words, one has to decide whether the scheme
should be fair to QoS flows or to non-QoS flows or to
all flows. In general, the term ‘fairness’ means that the
network is able to provide the same level of service to
all its users. However, fairness is a key issue in case
of schemes that have a provision of QoS in a wireless
network.
10.1 Wireless fair queueing
A topology-independent wireless fair queueing model
for a shared medium ad hoc network is proposed
in Luo and Lu (2005). The proposed fairness model
ensures coordinated fair channel access among the
contending flows so that spatial reuse of bandwidth
is maximised. Algorithms that realise the fluid fairness
model with analytical bounds are presented. Further,
a distributed implementation that approximates an
ideal centralised algorithm is also presented and the
performance of the proposed algorithm is analysed
through simulations.
10.2 Per-station fairness
A scheme called Fair QoS Agent (FQA) is proposed
in Park and Kim (2007) that provide per class QoS
enhancement and per station fair channel sharing in
Wireless Local Area Networks (WLAN). In FQA, there
are two components that operate above IEEE 802.11
MAC layer so that no change is required in IEEE
802.11 MAC layer. These components are a dual service
differentiator and a service level manager. The function
of dual service differentiator is to improve QoS for
different classes of traffic using appropriate algorithms
for scheduling and queue management. On the other
hand, service level manager is responsible for assuring
fair channel sharing by estimating the fair share of each
station and dynamically adjusting the service levels of
packets. The proposed scheme is simple and it does
not require to maintain per station queues. Further, an
assurance for weighted fairness among stations in terms
of channel access time is provided.
10.3 Fairness framework
As the IEEE 802.11 wireless LANs are widely deployed
for providing realtime service such as voice and
non-realtime service such as data, the provision
of fairness and QoS becomes a challenging task.
A framework for fairness and QoS assurance for current
IEEE 802.11 networks with multiple access points is
presented in Bejerano and Bhatia (2006). The framework
is called MiFi and relies on centralised coordination of
the APs. It is based on the fact that during any given
time of contention free period, APs that do not interfere
with one another are activated and others are silent. The
amount of service time granted to an AP is proportional
to the load on AP and the performance of the system
is optimised employing efficient scheduling algorithms.
Therein, it is pointed out that such a scheme can be
implemented without modifying either the underlying
MAC protocol or the behaviour of mobile stations and
the scheme takes care of the hidden terminal problem and
the overlapping cell problem. The proposed framework
is validated through simulations and it is established that
the system supports fairness. As a result, MiFi is able
to provide QoS guarantees for real time traffic and still
maintains relatively high throughput for non-realtime
traffic.
Future work
In future, one may address how to design a scheme
for QoS provisioning in an ad hoc network such that
it maintains fairness as per the requirements of the
application. Also, one may study the implications of
providing the fairness on realtime and non-realtime flows.
In what flows, we describe the research related to the
miscellaneous types of QoS
11 Miscellaneous QoS
In this section, we discuss the research reported in the
literature keeping in view of different approaches and
objectives.
11.1 User level QoS
In a network, functionalities of layers may vary
depending upon their boundaries. Consequently, there
can be different levels of the QoS provisioning (Tasaka
and Ishibashi, 2002) such as physical level, node level,
network level, end-to-end level, and user level. The user
level QoS is also called the perceptual QoS or user
perceived QoS. Out of these levels, the user level QoS
is the most important QoS as the users are the ultimate
recipient of the QoS. Therefore, there is a need to measure
and assess the QoS at the user level.
In Ito and Tasaka (2005), a scheme for quantitative
assessment of user level QoS for audio-video
transmissions using psychometric methods is presented.
The psychometric methods used are:
•
the method of paired comparisons
•
the law of comparative judgement.
A mapping of the QoS from application level to user
level is carried out using principal component analysis
and multiple regression analysis. During the assessment
of QoS, the transmission of an audio-video scheme over a
loaded network is simulated. The concept of control gain
through media synchronisation is proposed to reduce the
load on the network.
A scheme for user input driven QoS management
in ad hoc networks is proposed in Wang et al. (2009).
The motivation behind it is that the network QoS
is generally defined in terms of technical parameters
such as delay, jitter, and bandwidth. An end-user may
often find it difficult to express his desired QoS in
such technical parameters. Therefore, there is a need to
translate user inputs to technical parameters. An interface
design that uses translation functions to map user inputs
to parameters at routing and MAC layers is proposed.
Based on the translated parameters, the strategies are
chosen at specific layers so that the QoS desired by
the user is achieved. Further, the costs associated with
adoption of different strategies at different layer is also
examined in Wang et al. (2009).
11.2 Aggregate QoS
QoS can be provided to flows individually or to number
of flows taken together. The first type of QoS is called
the individual QoS or fine grain QoS or per flow QoS
and the later type of QoS is called as the aggregate
QoS or coarse grained QoS or scalable QoS. It is worth
mentioning that the pace of implementing the QoS
solutions in operational networks is slow due to many
reasons. One of the reasons is that the complexities
19
involved in methodologies for the provision of individual
QoS are relatively high. Therefore, there is a need to
explore methodologies for the provision of QoS that are
scalable or provide the QoS on an aggregate basis.
A loss performance study of individual QoS vs.
aggregate QoS is carried out in Xu and Guerin (2005).
Specifically, authors therein explore the differences
between the individual QoS and aggregate QoS in terms
of loss guarantees where guarantees are provided only
at the aggregate level. Analytical models to compute
both individual and aggregate loss guarantees for
ON-OFF and periodic sources are proposed. Further,
loss deviations for ON-OFF sources and periodic sources
are also analysed.
11.3 Integrated QoS
A support for integrating QoS with traffic engineering
and failure recovery is needed in a network that runs
the Transmission Control Protocol (TCP). Note that
TCP uses window based congestion control. In general,
the congestion control mechanism of TCP is called
Additive Increase Multiple Decrease (AIMD). If the
TCP anticipates congestion, the size of the window is
decreased.
Optimal algorithms for end-to-end integrated QoS,
traffic engineering, and failure recovery are proposed
in Movsichoff et al. (2007). Therein, laws for adapting
the data rate and laws of load balancing that can be
applied to networks with multiple paths between source
and destination and where multiple classes of traffic are
to be supported. The algorithms proposed need a minimal
amount of information for optimal control in the form of
whether a path is congested or not. The laws of control
proposed in the paper enable one to detect source inferred
congestion in the network. The proposed approach can
be used in networks with node or link failures in the
sense that when a link or node failure occurs the traffic
is rerouted away from the failed node or link. The work
presented can set a foundation for protocol that control
the traffic at different layers such as IP, transport or
higher layers.
Apart from that techniques for integration of
multihop wireless and wired networks with QoS
constraints are presented in Bejerano (2004). Specifically,
authors therein studied the problem of integrating
static multihop networks with a wired backbone in
such a manner so that QoS guarantees in terms of
bandwidth and delay are not violated. For that purpose,
the topology of the network is partitioned into a
minimal number of disjoint clusters that satisfy the QoS
constraints. The problem is formulated as an instance
of the Capacitated Facility Location Problem (CFLP)
with QoS constraints. A polynomial time approximation
algorithm is proposed that provides a solution with a
constant factor of the optimal one. Further, an adaptive
delivery mechanism is introduced that maximises the
throughput of each cluster without violating the QoS
constraints.
20
11.4 Self-adaptive QoS
It is presumed that the Next Generation Networks (NGN)
should be self-adaptive so that demands for changes in
traffic are satisfied. An example of such a network is
a Cognitive Packet Network (CPN). In CPN, paths are
selected adaptively in order to offer a best-effort QoS
to end users based on user defined QoS. This is done
by incorporating some kind of smartness or intelligence
inside the network. On the other hand an NGN should
also have a provision of guaranteed QoS. Thus, an
NGN should be able to support multiple QoS services.
In Sabella and Iovanna (2006), some key issues pertaining
to the provision of QoS and how one should react
to traffic changes in NGNs are presented. Therein, a
traffic engineering solution for supporting multiple QoS
services in an NGN through either Multi-Protocol Label
Switching (MPLS) or Generalised Multi-Protocol Label
Switching (GMPLS) is explored.
11.5 Capacity and QoS
A network where paths from a source to destination
fail frequently is called as an intermittently connected
network. Examples of such a network are an ad hoc
network or a sensor network. An analytical model for
QoS and capacity of intermittently connected networks
is proposed in Small and Haas (2007). In Small and
Haas (2007), a model that quantifies the resourcesdelay tradeoff and the capacity with QoS constraints
is presented. An intermittently connected network
is represented using a shared wireless infostation
model. The model is aimed to help network designers
in controlling the performance by adjusting the
communication bandwidth, fraction of time a node
spends in sleep mode, and the reliability in terms of
packet delivery.
11.6 QoS and multi-user diversity
The capacity of a time varying channel can be increased
with the use of diversity. The term ‘diversity’ means that
there are multiple signal paths between a transmitter
and a receiver. Diversity can be achieved in time, space,
and frequency; and traditional diversity methods are
applicable to a link with a single user. However, another
kind of diversity called multiuser diversity is introduced
in Knopp and Humblet (1995). It means that multiple
users share a time-varying channel and each of these users
has an independent channel gains for the shared medium.
In Wu and Negi (2005), the problem of supporting
QoS over a fading channel is considered that utilises
multiuser diversity. A downlink time slotted fading
channel is shared among a number of users, say K.
A statistical QoS that is characterised by <data rate,
delay bound, delay bound violation probability> is
provided explicitly. It is shown that when delay bounds
are not very tight, the approach can substantially increase
the delay-constrained capacity of the fading channel as
compared to a fixed slot assignment scheme. The delayconstrained capacity of a fading channel means the
maximum data rate that can be achieved when a given
probability of violation of delay bound is satisfied.
11.7 Adaptive QoS for multimedia applications
A mechanism for end-to-end delay control of multimedia
applications over multihop wireless links is proposed
in He et al. (2008). The motivation behind the
mechanism is that in case of multimedia applications, the
mechanism should adapt the client’s QoS requirements to
network resource constraints. The proposed mechanism
has separate adaptation components that operate at
application layer, middleware layer, and network layer.
At the application layer, there is a requirement
adapter that dynamically changes the requirement
levels according to end-to-end delay measurement and
QoS requirements that are acceptable to the enduser. At the middleware layer, there is a priority
adapter that dynamically adjusts the service classes
based on the feedback. At the network layer, there
is a service differentiation scheduler that assigns
different network resources to different service classes.
These three components coordinate to assign resources
to multimedia applications. The impact of adaptation
scheme is evaluated on a real ad hoc network testbed
and it is observed that the proposed mechanism
successfully adjusts multimedia applications to meet
delay requirements.
In what follows, we conclude the paper.
12 Conclusion
The provision of QoS in an ad hoc network is a
challenging task and the challenge comes due to inherent
characteristics of such a network. In this paper, we
presented an overview of the research related to the
provision of QoS in mobile ad hoc networks. The
contributions of the paper are as follows.
•
We pointed out issues and challenges in providing
the QoS in a mobile ad hoc network. Some of these
challenges are due to the typical nature of the
problem and others are due to inherent
characteristics and limitations associated with an
ad hoc network.
•
We discussed two classifications of the
methodologies used to provide QoS in mobile ad
hoc networks. The first classification called layered
classification is based on the layer of the protocol
stack to which the corresponding methodology
belongs. The second classification called functional
classification is based on the functionality provided
by the corresponding methodology.
•
We discussed various methodologies reported in the
literature that provide QoS in one form or the
other. We pointed out salient features of each
methodology and compared methodologies that
address somewhat similar issues.
•
After describing the methodologies proposed in the
literature, we pointed out issues that may be
addressed in future for almost all major
functionalities that are necessary for the provision
of QoS in an ad hoc network.
We believe that the survey presented in this paper would
be of some use to academia and professionals so as
to help them in selecting a strategy that is closer to
the requirements of their applications. It may also help
researchers to think directions that are either untouched
or not much work is carried out to address some of the
issues involved in providing QoS in a mobile ad hoc
network.
Acknowledgement
This work was carried out during the tenure of an
ERCIM ‘Alain Bensoussan’ Fellowship Programme.
References
Abdrabu, A. and Zhuang, W. (2006) ‘A position based QoS
routing scheme for UWB mobile ad hoc networks’, IEEE
Journal on Selected Areas in Communications, Vol. 24,
No. 4, April, pp.850–856.
Ahn, G.S., Campbell, A.T., Veres, A. and Sun, L.H. (2002)
‘Supporting service differentiation for real-time and
best-effort traffic in stateless wireless ad hoc networks’,
IEEE Transactions on Mobile Computing, September,
Vol. 1, No. 3, pp.192–207.
Akyildiz, I.F., Akan, O.B. and Morabito, G. (2007) ‘A rate
control scheme for adaptive real-time applications in IP
networks with lossy links and long round trip times’,
IEEE/ACM Transactions on Networking, Vol. 13, No. 3,
June, pp.554–567.
Bashandy, A.R., Chong, E.K.P. and Ghafoor, A. (2005)
‘Generalized quality-of-service routing with resource
allocation’, IEEE Journal on Selected Areas in
Communications, Vol. 23, No. 2, February, pp.450–463.
Bejerano, Y. (2004) ‘Efficient integration of multihop wireless
and wired networks with QoS constraints’, IEEE/ACM
Transactions on Networking, Vol. 12, No. 6, December,
pp.1064–1078.
Bejerano, Y. and Bhatia, R.S. (2006) ‘A framework for fairness
and QoS assurance for current IEEE 802.11 networks
with multiple access points’, IEEE/ACM Transactions on
Networking, Vol. 14, No. 4, August, pp.849–862.
Bianchi, G. (2000) ‘Performance analysis of the IEEE 802.11
distributed coordination function (DCF)’, IEEE Journal
on Selected Areas in Communications, Vol. 18, No. 3,
March, pp.535–547.
21
Chakeres, I.D., Royer, E.M.B. and Macker, J.P. (2007)
‘Perceptive admission control for wireless network quality
of service’, Elsevier Journal on Ad Hoc Networks, Vol. 5,
pp.1129–1148.
Chao, H.J. and Guo, X. (2002) Quality of Service Control
in Hgh Speed Networks, John Wiley and Sons, Inc.,
New York.
Chen, S. and Nahrstedt, L. (1999) ‘Distributed quality of
service routing ad hoc networks’, IEEE Journal on
Selected Areas in Communications, Vol. 17, No. 8, August,
pp.1488–1505.
Chen, L. and Heinzelman, W.B. (2005) ‘QoS aware
routing based on bandwidth estimation for mobile ad
hoc networks’, IEEE Journal on Selected Areas in
Communications, Vol. 23, No. 3, March, pp.561–572.
Chowdhury, K.R., Nandiraju, N., Chanda, P., Agrawal, D.P.
and Zeng, Q.A. (2009) ‘Channel allocation and medium
access control for wireless sensor networks’, Elsevier
Journal on Ad hoc Networks, Vol. 7, No. 2, March,
pp.307–321.
Dovrolis, C., Stiliadis, D. and Ramanathan, P. (2002)
‘Proportional differentiation services: delay differentiation
and packet scheduling’, IEEE/ACM Transactions on
Networking, Vol. 10, No. 1, February, pp.12–26.
Guimaraes, R., Cerda, L., Barcelo, J.M., Garcia, J.,
Voorhaen, M. and Blondia, C. (2009) ‘Quality of service
through bandwidth reservation on multirate ad hoc
wireless networks’, Elsevier Journal on Ad Hoc Networks,
Vol. 7, No. 2, March, pp.388–400.
Gupta, R., Jia, Z., Tung, T. and Walrand, J. (2005)
‘Interference-aware QoS routing (IQRouting) for ad
hoc networks’, Proc. IEEE Global Telecommunications
Conference (Globecom), IEEE Press, St. Louis, Missouri,
pp.2599–2604.
Gupta, R., Musacchio, J. and Walrand, J. (2007) ‘Sufficient
rate constraints for QoS flows in ad hoc networks’,
Elsevier Journal on Ad Hoc Networks, Vol. 5, No. 4, May,
pp.429–443.
Hanzo, L. and Tafazolli, R. (2007) ‘A survey of QoS
routing solutions for mobile ad hoc networks’, IEEE
Communications Surveys, Vol. 9, No. 2, pp.50–70.
He, W., Nahrstedt, K. and Liu, X. (2008) ‘End-to-end delay
control of multimedia applications over multihop wireless
links’, ACM Transactions on Multimedia Computing,
Communications and Applications, Vol. 5, No. 2,
November, Article 16, pp.1–20.
Hui, J. and Devetsikiotis, M. (2005) ‘A unified model
for the performance analysis of IEEE 802.11e EDCA’,
IEEE Transactions on Communications, Vol. 53, No. 9,
September, pp.1498–1510.
Hwang, I.S. and Wu, J.H. (2008) ‘Performance assessment of
service differentiation in IEEE 802.11e wireless LANs’,
InderScience International Journal of Ad hoc and
Ubiquitous Computing (IJAHUC), Vol. 3, No. 1, pp.21–32.
Ito, Y. and Tasaka, S. (2005) ‘Quantitative assessment of
user-level QoS and its mapping’, IEEE Transactions on
Multimedia, Vol. 7, No. 3, June, pp.572–584.
Ivascu, G.I., Pierre, S. and Quintero, A. (2009) ‘QoS routing
with traffic distribution in mobile ad hoc networks’,
Elsevier Journal on Computer Communications, Vol. 32,
No. 2, February, pp.305–316.
22
Jawhar, I. and Wu, J. (2005) ‘QoS support in TDMA-based
mobile ad hoc networks’, Journal of Computer Science
and Technology, Vol. 20, No. 6, November, pp.797–810.
Jia, Y., Nikolaidis, I. and Gburzynski, P. (2004) ‘On the
effectiveness of alternate paths in QoS routing’, John Wiley
International Journal on Communication Systems, Vol. 17,
pp.1–26.
Jia, Z., Gupta, R., Walrand, J. and Varaiya, P. (2005)
‘Bandwidth guaranteed routing for ad hoc networks
with interference considerations’, Proc. Eleventh IEEE
Symposium on Computers and Communications, IEEE CS
Press, Cartagena, Spain, pp.3–9.
Kanodia, V., Li, C., Sabharwal, A., Sadeghi, B. and Nightly, E.
(2002) ‘Distributed priority scheduling and medium access
in ad hoc networks’, ACM Wireless Networks, Vol. 8,
No. 5, September, pp.455–466.
Knopp, R. and Humblet, P.A. (1995) ‘Information
capacity and power control in single-cell multiuser
communications’, Proc. IEEE International Conference
on Communications (ICC), IEEE Press, Seattle,
pp.331–335.
Kuipers, F.A. and Mieghem, P.F.A.V. (2005) ‘Conditions
that impact the complexity of QoS routing’, IEEE/ACM
Transactions on Networking, Vol. 13, No. 4, August,
pp.717–730.
Kumar, S., Raghavan, V.S. and Deng, J. (2006) ‘Medium access
control protocols for ad hoc wireless networks: a survey’,
Elsevier Journal on Ad Hoc Networks, Vol. 4, pp.326–358.
Kumar, A., Altman, E., Miorandi, D. and Goyal, M. (2007)
‘New insights from a fixed-point analysis of single cell
IEEE 802.11 WLANS’, IEEE/ACM Transactions on
Networking, Vol. 15, No. 3, June, pp.588–601.
Kurose, J.F. and Ross, K.W. (2006) Computer Networking:
A Top Down Approach Featuring the Internet, 3rd ed.,
Pearson Education, New Delhi.
Lee, S.B., Ahn, G.S., Zhang, X. and Campbell, A.T. (2000)
‘INSIGNIA: an IP-based quality of service framework
for mobile ad hoc networks’, Journal of Parallel and
Distributed Computing: Special Issue on Wireless and Ad
hoc Networks, Vol. 60, No. 4, April, pp.374–406.
Li, K. and Wang, X. (2006) ‘Cross-layer optimization
for LDPC-coded multirate multiuser systems with QoS
constraints’, IEEE Transactions on Signal Processing,
Vol. 54, No. 7, July, pp.2567–2578.
Liang, B. and Dong, M. (2007) ‘Packet prioritization in
multihop latency aware scheduling for delay constrained
communication’, IEEE Journal on Selected Areas in
Communications, Vol. 25, No. 4, May, pp.819–830.
Liao, W.H., Tseng, Y.C. and Shih, K.P. (2002)
‘A TDMA-based bandwidth reservation protocol for QoS
routing in a mobile ad hoc network’, IEEE International
Conference on Communications (ICC), IEEE Press,
New York, Vol. 5, pp.3186–3190.
Lin, C.R. and Liu, J.S. (1999) ‘QoS routing in ad hoc
wireless networks’, IEEE Journal on Selected Areas in
Communications, Vol. 17, No. 8, August, pp.1426–1438.
Liu, H., Jia, X., Li, D. and Lee, C.H. (2005) ‘Bandwidth
guaranteed call admission in TDMA/CDMA ad hoc
wireless networks’, Elsevier Journal on Ad Hoc Networks,
Vol. 3, pp.689–701.
Liu, Q., Wang, X. and Giannakis, G.B. (2006) ‘A cross-layer
scheduling algorithm with QoS support in wireless
networks’, IEEE Transactions on Vehicular Technology,
Vol. 55, No. 3, May, pp.839–847.
Luo, H. and Lu, S. (2005) ‘A topology-independent wireless
fair queueing model in ad hoc networks’, IEEE Journal on
Selected Areas in Communications, Vol. 23, No. 3, March,
pp.585–597.
Meshkati, F., Goldsmith, A.J., Poor, H.V. and Schwartz,
S.C. (2007) ‘A game-theoretic approach to energyefficient modulation in CDMA networks with delay
QoS constraints’, IEEE Journal on Selected Areas in
Communications, Vol. 25, No. 6, August, pp.1069–1078.
Mieghem, P.F.A.V., Neve, H.D. and Kuipers, F.A. (2001)
‘Hop-by-hop quality of service routing’, Elsevier Journal
on Computer Networks, Vol. 37, Nos. 3–4, pp.407–423.
Mieghem, P.F.A.V. and Kuipers, F.A. (2004) ‘Concepts of
exact QoS routing algorithms’, IEEE/ACM Transactions
on Networking, Vol. 12, No. 5, October, pp.851–864.
Movsichoff, B.A., Lagoa, C.M. and Che, H. (2007) ‘End-to-end
optimal algorithms for integrated QoS, traffic engineering,
and failure recovery’, IEEE/ACM Transactions on
Networking, Vol. 15, No. 4, August, pp.813–823.
O’Farrell, T. and Omiyi, P. (2005) ‘Low-complexity medium
access control protocols for QoS support in thirdgeneration radio access networks’, IEEE Transactions
on Wireless Communications, Vol. 4, No. 2, March,
pp.743–756.
Park, E.C. and Kim, D.Y. (2007) ‘Improving quality of service
and assuring fairness in WLAN access networks’, IEEE
Transactions on Mobile Computing, Vol. 6, No. 4, April,
pp.337–350.
Paschos, G.S., Politis, I.D. and Kotsopoulos, S.A. (2005)
‘A quality of service negotiation-based admission control
scheme for WCDMA mobile wireless multiclass services’,
IEEE Transactions on Vehicular Technology, Vol. 54,
Patil, S. and Veciana, G. (2007) ‘Managing resources and
quality of service in heterogeneous wireless systems
exploiting opportunism’, IEEE/ACM Transactions on
Networking, Vol. 15, No. 5, October, pp.1046–1058.
Perkins, D.D. and Hughes, H.D. (2002) ‘A survey on quality
of service support for mobile ad hoc networks’, Kluwer
Academic Publishers Journal on Wireless Communications
and Mobile Computing, Vol. 2, No. 5, September,
pp.503–513.
Reddy, T.B., Karthigeyan, I., Manoj, B.S. and Murthy, C.S.R.
(2006) ‘Quality of service provisioning in ad hoc wireless
networks: a survey of issues and solutions’, Elsevier
Journal on Ad hoc Networks, Vol. 4, No. 1, January,
pp.83–124.
Reddy, T.B., John, J.P. and Murthy, C.S.R. (2007) ‘Providing
MAC QoS for multimedia traffic in 802.11e based
multihop ad hoc wireless networks’, Elsevier Journal on
Computer Networks, Vol. 51, pp.153–176.
Sabella, R. and Iovanna, P. (2006) ‘Self-adaptation in
next-generation internet networks: how to react to traffic
changes while respecting QoS?’, IEEE Transactions on
Systems, Man, and Cybernetics–Part B: Cybernetics,
Vol. 36, No. 6, pp.1218–1229.
Sarr, C., Chaudet, C., Chelius, G. and Lassous, I.G. (2008)
‘Bandwidth estimation for IEEE 802.11-based ad hoc
networks’, IEEE Transactions on Mobile Computing,
Vol. 7, No. 10, October, pp.1228–1241.
Shih, K.P., Chang, C.Y., Chen, Y.D. and Chuang, T.H.
(2006) ‘Dynamic bandwidth allocation for QoS routing on
TDMA-based mobile ad hoc networks’, Elsevier Journal
on Computer Communications, Vol. 29, pp.1316–1329.
Singh, S., Acharya, P.A.K., Madhow, U. and Royer, E.M.B.
(2007) ‘Sticky CSMA/CA: implicit synchronization and
real-time QoS in mesh networks’, Elsevier Journal on
Ad Hoc Networks, Vol. 5, No. 6, August, pp.744–768.
Sivakumar, R., Sinha, P. and Bharghavan, V. (1999) ‘CEDAR:
A core-extraction distributed ad hoc routing algorithm’,
IEEE Journal on Selected Areas in Communications,
Vol. 17, No. 8, August, pp.1454–1465.
Small, T. and Haas, Z.J. (2007) ‘Quality of service and capacity
in constrained intermittent-connectivity networks’, IEEE
Transactions on Mobile Computing, Vol. 6, No. 7, July,
pp.803–814.
Sobrinho, J. and Krishnakumar, A.S. (1999) ‘Quality of service
in ad hoc carrier sense multiple access wireless networks’,
IEEE Journal on Selected Areas in Communications,
Vol. 17, No. 8, August, pp.1353–1368.
Sun, W. and Shin, K.G. (2005) ‘End-to-end delay bounds
for traffic aggregates under guaranteed-rate scheduling
algorithms’, IEEE/ACM Transactions on Networking,
Vol. 13, No. 5, October, pp.1188–1201.
Sun, W., Yamaguchi, H., Yukimasa, K. and Kusumoto, S.
(2006) ‘GVGrid: a QoS routing protocol for vehicular
ad hoc networks’, Proc. Fourteenth IEEE International
Workshop on Quality of Service (IWQoS), IEEE Press,
New Haven, Connecticut, pp.130–139.
Tang, J. and Zhang, X. (2007) ‘Cross-layer resource allocation
over wireless relay networks for quality of service
provisioning’, IEEE Journal on Selected Areas in
Communications, Vol. 25, No. 4, May, pp.645–656.
Tasaka, S. and Ishibashi, Y. (2002) ‘Mutually compensatory
property of multimedia QoS’, Proc. IEEE International
Conference on Communications (ICC), IEEE Press,
Anchorage, Alaska, pp.1105–1111.
Tay, Y.C. and Chua, K.C. (2001) ‘A capacity analysis for
the IEEE 802.11 MAC protocol’, Springer Journal on
Wireless Networks, Vol. 7, No. 2, July, pp.159–171.
Tsigkas, O. and Pavlidou, F.N. (2006) ‘A dynamic priority
MAC protocol for time-bounded services in wireless
networks’, International Journal of Wireless Personal
Communications, Vol. 38, No. 3, pp.301–324.
Tsigkas, O. and Pavlidou, F.N. (2008) ‘An adaptive medium
access control protocol using m-ary tree algorithms for
quality of service support in single cell ad hoc networks,
Elsevier Journal on Ad Hoc Networks, Vol. 6, No. 2, April,
pp.245–259.
Wang, Z. and Crowcroft, J. (1996) ‘Quality of service routing
for supporting multimedia applications’, IEEE Journal
of Selected Areas in Communications, Vol. 14, No. 7,
pp.1228–1234.
Wang, K.C. and Ramanathan, P. (2003) ‘End-to-end delay
assurances in multihop wireless local area networks’,
Proc. IEEE Conference on Global Telecommunications
(GLOBECOM), San Francisco, California, pp.2962–2966.
23
Wang, K.C. and Ramanathan, P. (2005) ‘QoS assurances
through class selection and proportional differentiation in
wireless networks’, IEEE Journal on Selected Areas in
Communications, Vol. 23, No. 3, March, pp.573–584.
Wang, P., Jiang, H. and Zhuang, W. (2008) ‘A new MAC
scheme supporting voice/data traffic in wireless ad hoc
Vol. 7, No. 12, December, pp.1491–1503.
Wang, W., Chatterjee, M. and Kwiat, K. (2009) ‘User input
driven QoS management in ad hoc networks’, Elsevier
Journal on Computer Communications, Vol. 32, No. 11,
July, pp.1306–1315.
Wu, D. and Negi, R. (2005) ‘Utilizing multiuser diversity
for efficient support of quality of service over a fading
channel’, IEEE Transactions on Vehicular Technology,
Vol. 54, No. 3, May, pp.1198–1206.
Wu, H. and Jia, X. (2007) ‘QoS multicast routing by using
multiple paths/trees in wireless ad hoc networks’, Elsevier
Journal on Ad hoc Networks, Vol. 5, No. 5, July,
pp.600–612.
Xiao, H., Seah, K.G., Lo, A. and Chua, K.C. (2000) ‘A flexible
quality of service model for mobile ad hoc networks’, Proc.
Fifty First IEEE Vehicular Technology Conference IEEE
Press, Tokyo, Vol. 1, pp.445–449.
Xiao, Y. and Li, H. (2004) ‘Local data control and admission
control for QoS support in wireless ad hoc networks’,
IEEE Transactions on Wireless Communications, Vol. 53,
Xiao, Y. (2005) ‘Performance analysis of priority schemes for
IEEE 802.11 and IEEE 802.11e wireless LANs’, IEEE
Transactions on Wireless Communications, Vol. 4, No. 4,
July, pp.1506–1515.
Xu, Y. and Guerin, R. (2005) ‘Individual QoS versus aggregate
QoS: a loss performance study’, IEEE/ACM Transactions
on Networking, Vol. 13, No. 2, April, pp.370–383.
Xu, J. and Lipton, R.J. (2005) ‘On fundamental tradeoffs
between delay bounds and computational complexity in
packet scheduling algorithms’, IEEE/ACM Transactions
on Networking, Vol. 13, No. 1, February, pp.15–28.
Xue, G., Sen, A., Zhang, W., Tang, J. and Thulasiraman,
K. (2007) ‘Finding a path subject to many additive QoS
constraints’, IEEE/ACM Transactions on Networking,
Vol. 15, No. 1, February, pp.201–211.
Yang, Y. and Kravets, R. (2005) ‘Contention-aware admission
control for ad hoc networks’, IEEE Transactions on
Mobile Computing, Vol. 4, No. 4, August, pp.850–856.
Yang, Y. and Kravets, R. (2007) ‘Throughput guarantees for
multipriority traffic in ad hoc networks’, Elsevier Journal
on Ad Hoc Networks, Vol. 5, pp.228–253.
Zhai, H., Chen, X. and Fang, Y. (2005) ‘How well can the
IEEE 802.11 wireless LAN support quality of service?’,
IEEE Transactions on Wireless Communications, Vol. 4,
No. 6, November, pp.3084–3094.
Zhang, B. and Mouftah, H.T. (2005) ‘QoS routing for wireless
ad hoc networks: problems, algorithms, and protocols’,
IEEE Communications Magazine, October, pp.110–117.
Zhao, Q. and Tsang, D.H.K. (2006) ‘An equal-spacing-based
design for QoS guarantee in IEEE 802.11e HCCA wireless
Vol. 7, No. 12, December, pp.1491–1503.
24
Zhu, C. and Corson, M.S. (2002) ‘QoS routing for mobile
ad hoc networks’, Proc. Twenty First Annual Joint
Conference of IEEE Computer and Communication
Societies (INFOCOM), IEEE CS Press, New York, Vol. 2,
pp.958–967.
Znati, T.F. and Melhem, R. (2004) ‘Node delay assignment
strategies to support end-to-end delay requirements in
heterogeneous networks’, IEEE/ACM Transactions on
Networking, Vol. 12, No. 5, October, pp.879–892.
Notes
1
Note that a constraint that is multiplicative can be converted
to an additive constraint by using logarithms.
2
In this paper, some of the papers that describe the provision
of QoS in case of networks with a centralised infrastructure
have also been considered with the view that those works can
serve as the foundations or basics for the issues and challenges
that are to be faced in providing QoS in case of an ad hoc
network too. The issues and challenges to be tackled in case
of an ad hoc network are more stringent as compared to other
types of networks whether wired or wireless.
3
The literal meaning of the word ‘insignia’ is “a badge or
distinguishing mark”. However, INSIGNIA here stands for
in-band signalling.
4
Although, CMAC is proposed for sensor networks, we believe
that it can also be used in ad hoc networks in situations when
either the nodes do not move or the mobility of nodes is very
low.
5
Though the strategies presented in Znati and Melhem (2004)
are for the internet, we believe that similar kinds of problems
can also be faced while scheduling packets at a node in an ad
hoc network.

Quality of Service in mobile ad hoc networks: a

Transcription

Similar documents

from polimi.it - Dipartimento di Elettronica ed informazione

Quality of Service Monitoring Tool Tahitii OBIOHA Network Planning & Optimisation Engineer

Wireless and Mobile Networks Syllabus

Seamless quality of Service for the Users for the Users - Docbox

0.1 Adaptive Multimedia Streaming over Wireless Links

QoS for broadcasting

Zaplet 3: Streamlining Collaborative Business Processes

fiDDi[ im ^ m ta?

Voice over IP Implementation

Reliable network solutions enabling Smart Grid Applications