An Empirical Study of Effective Capacity Throughputs in 802.11 Wireless Networks

Abstract

Current literature defines the effective capacity of a wireless link as the maximum throughput that can be supported while meeting specific Quality of Service targets on packet delay. This metric can be harnessed for a wide variety of QoS control routines within wireless networks such as traffic optimisation and delay sensitive admission control and routing. However to date, no empirical evaluation of the effective capacity of 802.11 wireless links has been carried out. We present an empirical study of the effective capacity throughput of 802.11 wireless links under a number of network scenarios. We evaluate an analytical effective capacity model and compare the result with an empirical evaluation. We find that with an accurate measurement of the channel service delay, the effective capacity model can approximate the empirical measurement quite well. We also evaluate the relationship between the effective bandwidth of multimedia traffic and demonstrate that when the effective bandwidth exceeds the effective capacity threshold of a wireless link, the probability of QoS violations increases. We conclude that the effective capacity measurement is usable within an operational setting, and can lead to optimized utilization of bandwidth in a wide range of delay sensitive control operations.

Full text

An Empirical Study of Effective Capacity
Throughputs in 802.11 Wireless Networks
Alan Davy∗, Brian Meskill∗, Jordi Domingo-Pascual†
∗Waterford Institute of Technology,
Cork Road, Waterford, Ireland.
{adavy, bmeskill}@tssg.org
†Universitat Polit`
ecnica de Catalunya,
Barcelona, Spain.
jordi.domingo@ac.upc.edu
Abstract—Current literature defines the effective capacity of a
wireless link as the maximum throughput that can be supported
while meeting specific Quality of Service targets on packet
delay. This metric can be harnessed for a wide variety of
QoS control routines within wireless networks such as traffic
optimisation and delay sensitive admission control and routing.
However to date, no empirical evaluation of the effective capacity
of 802.11 wireless links has been carried out. We present an
empirical study of the effective capacity throughput of 802.11
wireless links under a number of network scenarios. We evaluate
an analytical effective capacity model and compare the result
with an empirical evaluation. We find that with an accurate
measurement of the channel service delay, the effective capacity
model can approximate the empirical measurement quite well.
We also evaluate the relationship between the effective bandwidth
of multimedia traffic and demonstrate that when the effective
bandwidth exceeds the effective capacity threshold of a wireless
link, the probability of QoS violations increases. We conclude
that the effective capacity measurement is usable within an
operational setting, and can lead to optimized utilization of
bandwidth in a wide range of delay sensitive control operations.
I. INTRODUCTION
In recent years there has been a dramatic increase in the
delivery of delay sensitive multimedia services over wireless
networks. With the introduction of Quality of Service (QoS)
provisioning support in standards such as 802.11n [1] and
3GPP QoS architecture for LTE [2], wireless networks now
have the capability of offering QoS provisioning of delay sen-
sitive services to end user devices. However there still remains
the challenge of optimizing these QoS provisioning schemes
to maximize utilization of resources while maintaining QoS
targets. A major challenge in this area is the issue of accurately
measuring the availability of resources over wireless links to
meet the QoS requirements of delay sensitive services given
the varying state of the wireless link.
The objective is to accurately estimate the maximum service
rate that can be supported by a wireless link while meeting
specific QoS bounds on packet delay. Current literature has
termed this the effective capacity (EC) of a wireless link [3].
An accurate estimate of EC can ensure that QoS provisioning
of multimedia traffic over a wireless link can be optimal with
respect to the wireless link properties. As wireless channels are
commonly shared mediums which are loss prone and fading,
this can impact on the capacity of the wireless link over
time. This may also be specific to the particular environmental
context of operation (e.g. mobility, urban setting, neighbor
nodes, cross traffic), and the coding and modulation techniques
employed. When estimating the EC of a wireless link, all
factors impacting capacity must be accounted for, which can
be very challenging indeed.
Recent research [3] has proposed a method of modeling
the EC of a wireless fading channel through analysis of the
data link layer queue behavior. The authors have demonstrated
that this model in theory can accurately provide statistical
guarantees on QoS targets over wireless links. This model
has spawned a wide range of theoretical research papers
in the area of wireless resource management with a view
to QoS provisioning of delay sensitive services. However
to the best of our knowledge, no validation of this model
has been carried out on IEEE 802.11 wireless links. We
aim to evaluate this model for a range of 802.11 physical
technologies and varying environmental conditions using the
OPNET ModelerTMsimulation software [4].
The objective of this paper is to first validate the theoretical
effective capacity model for IEEE 802.11 wireless links,
studying the impact of various assumptions within the EC
model and then evaluate the EC capabilities and performance
of the IEEE 802.11 wireless links under varying environmen-
tal contexts. We also evaluate the relationship between the
bandwidth requirements of multimedia traffic and the EC of
a wireless link. In this paper we carry out a simulation based
analysis of effective capacity measurements over 802.11 wire-
less links, where we vary both the range of communication,
the channel coding and modulation schemes, the number of
nodes competing for the shared channel and also the level of
cross traffic within the shared channel.
We compare the empirically collected packet delay distribu-
tions at the wireless data link layer to the predicted effective
capacity model to evaluate the QoS provisioning capabilities
of the wireless links. We find that under ideal conditions with
accurate measurement of channel service delay, the effective
capacity model proposed by [3] matches well to the empirical
results. However as the level of channel interference increases
Globecom 2012 - Communications QoS, Reliability and Modelling Symposium
1788

the accuracy of the model can reduce to a margin of error of
up to 9%. In such a case a more in-depth empirical analysis
of the data link layer queue is required to ensure an accurate
measurement. We also find that when the EB of a multimedia
traffic flow increases beyond the measured EC of a wireless
link, the probability of QoS target violations increases.
The paper is organized as follows. Section II provides a
discussion of related work in the area of EC estimation and
usage and motivates the particular EC model under analysis.
Section III provides an overview of the EC model. Section
IV discusses the measurement methodology used to estimate
the EC of a wireless link. Section V defines the experimental
setup used for this evaluation. Section VI discusses the results
attained and evaluates both the validity of the EC model and
the impact various factors have on this measurement. Finally
section VII will conclude the paper with a discussion and
outline future work.
II. RE LATE D WOR K
The meeting of statistical delay target in packet based com-
munication is a fundamental requirement of delay sensitive
services such as voice and video delivery. Within a wired
network, there has been much research into understanding the
resource requirements of various traffic types to ensure delay
targets are met. The concept of effective bandwidth has been
extensively studied in the past, for example in [5] and [6].
This concept captures the asymptotic stochastic behavior of
queuing delay for a traffic source. Understanding the effective
bandwidth requirements of traffic within a network can aid
a wide range of management processes in the provisioning
of QoS within a network, such as admission control [7] and
network planning [8], [9]. Active based measurement tools
have also been proposed to measure the available bandwidth
between two hosts within a wired network with the aim of
improving end to end QoS provisioning [10].
More recent research has emerged that focuses on the QoS
provisioning capabilities of a wireless link [3]. In contrast to
the deterministic nature of capacity within a wired network,
where components of delay across a link (e.g. transmission and
propagation delay) are static, the capacity of a wireless link
is considered non-deterministic. In this respect, deterministic
QoS targets can not be guaranteed. Wu and Negi [3] proposed
an EC model which determined the statistical probability of
meeting QoS targets of packet delay across a wireless link.
This model is based on the dual of the effective bandwidth
model of a time-varying traffic source. The EC model captures
the asymptotic stochastic behavior of queuing delay for a time-
varying wireless link. By understanding these properties of the
wireless link, one can appropriately manage traffic to ensure
specific QoS targets on packet delay are met.
The effective capacity model proposed by Wu and Negi [3]
has motivated a broad range of research papers investigating
the provisioning of delay sensitive services over wireless links
such as voice and video. The authors of [11],[12],[13] and
[14] focus on using the EC model of a wireless link as the
basis of a cross-layer approach to power and channel allocation
management for delay sensitive services over wireless links.
The effective capacity model has also been used as a delay
sensitive performance metric over different wireless MAC
techniques. For example in [15],[16] and [17], the authors
analyzed the EC of cognitive radio channels given varying
primary and secondary user scenarios. The authors in [18]
analyzed the EC capabilities of a time division downlink
scheduling system. Delay sensitive routing within a wireless
multi hop network was also studied in [19] where the QoS
capabilities of the wireless links were modeled using the EC
model. The diversity and utilization of the proposed EC model
in delay sensitive wireless networking scenarios motivates the
authors to choose this model as the bases of the evaluation of
EC over 802.11 wireless links.
These papers are primarily theoretical in nature and model
the wireless link using Rayleigh fading channel models [20].
However no papers have attempted to evaluate the model
empirically over 802.11 wireless links. The objective of this
paper is to focus on analyzing the EC of the IEEE 802.11 MAC
layer for a variety of settings within a simulated environment,
and to evaluate the effective capacity of the wireless links. We
also evaluate the relationship between EB of multimedia traffic
and the EC of wireless links, to demonstrate the impact on
packet delay, motivating the appropriateness of such a metric
for delay sensitive QoS control operations. In the following
section, we present the necessary knowledge to understand
the effective capacity theory and the proposed measurement
methodology.
III. EFFE CT IV E CAPACITY MODEL
The theory of EC was developed by Wu and Negi [3] to
define a data link layer model of buffer overflow probability for
wireless links. This theory is the dual of effective bandwidth,
which has been studied for many years in the area of wired
networks [5], [6]. As the dual of this theory, the EC defines
the maximum arrival rate a wireless link can support while
maintaining specific targets of packet delay.
It has been shown [3] that the EC of a queuing system with
time varying service rate process S(t)at time t, being supplied
by a traffic source of constant rate µis:
EC (θ) = −lim
t→∞
1
tθ log E[γe−θS(t)](1)
Using the theory of large deviations, Wu and Negi demon-
strates that the delay distribution of this service process
satisfies:
Pr{D(t)≥Dmax} ≈ γ(µ)e−θ(µ)Dmax (2)
where both γ(µ)and θ(µ)are functions of the constant source
rate µ. The justification for using a constant arrival rate in this
model has been discussed in [21]. In summary, the authors
demonstrate that if a constant arrival rate arrives at a queue
system with a time varying service process, the effective
capacity of the system can be measured through analysis of
the system queue behavior. In equation (2), the parameters
1789

{γ(µ), θ(µ)}provide the statistical characterization of the
queue behavior of a time varying service process being offered
a constant arrival rate µ.
Assuming that (2) holds at equality, the authors of [22]
have shown that through estimating the expected queue size
E[Q(t)], expected utilization of the queue E[Sr(t)] and the
average remaining service time of a packet τof a given queue
system being supplied with an arrival rate of µ, the parameters
γ(µ)and θ(µ)can be measured as follows:
γ(µ)
θ(µ)=E[D(t)] (3)
=τE[S r(t)] + E[Q(t)]
µ(4)
where Sr(t)∈ {0,1}is the indicator as to whether a packet
is in service or not at a given random time tand Q(t)is
the size of the queue in bits at a random time t.τin (4)
is the average remaining service time for a packet in service
as seen by an arriving packet (conditional on there being a
packet in service). D(t)is the total delay experienced by a
packet combining queue delay with channel service delay at a
random time t. We can estimate γ(µ), the channel utilization,
by noting any instance the queue is empty:
E[Sr(t)] = P r(Q(t)>0) = γ(µ)(5)
We can use equation (3) to obtain θ(µ)as a function of delay
and channel utilization through the following relationship.
θ(µ) = γ(µ)
E[D(t)] (6)
We can use equation (4) to obtain θ(µ)as a function of
the queue size, channel service delay and channel utilization,
through the following relationship.
TABLE I
EFFE CTI VE CA PACIT Y MODE L NOTATIO N
Notation Description
µThroughput rate of at traffic source.
Dmax Target upper bound on packet delay.
S(t)Amount of data serviced by the service process at time [0, t).
Q(t)Amount of data in the queue waiting to be serviced at time
[0, t).
D(t)Delay experienced by a source packet at time [0, t].
Sr(t)An indicator at to whether a packet is in service or not at a
randomly chosen time t.
γProbability of a non-empty queue.
θQoS exponent.
EC (θ)Effective capacity of a service process given θ.
τThe average remaining service time of a packet in service
(assuming there is one).
ˆγA sample probability of a non-empty queue.
ˆqA sample measure of average queue length (in bits).
ˆτA sample measure of the average remaining service time of
a packet in service (assuming there is one).
ˆ
θAn empirical estimation of the QoS exponent.
εTarget QoS violation probability.
θ(µ) = γ(µ)µ
µτ(µ)γ(µ) + E[Q(t)] (7)
With a method of calculating both γ(µ)and θ(µ)from
equations (5) and (6) or (7) we can then substitute these values
into the effective capacity model in equation (2) to estimate the
probability of queue overflow given a particular packet delay
target Dmax. The two options for calculating θ(µ)depend on
whether one has direct access to calculating the total delay
experienced by a packet D(t)or just access to the queue size
Q(t)at any given random time t. We now discuss practical
methods of estimating the effective capacity model parameters.
IV. ESTIMATION OF EFFECTIVE CAPACIT Y MOD EL
PARAMETERS
The estimation of the EC parameters are based on observa-
tions at the data-link layer queue over a measurement interval
T. The basic procedure is as follows, over interval T, a number
of samples are taken, say N. At the nth sampling interval,
the following quantities are recorded: Srnthe indicator of
whether a packet is in service Srn∈ {0,1},Qnthe number
of bits in the queue, Dnthe total delay experienced by a
packet including queue delay and contention delay, and Tn
the remaining service time of a packet in service, if one is in
service. The following sample means are calculated:
ˆγ=1
N
N
X
n=1
Srn(8)
ˆq=1
N
N
X
n=1
Qn(9)
ˆ
d=1
N
N
X
n=1
Dn(10)
ˆτ=1
ˆγN
N
X
n=1
Tn.(11)
Then ˆ
θis calculated via 6 or 7 using:
ˆ
θ=ˆγ
ˆ
d(12)
ˆ
θ=ˆγµ
µˆτˆγ+ ˆq≈ˆγµ
ˆγ/2 + ˆq(13)
As the remaining service time of a packet in service is quite
difficult to measure in practice due to random contention delay,
authors of [22] claim that ˆτcan be approximate with 1/2µ.
This implies that the remaining service time of a packet is a
function of the arrival rate µ. These equations make up the
channel estimation algorithm used to estimate the EC model
parameters {γ(µ), θ(µ)}. They will be used to predict the
Quality of Service provisioning capabilities of a time varying
wireless channel by approximating (2) with:
Pr{D(t)≥Dmax} ≈ ˆγe−ˆ
θDmax .(14)
1790

AP Data
Source
Wireless channel Wired link
d 100Mbps
Wireless
Stations
Fig. 1. Network Topology
We therefore wish to find an approximate arrival rate µfor
a given Dmax that will satisfy a given probability violation
target ε. The objective is to use the model to iteratively test
different values of µand to report when the probability of
queue overflow is equal to a predefined probability violation
target ε.
V. SIMULATION SET UP
We use the OPNET ModelerTM simulation environment [4]
to both study the EC model and to evaluate the EC of 802.11
wireless links under varying network scenarios. We use a
single hop topology as depicted in Fig. 1 for our network setup.
The topology comprises of one data source node connected to
a wireless access point (AP) using a wired link of throughput
100Mbps. One or more wireless stations (WS) are connected
to this AP using the specified wireless channel. The WSs
are located equal distances from the AP requiring only one
hop from the AP to receive data. All 802.11 MAC settings
are configured to standard. The following 802.11 MACs are
evaluated; Direct Sequence (802.11b), OFDM (802.11a).
Traffic is generated by the data source and is directed
towards the WS through the AP to determine EC throughput.
This traffic is UDP with a fixed payload of 1000 bytes.
This is to ensure that each packet is small enough to avoid
any fragmentation at the MAC layer. The throughout rate of
the traffic is therefore controlled by the inter-departure time
of each packet, which is uniformly distributed to ensure a
constant bit rate throughout the simulation.
The AP model has been extended to collect the required
samples to calculate the EC parameters as outline in section
IV, namely the queue size ( Qn) and service indicator (Srn).
Samples are taken every 0.01 seconds for a duration of 200
seconds. To ensure the simulation has reached a steady state,
we run each simulation for 400 seconds and use samples
taken for the latter 200 seconds. This therefore means that
a maximum of 20000 samples are used for the calculation of
each parameter. As the EC model calls for the estimation of
the remaining service time of a packet τ, we do not collect
this sample explicitly. We will however collect the total delay
experienced by a packet at the wireless interface of the AP
,(Dn). This is to evaluate the accuracy of the model with
complete knowledge and to evaluate the assumption that the
parameter τcan be estimated with the EC model.
For each experiment we use the samples collected to esti-
mate the parameters ˆγand ˆ
θusing equations (8) and (12) or
(13). Table. II depicts the set of parameters we will use within
TABLE II
EXP ERI ME NT PARAMETERS
Distance (Meters) 50, 100
Rates (Mbps) 11, 54
Propagation Model Free Space Path Loss
Transmitter Power (Watts) 0.005
Transmitter Antenna Gain (dBi) 2
Receiver Antenna Gain (dBi) 2
MAC Protocols 802.11b, 802.11a
Nodes 1,3,5
Cross Traffic (bps) 600K, 1M, 3M, 5M
QoS delay targets Dmax = 0.01 sec, ε= 0.0001
the experimental evaluation. Each experiment proceeds as
follows: for a given set of parameters, an inter-departure time
is chosen for the UPD constant bit rate traffic source. Traffic
is then sent towards a single WS passing through the single
AP. Samples are taken from the AP data link layer queue,
and are processed offline. We present our proposed empirical
measurement methodology within the following section.
A. Empirical Measurement Methodology
To evaluate the accuracy of the EC model, we compare all
estimates to a base set of empirical data. As the EC model aims
to estimate the packet delay distribution through the use of the
exponential function denoted in equation 2, we will directly
collect the packet delay distribution Dnfrom the AP within
the simulation also at a sampling interval of 0.01 seconds. This
is achieved by collecting the media access delay parameter of
the wireless link within the OPNET modeler. OPNET provides
this parameter directly and is defined as the total of queuing
delay and contention delay of all packets transmitted by the
MAC. From this sample set, we can create a cumulative
distribution function (CDF) for the following Pr{Dn≥D}
where D∈ {0. . . Dmax}. We can test our QoS target against
both the CDF and the effective capacity model to determine
the probability of packet delay being greater than the Dmax
target. At the end of each experiment, the collected data is
analyzed to determine if the chosen traffic rate successfully
meets the QoS target imposed. Fig. 2 depicts the search rule
used to identify whether the EC has been found. Once found,
the experiments are complete, however if the probability of
violating the QoS target Dmax is either higher or lower than
the violation target εthen the search will continue and a new
simulation will be started with a new traffic inter-departure
time. We employ a straight forward binary search algorithm to
perform the search between the traffic rates of 0Mbps and the
maximum theoretical rate of the physical technology within
the experiment.
VI. RE SU LTS
We began our evaluation of the EC model within an ideal
scenario i.e. with only one WS connected to one AP and no
competing cross traffic. This serves to determine the accuracy
of the EC model in ideal conditions. The evaluation then
progresses with the addition of various contending cross traffic
scenarios at two different distances. Table. III depicts four
experimental results comparing the empirically collected data
1791

if [(Pr(D≥Dmax)−ε)]+<(ε∗error)then
QoS Target Matched
else if Pr(D≥Dmax)> ε then
QoS Target Too High
else if Pr(D≥Dmax)< ε then
QoS Target Too Low
end if
Fig. 2. Effective Capacity Search Rule
to the theoretical EC model with perfect estimation of channel
service delay through the collection of parameter Dn.
The two wireless MAC standards were evaluated, Direct
Sequence (802.11b) and OFDM (802.11a), with transmission
rates of 11Mbps and 54Mbps respectively. We first evaluated
the empirical data to the EC model with perfect channel ser-
vice delay measurement at the lower rate of 11Mbps. We found
the model to be highly accurate in estimating the EC with a
relative error of 0.075%. EC was found to be about 4.86Mbps
in both cases. In proportion to the maximum theoretical rate
of 11Mbps, the EC is a ratio of 0.442 approximately.
As cross traffic was introduced, there was a very slight
increase in the relative error of approximately 1.2% between
the empirical EC value and the EC model estimation. We also
notice that as the cross traffic increases, the EC of the wireless
link decreases in both the empirical data and the EC model,
as can be seen in Table. III. This demonstrates that the EC
model can accurately capture the impact cross traffic has on
the EC of a wireless link.
Fig. VI depicts a semi-log scale graph of the empirical data
of the packet delay distribution collected for Experiment 1
with 3 cross traffic nodes generating 1Mbps traffic each and
sending this traffic towards the AP. The graph also depicts the
distribution generated by the EC model at the same arrival
rate. One can see that at this particular arrival rate, the
empirical distribution shows that 0.001 of packets are delayed
greater than 0.01 seconds. However the EC estimates that
approximately 0.0004 of packets are delayed greater than 0.01
seconds. As can be seen in the results in Table. III, the EC
Fig. 3. Semi-Log Scale Empirical Data vs EC Model @ 1,934 Mbps for
Experiment 1
TABLE III
EMPIRICAL EC V S EC MO DE L
Experiment 1: 11Mbps, 802.11b links, distance of 50 Meters
Nodes Empirical EC
(Mbps)
EC Model
(Mbps)
Relative Error
0 4.867 4.863 0.08
1 @ 3Mbps 2.019 2.094 3.72
3 @ 1Mbps 1.933 2.041 5.56
5 @ 1Mbps 1.851 2.006 8.38
Experiment 2: 11Mbps, 802.11b links, distance of 100 Meters
Nodes Empirical EC
(Mbps)
EC Model
(Mbps)
Relative Error
0 4.864 4.862 0.06
3 @ 1Mbps 1.890 2.062 9.09
5 @ 600Kbps 3.137 3.168 1.00
5 @ 1Mbps 1.890 2.065 9.23
Experiment 3: 54Mbps, 802.11a links, distance of 50 Meters
Nodes Empirical EC
(Mbps)
EC Model
(Mbps)
Relative Error
0 24.495 24.476 0.07
3 @ 5Mbps 9.755 10.019 1.27
5 @ 3Mbps 15.906 15.905 0.002
5 @ 5Mbps 9.755 9.886 1.34
Experiment 4: 54Mbps, 802.11a links, distance of 100 Meters
Nodes Empirical EC
(Mbps)
EC Model
(Mbps)
Relative Error
0 24.528 24.510 0.07
3 @ 5Mbps 9.439 9.559 1.27
5 @ 3Mbps 15.609 15.942 2.13
5 @ 5Mbps 9.334 9.729 4.23
TABLE IV
EFFE CTI VE CA PACIT Y MODE L PARAMETERS USIN G Qn: EX PER IM ENT 3
Nodes EC Model
(Mbps)
ˆ
θˆγRelative Error
0 24.469 697.70 0.98 0.11
3 @ 5Mbps 9.927 627.58 0.59 1.76
5 @ 5Mbps 9.877 635.56 0.60 1.24
model conservatively estimates the EC to be at 2.041 Mbps
which is a relative error of 5.56%.
At the higher transmission rate of 54Mbps, the EC model
with perfect channel service delay measurements using the
media access delay parameter Dn, achieved high accuracy in
comparison to the empirical data with relative error as low as
0.072% with no interfering cross traffic. EC in both cases of
50 and 100 meters was found to be approximately 24.5Mbps
which in comparison to the theoretical maximum is a ratio of
0.454. As cross traffic was introduced, the accuracy of the EC
model estimation in comparison to the empirical data remained
low with a relative error of up to 4.234%.
At the higher transmission rate, however, solely using the
queue size and the service indicator to calculate the EC model
parameters, returned a high level of accuracy which is depicted
in Table. IV for Experiment 3 parameters. We repeated the
experiment for the distance of 100 meters also and found the
results to be consistent. In addition to this, as cross traffic is
introduced the accuracy of the EC estimates in comparison
1792

to the empirical data still remains low at about 1.7% relative
error.
VII. CONCLUSIONS
We have presented an empirical analysis of the effective
capacity throughput of 802.11 wireless links under varying
network scenarios. We evaluated the proposed effective ca-
pacity model against the empirical data for a number of
wireless network scenarios. With direct access to packet delay
across a wireless link being available within the simulation
environment, we determine that the effective capacity model
delivers a high degree of accuracy ranging from 0.074% to
9.09% when compared to the empirical data. We therefore
conclude that based on our analysis the effective capacity
model is usable and the assumptions on estimation of channel
service delay can yield a conservative estimates of effective
capacity under varying network conditions. We plan to develop
the effective capacity measurement methodology into a usable
tool for wireless network delay sensitive control operations,
such as admission control and routing.
ACKNOWLEDGMENT
This work has been funded by the Irish Research Council
INSPIRE programme co-funded by Marie Curie Actions under
FP7 and the COST Action IC0703 Data Traffic Monitoring
and Analysis (TMA) and supported by the OPNET Modeler
software through the OPNET Technologies, Inc. University
Program.
REFERENCES
[1] IEEE Computer Society, “Wireless lan media access control (mac)
and plysical layer (phy) specifications,” IEEE Standard for Information
technology - Telecommuncations and information exchange between
systems - Local and metrapolitan area networks - Specific requirements,
no. 802.11n, October 2009.
[2] 3GPP, “Quality of service (qos) concept and architecture (release 10),”
3rd Generation Partnership Project: Technical Specification Group
Services and System Aspects, vol. 10.0.0, no. TS 23.107, p. 41, 03 2011.
[3] D. Wu and R. Negi, “Effective capacity: a wireless link model for sup-
port of quality of service,” Wireless Communications, IEEE Transactions
on, vol. 2, no. 4, pp. 630 – 643, july 2003.
[4] OPNET, “Discrete event smulation model library,” OPNET ModelerTM ,
2011. [Online]. Available: http://www.opnet.com/
[5] F. P. Kelly, S. Zachary, and I. Zeidins, Eds., Notes on effective bandwidth,
vol. 4. Oxford University Press, 1996.
[6] G. Choudhury, D. Lucantoni, and W. Whitt, “Squeezing the most out of
atm,” Communications, IEEE Transactions on, vol. 44, no. 2, pp. 203
–217, feb 1996.
[7] A. Davy, D. Botvich, and B. Jennings, “Revenue Optimized IPTV
Admission Control using Empirical Effective Bandwidth Estimation,”
IEEE Transactions on Broadcasting, vol. 54, issue 3 part 2, pp. 599 –
611, Sept 2008.
[8] A. Davy, B. Jennings, and D. Botvich, “QoSPlan: Process for QoS-
Aware IP Network Planning Using Accounting Data and Effective Band-
width Estimation,” in Proc. of 2007 IEEE Global Telecommunications
Conference (GLOBECOM), 2007, pp. 2690–2695.
[9] A. Davy, D. Botvich, and B. Jennings, “QoSPlan: A Measurement Based
Quality of Service aware Network Planning Framework,” Journal of
Network and Systems Management, June 2012.
[10] A. Cabellos-Aparicio, F. Garcia, and J. Domingo-Pascual, “A novel
available bandwidth estimation and tracking algorithm,” in Network
Operations and Management Symposium Workshops, 2008. NOMS
Workshops 2008. IEEE, April 2008, pp. 87 –94.
[11] X. Li and D. Wu, “Power control and channel allocation for real-time
applications in cellular networks,” Wireless Communications & Mobile
Computing, vol. 8, no. 6, pp. 705–713, AUG 2008.
[12] Y. Ma, H. Zhang, D. Yuan, and H.-H. Chen, “Adaptive power allocation
with quality-of-service guarantee in cognitive radio networks,” Computer
Communications, vol. 32, no. 18, Sp. Iss. SI, pp. 1975–1982, DEC 15
2009.
[13] S. Vassilaras, “A cross-layer optimized adaptive modulation and cod-
ing scheme for transmission of streaming media over wireless links,”
Wireless Networks, vol. 16, no. 4, pp. 903–914, MAY 2010.
[14] H. Zhang, Y. Ma, D. Yuan, and H.-H. Chen, “Quality-of-Service Driven
Power and Sub-Carrier Allocation Policy for Vehicular Communication
Networks,” IEEE Journal on Selected Areas In Communications, vol. 29,
no. 1, pp. 197–206, JAN 2011.
[15] S. Akin and M. C. Gursoy, “Effective Capacity Analysis of Cognitive
Radio Channels for Quality of Service Provisioning,” IEEE Transactions
on Wireless Communications, vol. 9, no. 11, pp. 3354–3364, NOV 2010.
[16] L. Musavian and S. Aissa, “Effective Capacity of Delay-Constrained
Cognitive Radio in Nakagami Fading Channels,” IEEE Transactions on
Wireless Communications, vol. 9, no. 3, pp. 1054–1062, MAR 2010.
[17] L. Musavian, S. Aissa, and S. Lambotharan, “Effective Capacity for
Interference and Delay Constrained Cognitive Radio Relay Channels,”
IEEE Transactions on Wireless Communications, vol. 9, no. 5, pp. 1698–
1707, MAY 2010.
[18] A. Balasubramanian and S. L. Miller, “The Effective Capacity of a
Time Division Downlink Scheduling System,” IEEE Transactions on
Communications, vol. 58, no. 1, pp. 73–78, JAN 2010.
[19] A. Abdrabou and W. Zhuang, “Statistical QoS Routing for IEEE
802.11 Multihop Ad Hoc Networks,” IEEE Transactions on Wireless
Communications, vol. 8, no. 3, pp. 1542–1552, MAR 2009.
[20] B. Sklar, “Rayleigh fading channels in mobile digital communication
systems .i. characterization,” Communications Magazine, IEEE, vol. 35,
no. 7, pp. 90 –100, jul 1997.
[21] C.-S. Chang and J. Thomas, “Effective bandwidth in high-speed digital
networks,” Selected Areas in Communications, IEEE Journal on, vol. 13,
no. 6, pp. 1091 –1100, aug 1995.
[22] B. L. Mark and G. Ramamurthy, “Real-time estimation and dynamic
renegotiation of upc parameters for arbitrary traffic sources in atm
networks,” IEEE/ACM Trans. Netw., vol. 6, pp. 811–827, December
1998. [Online]. Available: http://dx.doi.org/10.1109/90.748091
[23] F. H. P. Fitzek and M. Reisslein, “MPEG-4 and H. 263 video traces
for network performance evaluation,” IEEE Network, vol. 15, no. 6, pp.
40–54, 2001.
1793