Elsarticle Template Libre

8/10/2019 Elsarticle Template Libre

1/23

Improving QoS of IPTV

and VoIP over IEEE 802.11n

Saad Saleha,, Zawar Shahb, Adeel Baiga,c

aSchool of Electrical Engineering and Computer Science (SEECS),

National University of Sciences and Technology (NUST), Islamabad, PakistanbWhitireia Community Polytechnic, Auckland, New Zealand

cCollege of Computer and Information Systems, Al Yamamah University, Kingdom of Saudi Arabia

Abstract

Tremendous growth rates of Internet Protocol Television (IPTV) and Voice over Internet Protocol (VoIP) have de-

manded the shift of paradigm from wired to wireless applications. Increased packet loss with continuously varyingwireless conditions make the transmission a challenging task in wireless environment. Our study investigates and

proposes improvement in the transmission of combined IPTV and VoIP over the IEEE 802.11n WLAN. Our major

contributions include the analytical and experimental investigations of (1) Transport layer protocol UDP/TFRC for

IPTV and VoIP, (2) Optimal physical layer parameters for IPTV and VoIP, (3) Proposition of wireless enhancement

of TFMCC (W-TFMCC) to enhance the capacity and Quality of Service (QoS) of wireless IPTV and VoIP. Analytical

and experimental evaluations show a 25% increase in capacity using TFRC with 167% more bandwidth share to TCP.

Our study shows that use of W-TFMCC with optimal parameters can enhance IPTV and VoIP capacity by 44%.

Keywords: IPTV; VoIP; DCCP; TFRC; Multi-casting; TFMCC.

1. Introduction

Internet Protocol Television (IPTV) is one of the fastest growing applications which has gained huge growth rates

in the past few years. Number of IPTV users are expected to increase by 500% from 2011 to 2016 [1]. Large growth

rate with increased users interest motivate us to study transmission of IPTV with an aim to provide better Quality of

Service (QoS). IPTV offers a number of advantages over its predecessor analog technologies. Major advantages of

IPTV include user interaction, video on demand service, economic and better Quality of Service (QoS). Architecture

of IPTV includes three entities: video head end, transport network and video receiver. Video head end is placed at

the server side and it has the tasks of video encoding and transmission of video and audio to the user end. Transport

network is the entity which plays the most crucial rule because it incorporates jitter, delay, scrambling and packet

loss effects during the transmission of video. Transport network includes both wired and wireless medium. Inside

the transport network, a number of queues having the parallel storing capabilities which shuffle the packets. Video

receiver is the last entity which has the task of decoding information, eliminating delay and jitter factors and managinga reliable QoS at the user end.

Voice over Internet Protocol (VoIP) is another fastest growing internet application which has obtained huge growth

rates in the past few years. There are 10 times more VoIP users than IPTV users[1]. Major factors for VoIP success

are cheap calling rates, better QoS and better penetration among end users. VoIP uses bi-directional traffic and has

more challenging requirements for packet loss and delay than IPTV. Transmission of VoIP requires limited packet loss

and delay for all users which becomes challenging when optimum route changes for all users.

Corresponding author

Email addresses:[email protected](Saad Saleh),[email protected](Zawar Shah),

[email protected] (Adeel Baig)

Preprint submitted to Journal of Computers and Electrical Engineering October 23, 2014


2/23

Currently, wired access links are preferred by service providers for transmission of IPTV and VoIP services owing

to minimum packet loss and delay in the wired links. Transmission of IPTV and VoIP becomes challenging in the

wireless environment because major bandwidth restriction occurs at the user end having wireless Access Point (AP)

[2]. Packets drop from queues of the wireless AP which make it difficult to meet the QoS constraints of IPTV and

VoIP. Moreover, range and data rate are also limited in wireless links owing to the continuously varying wireless

conditions. Users demand, ease of access and freedom of mobility require an insight investigation for transmission of

IPTV and VoIP over wireless networks.

IEEE 802.11n Wireless Local Area Network (WLAN) is the latest standard proposing data rates upto 600 Mbps

(theoretically) and 300 Mbps (practically). IEEE 802.11n is equipped with a number of features which include its

Multiple Input Multiple Output (MIMO) technology and frame aggregation mechanisms at Medium Access Control

(MAC) layer and Physical (PHY) layer. Frame aggregation combines multiple frames at MAC layer and PHY layer

level. Major advantage of frame aggregation includes the reduction of header over-head time and also the reduction

in the collision time. Our previous study for IPTV and VoIP capacity over IEEE 802.11n shows that aggregation of 4

packets is the optimal aggregation size for capacity enhancement of IPTV and VoIP [3]-[4].

Enhancing QoS of IPTV and VoIP with increased capacity over IEEE 802.11n has been actively discussed in

various studies due to the challenging IPTV constraints over WiFi. Packet loss is a major factor which results incapacity reduction for IPTV and VoIP. IPTV and VoIP are extremely susceptible to packet loss because both use User

Datagram Protocol (UDP) at the transport layer.

UDP is a constant bit rate protocol. By use of UDP, packets accumulate at the AP which results in congestion

at the AP. All packets crossing the limits of queue size are dropped at the AP. Our study shows that UDP provides

less delay but it increases packet loss which becomes the bottleneck for other users. Our previous study [ 3]-[4] shows

that Datagram Congestion Control Protocol (DCCP) is the better suited protocol for transmission of IPTV and VoIP.

DCCP has two variants namely TCP-like and TCP Friendly Rate Control (TFRC). TCP-like offers high reliability

and decreases its data rate much more rapidly than TFRC. This makes it suitable for all applications demanding less

packet loss. On contrary, TFRC offers a nearly constant data rate by maintaining its data rate according to varying

conditions of the network. Behaviour of TFRC makes it suitable for all applications which require less delay. Our

investigations [3]-[4] reveal through simulations that TFRC gives better performance than UDP for transmission of

IPTV and VoIP over IEEE 802.11n.In this paper, we aim to develop an analytical model for transmission of IPTV and VoIP over IEEE 802.11n.1

Transport layer protocols UDP and TFRC are modelled by their behaviour in the wireless environment. Extensive

experiments are performed to validate the analytical results. Various physical layer parameters are modelled through

SIFS, DIFS and default behaviour of wireless environment. Optimal values of queue size, contention window, SIFS

and DIFS are proposed. Analytical values of physical layer parameters are compared with experimental results.

We propose Wireless TCP Friendly Multicast Congestion Control Protocol (W-TFMCC). TFRC and TCP-like suffer

low capacity because both use the unicast mechanisms. Capacity can be enhanced significantly by shifting unicast

transmissions to multicast transmissions. In this paper we present the results of TCP Friendly Multicast Congestion

Control Protocol (TFMCC). TFMCC is designed for wired networks which suffer low packet loss and all users are

nearly in the same conditions. TFMCC keeps a track of the user facing worst packet loss conditions and adjusts its

sending rate according to the worst case user. This is highly unsuitable for the wireless medium because all users

are present in different environments. TFMCC forms channel groups based only upon users demands. Transmission

for the worst case user would lead to lower data rate even if a single user is having high packet loss rates or RoundTrip Times (RTT).We suggest a group based protocol which keeps a track of the various conditions experienced by

different users. Our study shows that performance of W-TFMCC is greater than UDP/TFRC/TFMCC if at least two

or more users are watching same channels. Performance of W-TFMCC is equal to TFRC/TFMCC when all users are

watching different channels.

1Initial results of this research appeared in

Saad Saleh, Zawar Shah, Adeel Baig, Capacity Analysis of Combined IPTV and VoIP Over IEEE 802.11n, In the IEEE Conference on

Local Computer Networks (LCN), Sydney, Australia, Oct. 2013.

Saad Saleh, Zawar Shah, Adeel Baig, IPTV Capacity Analysis using DCCP over IEEE 802.11n, In the IEEE proceedings of Vehicular

Technology Conference (VTC), Las Vegas, USA, Sep. 2013.

2


3/23

Our contributions in this work are (i) Analytical and experimental evaluation of transport layer protocols UDP/TFRC

for transmission of combined IPTV and VoIP over IEEE 802.11n, (ii) Analytical and experimental investigation of

optimum physical layer parameters of combined IPTV and VoIP over IEEE 802.11n, (iii) Proposition of a new group-

based multicast protocol W-TFMCC with performance analysis over UDP/TFRC/TFMCC through simulations and

experiments.

The rest of the paper is organized as follows. Section-2presents the related work for IPTV and VoIP over WLANs.

Section-3presents the experimental scenario and data rate estimation for IPTV and VoIP. Section-4presents the IPTV

and VoIP capacity analysis over UDP and TFRC along with fairness analysis with TCP traffic. In section-5, we show

the optimal values of IEEE 802.11n analytically and experimentally. Section-6presents the performance improvement

using multicast mechanism and our proposed W-TFMCC protocol. Section-7presents the comparison of our results

with previous state of the art approaches. Section-8concludes the paper.

2. Related Work

Transmission of IPTV over Wireless Local Area Networks (WLANs) is a challenging task because of large packet

loss, large delay and minimum available bandwidth. A number of investigations have been made for availability ofwireless IPTV. In [5], authors evaluate the capacity trends of IPTV over IEEE 802.11b and IEEE 802.11g networks.

They conclude that IPTV users and WLAN data rate have non-linear relationship with each other. Their findings show

that IEEE 802.11b and IEEE 802.11g networks can support 2 and 6 IPTV streams respectively. In [6], authors apply

the fluid model flow analysis to determine IPTV capacity. They incorporate buffer size, network hops and show a

non-linear relationship between them for reliable QoS of IPTV. In [7], an experimental investigation has been made

to evaluate the capacity of IPTV over IEEE 802.11n WLAN. Their QoS findings show that outdoor environment

deteriorates IPTV performance significantly while indoor environment can support dozens of low resolution users

with reliable QoS. In [8], Kilik and Amadou implemented a practical test bed of IPTV to observe the behaviour of

various users in IPTV channel streaming. However, their research is not focused over the improvement in various

layers of IPTV but is limited to performance of current IPTV architecture. In another study [9], Piamrat et al. study

the transmission of IPTV over the wireless home network. Authors analyze the performance of IPTV over UDP

and TCP transport layer protocols as well as various MAC layer protocols. Authors propose a solution of combinedusage of TCP with a coordinated link layer protocol. In [10], Chaparro et al. evaluate the bandwidth requirements

for transmission of high quality television content. Authors show that granularity of the estimation can be utilized

efficiently for content generator to react to changes in utilization of the network.

VoIP has been a major focus of numerous studies owing to its high demand. In [11], authors evaluate the capacity

of VoIP using various codec and packetization intervals over IEEE 802.11b network. They show through simulations

that IEEE 802.11b WLAN can support 3 to 11 users depending upon the wireless channel loss conditions. In[12],

authors evaluate the capacity of VoIP along with tracking capacity over IEEE 802.11b/g WLAN. They show through

simulations and experiments that combined VoIP and tracking capacity is 30% less than VoIP only capacity at higher

packetization intervals. In [13], authors show through experimental setup that IEEE 802.11b WLAN can support

15 calls having a packetization interval of 20 ms. They prove their experimental results through simulations and

analytical work. Transmission of combined IPTV and VoIP introduces more challenges by incorporating different

delays and packet loss thresholds for different devices. In [14], authors study the transmission of combined IPTV and

VoIP over IEEE 802.11n with varying number of hops. Authors have shown that it is possible to run 3 IPTV streamsalong with VoIP connected for 2 hops only. They prove their findings through simulations and experiments and prove

that hop count is inversely related to capacity of IPTV and VoIP. Authors conclude that performance of IPTV and VoIP

is limited in IEEE 802.11b/g networks due to less throughput. Performance is expected to enhance in IEEE 802.11n

WLAN due to provision of high data rates.

Analytical model of Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) has been developed in

[15]. Author develops a markov model to estimate the packet loss probabilities and estimates the throughput of IEEE

802.11b WLAN. Performance of IEEE 802.11n has been evaluated in [16]. Authors evaluate the MAC and PHY

layer mechanisms of VoIP and show that performance of IEEE 802.11n for VoIP is significantly enhanced. Frame

aggregation mechanism of IEEE 802.11n has been evaluated in [17]. Authors show that IEEE 802.11n can enhance

channel utilization upto 95% for UDP traffic by using frame aggregation mechanisms. They conclude that MAC

level aggregation is less effective than PHY layer aggregation. In [18], authors study the IEEE 802.11n mechanisms

3


4/23

in comparison to the legacy frame protection mechanisms. They show that IEEE 802.11n has enhanced QoS as

compared to its predecessor standards.

Performance of DCCP for delay sensitive applications has been evaluated in [12]. Authors show that DCCP

adjusts its data rate continuously which results in a decrease in packet loss at the queues of AP. They show through

simulations that DCCP gives much more fair share in bandwidth to TCP traffic than UDP. Our previous studies [3]

and [4] show through simulations that TFRC can increase network capacity for IPTV and VoIP. Performance of

DCCP is limited because it provides unicast mechanisms. A number of investigations have been made to increase

efficiency by proposing multicast mechanisms. In [19], authors propose a DCCP enhancement, called Multi(Uni)

DCCP, which transmits its streams based upon the number of receivers in the network. Authors show that their

protocol can not only provide an increase in capacity but also decrease network congestion. In [20], authors propose

an improvement to currently developed protocol TFMCC by changing the data rate equations. Authors conclude that

incorporation of the uni-directional delay improves the performance of TFMCC. In [21], authors develop a framework

for multicast transmission of multimedia services through wireless networks. Authors show that their framework

increases efficiency with better QoS as compared to TFRC.

Comparison of various studies [5]-[21] show that capacity of IPTV and VoIP is limited over WLANs due to low

data rates of IEEE 802.11b/g. High data rates of IEEE 802.11n motivate the concept of wireless IPTV and VoIP overIEEE 802.11n. To the best of our knowledge, very limited studies exist on the performance of IPTV and VoIP over

IEEE 802.11n.

2.1. Transport Layer Protocols

A number of transport layer protocols exist which vary in their performance. Before probing into the performance

of IPTV and VoIP, we define various transport layer protocols used in this study.

2.1.1. User Datagram Protocol (UDP)

UDP provides a constant data rate with no handshaking dialogues and no guarantee of service, ordering or dupli-

cate protection. UDP has no congestion control mechanism with extensive voice and video applications along with

the use in Domain Name System (DNS) and Routing Information Protocol (RIP) etc.

2.1.2. Datagram Congestion Control Protocol (DCCP)

DCCP implements reliable connection setup, congestion control, explicit congestion notification, feature negoti-

ation and tear down. Provides flow based semantics similar to TCP but does not provide reliable in-order delivery.

DCCP has two variants TCP-like and TFRC. Applications of DCCP include internet telephony, online multiplayer

games and streaming media.

2.1.3. TCP Friendly Rate Control (TFRC) Protocol

TFRC provides a congestion control mechanism for unicast flows operating in the internet and competing fairly

with the TCP traffic. TFRC varies its data rate continuously based upon the network congestion, packet loss rate and

round trip time. Internet telephony and streaming media are the popular applications of TFRC.

2.1.4. TCP Friendly Multicast Congestion Control (TFMCC) ProtocolTFMCC provides an equation-based congestion control mechanism for multicast connections by extending the

TFRC protocol from unicast to multicast domain. TFMCC is most suitable for multicast applications demanding a

smooth rate including streaming media based applications.

3. Modelling of IPTV and VoIP over IEEE 802.11n

This section presents the modelling of IPTV and VoIP over IEEE 802.11n. Experimental setup for transmission

of IPTV and VoIP is discussed. Bandwidth estimations for IPTV and VoIP are also presented for simulations and

analytical evaluations.

4


5/23

3.1. Network Scenario

We develop an analytical model and make experimental setup for performance of IPTV and VoIP over IEEE

802.11n. The analytical model assumesNnodes communicating with each other inside the wireless AP range asshown in Fig. 1. Model assumes that network nodes are lying inside the access point coverage area and all nodes

get small packet loss and RTT from AP. AP is connected to the internet which joins AP to the IPTV and VoIP

servers present on the opposite side of the internet. Matlab has been used to generate results of analytical model. All

simulations referenced from our previous study have been cited. For simulations, ns2 has been used as the simulation

tool 2 . Support of our patch [4]has been extended for simulations of IPTV and VoIP over IEEE 802.11n.

TV

TV

TV

TV

PC

TV

VoIP

VoIPVoIP

VoIP

VoIP

16'-0"*14'-0"

16'-0"*13'-0"

18'-0"*13'-0"

8'-0"*8'-0"

16'-0"*7'-0" 10'-0"*6'-0"

11'-0"*7'-0"

22'-0"*29'-0"

63'0"

40'0"

IEEE 802.11n WLAN

Laptop

TV

Figure 1: Network Scenario for IPTV and VoIP.

For experimental analysis, we setup an experimental test-bed. Distributed Internet Traffic Generator (DITG) ver-

2.8.1 is used to generate IPTV, VoIP and FTP packets from the application layer [22]. Sender and receiver devices

have DITG installed. Packet loss readings, RTT and delays are collected from the receiver devices. Data rate of

DITG is adjusted for the data rates of High Definition Television (HDTV) and Standard Definition Television (SDTV)

streams of IPTV. DITG can be tuned for DCCP and UDP streams. Results of DITG were validated with results of

previous studies [11]. IEEE 802.11n AP used for experiments has model no. AN102025 and name ADSL WirelessModem. Transmitter and receiver devices are connected to each other through the wireless AP. The experimental

setup developed is shown in Fig. 1. Various parameters, adopted from [16] used in simulations and experiments are

shown in Table-1.

3.2. Data rate estimation of IPTV and VoIP

IPTV requires a picture resolution which takes into account a number of factors including the pixel quality given

by luminance and chrominance. Luminance is the light intensity and chrominance is the colour depth. Moreover, a

moving picture is composed of a number of frames which move in series to make a moving picture. We take into

2S. McCanne and S. Floyd. ns Network Simulator. http://www.isi.edu/nsnam/ns

5
http://www.isi.edu/nsnam/nshttp://www.isi.edu/nsnam/ns


6/23

Table 1: IEEE 802.11n Access Point Parameters.

Parameter Value Parameter Value

DIFS 34s SIFS 16sSlot time 9s Physical Header 20s

Contention Window (min) 15 TXOP limit 5

Channel Bandwidth 40 MHz Bit error rate 0.000008

account the frames per second effect for IPTV bandwidth estimation. Another important factor is the size of the video

resolution. Although a number of resolutions exist but there are two standard sizes used namely SDTV and HDTV

using 16:9 or 4:3 resolutions.

Compression schemes play an important role in estimating the amount of data to be transmitted through the net-

work. Moving Picture Expert Group (MPEG) has suggested a number of compression schemes and the most popular

schemes are MPEG-2 and MPEG-4. Studies have revealed that H.264 (MPEG-4) gives a much better compression

ratio than the currently used MPEG-2 standard[23]. Data rate requirement plays the most important role in evaluating

the capacity of IPTV for deploying in a given network. We evaluate the data rate by taking into account all the factors.The various factors used in the data rate calculation are shown in Table-2.Using all these factors the data rate required

in uncompressed form is given by Eq.(1).

D = RHRVCF (1)

HereRH is the horizontal resolution and RV is the vertical resolution for the picture resolution. C is the chromi-

nance factor andFis the intensity of frames per sec used for the pictures.

The data rate obtained without any compression scheme is high (797 Mbps) which is not achievable for wireless

environment. Performance of various compression schemes specially MPEG-2 and MPEG-4 motivates the use of

compression to raw data in order to decrease the data rate. Compression ratios of two popular schemes are given as

follows[23].

MPEG-2 and H.263 Compression ratio (Hcomp) = 30:1

MPEG-4 and H.264 Compression ratio (Hcomp) = 50:1

A group of picturesNgop is generated which is arranged in a certain priority given by I,P and B frames. The datarate equation after applying compression schemes is shown in Eq. (2).

D = RHRVCF Ngop/Hcomp (2)

Table-2presents the data rates required after applying compression schemes. It is worth mentioning that our data

rate calculation is with the latest standards (resolutions, compression schemes and frames per seconds etc.) which is

in accordance with the previous researches [5][6][7][9]. We used the standard IPTV packet size of 1366 bytes which

contains 1288 bytes of payload data [7]. On contrary to IPTV, VoIP requires less data rate. Data rate of VoIP depends

Table 2: Data Rate Requirement for various compression and resolution schemes.

TV F C Resolution Compression Required

(fps) RHRV Scheme Rate (Mbps)

SDTV 24 2 640 480 MPEG-2 3.93SDTV 24 2 640 480 MPEG-4 2.36

HDTV 24 3 1920 1080 MPEG-2 26

HDTV 24 3 1920 1080 MPEG-4 15.92

upon the packetization interval, amount of payload data and the transmission schemes G.711 and G.729 etc.3 For our

study, we use the popular VoIP codec G.711 (64 kbps) with a 10 ms codec sample interval.

IPTV requires a one-way delay constraint which is 50 msec [24]. On contrary, VoIP requires two-way delay

constraints which must be less than 150 msec for both directions [12]. IPTV cannot tolerate a packet loss greater than

1% while VoIP allows a packet loss upto 2%[24][12].

3Cisco, Voice Over IP - Per Call Bandwidth Consumption, available at http://www.cisco.com

6
http://www.cisco.com/http://www.cisco.com/


7/23

4. Optimal transport layer protocol for IPTV and VoIP

In this section, we investigate the optimal transport layer protocol for transmission of combined IPTV and VoIP

over IEEE 802.11n. Only simulations were performed in our previous studies [3]-[4]. In this section, we extend ouranalysis by conducting extensive experiments and analytical evaluations. Firstly, we investigate IPTV and VoIP with

UDP and TFRC for both applications. Secondly, we study the performance of TFRC based applications in presence

of UDP based applications. Lastly, we evaluate the performance of TFRC based applications in presence of non-real

time TCP traffic.

4.1. IPTV and VoIP capacity analysis over UDP and TFRC

In this subsection, we present the analytical framework for transmission of IPTV and VoIP over IEEE 802.11n.

Our analytical model estimates the packet loss probabilities, queue utilization ratio and expected CSMA/CA wait-off

time to estimate the throughput of IPTV and VoIP over IEEE 802.11n. As layers of computer network protocols are

independent of each other, so analytical modelling of physical layer does not change by incorporation of any other

protocol at transport layer of IPTV or VoIP. Our aim is to investigate the capacity of IPTV and VoIP over IEEE 802.11n

using UDP and TFRC at transport layer.Let traffic arrival rate be defined by Ai at any time instant i and frame service rate at queue of access point be

defined bySi at time instanti. Based upon the frame service rate and frame arrival rate, queue utilization ratioQi is

given by following Eq. (3).

Qi = Ai

Si(3)

Based upon Eq. (3), average queue utilization ratio Q for a period ofT time units is given by the average of

summation of queue sizes on every instant as shown in Eq. (4).

Q=

Ti=1QiTi=11

(4)

To determine the transmission rate, we evaluate the probability of successful transmission by station i. Let pibe the probability of successful transmission andj be the transmission probability of station i. Station i transmits

successfully if no other station is transmitting. Eq. (5) shows the probability of successful transmission.

pi =

N1j=0ji

(1 j) (5)

If station i is transmitting, a collision occurs if at least one of the remaining stations transmits. Let ci be the

collision probability of stationi. Conditional collision probability is given by Eq. (6).

ci =1

N1j=0ji

(1 j) (6)

ci =1 pi (7)

IEEE 802.11n uses Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). LetWibe the expected

wait-off time for any station i with C Wj as contention window size and pi as the collision probability with Ai as

expected number of transmission attempts. Let m be the retry limit. Average wait-offtime E[Wi] and average trans-

mission attemptsE[Ai] are given by Eq. (8)and Eq. (9), respectively[15]. It is worth mentioning thatCWjrepresents

the size of the contention window while p irepresents the collision probability.

E[Wi]=

m1k=0

pik(1 pi)

kj=0

CWj

2 + pi

m

mj=0

CWj

2 (8)

7


8/23

E[Ai]=

m1

k=0

pik(1 pi)(k+ 1) +pi

m(m + 1) (9)

Based on the expected wait offtime and expected number of transmission attempts, transmission probability Tiis

given by Eq. (10).

Ti = E[Ai]

E[Wi] +E[Ai] (10)

IEEE 802.11n transmits frame on the physical layer level which can be modelled by incorporating SIFS, DIFS,

frame transmission time and acknowledgement time. Let Ts be the time taken by a frame if it is transmitted success-

fully. LetTc be the time if a frame is not transmitted successfully. Ts andTc can be modelled by viewing the frame

transmission timeTf rame,S I F S ,Tackand DIFSas given in Eq.(11) and Eq. (12).

Ts =Tf rame+ S I F S + Tack+ DI FS (11)

Tc =Tf rame+ Tacktimeout (12)

Also, collision time is a function of collision probability to success probability. Based on [15], average collision

timeTavg(c) can be derived from the packet loss probability p iand packet collision timeTc as shown in Eq. (13).

Tavg(c)= pi

1 piTc (13)

Eq. (3)-Eq. (13)are used to evaluate the throughput using collision time and transmission probability. Analytical

results for capacity of IPTV and VoIP using UDP are shown in Table-3.Results show that IPTV and VoIP users have

an inverse relationship with each other. This behaviour shows that IPTV users occupy majority share in bandwidth

which reduces share for VoIP users. Network congestion also increases with large number of IPTV users which

increases delay for further VoIP users. Similar trends are observed experimentally for IPTV and VoIP users. Resultsfrom Table-3show that 4 IPTV users can be accommodated with 1 VoIP user maximally. All of our results represent

the steady state conditions where throughput of all supportable IPTV and VoIP streams is at saturation level. It is

important to mention that all of our simulation, experimental and analytical results differ by a small amount based

upon the following reasons.

Analytical results show maximum capacity because all analytical results present a mean estimate of capacity of

IPTV and VoIP users based upon their bandwidth requirements in wireless medium.

Simulations present an estimate of users obtained by simulating the environment in ns2. Collisions at a partic-

ular instant decrease the capacity in simulations.

Experimental results show minimum capacity because of the particular environment e.g. room walls, obstacles,

transmitting device location and receiving device location and reception system etc deteriorate the capacity

from the ideal analytical capacity.

Table 3: IPTV and VoIP capacity over IEEE 802.11n using UDP.

Simulation Analytical Experimental

IPTV VoIP IPTV VoIP IPTV VoIP

1 36 1 39 1 32

2 24 2 28 2 21

3 11 3 14 3 8

4 2 4 6 4 1

8


9/23

IPTV using UDP encounters packet loss at the AP because UDP is a constant bit rate protocol and packets drop from

queues of AP. UDP cannot cope fairly with the network congestion and maintains its data rate at any situation. To

resolve congestion less mechanism of UDP, DCCP has been proposed. DCCP has two variants namely TCP-like and

TFRC.

TCP-like uses the congestion control mechanism similar to TCP. TCP-like has no retransmissions. TCP-like

applies congestion control to acknowledgements, it works on units of packets instead of bytes and TCP-like uses no

retransmissions. These features make it different from TCP but in reliability it resembles TCP because it decreases its

data rate much more sharply in congested situations similar to TCP.

We have shown previously that TCP-like gives poor performance for IPTV and VoIP because TCP-like decreases

its data rate in congested situations. Decrease in data rate increases the delay for all packets reaching the receiver end.

IPTV is a real time service with limitations of packet loss and delay. Large delay disrupts the support for many users.

On contrary to TCP-like, TFRC is the protocol which resembles UDP in some aspects like no retransmissions

and less reliability etc. Instead of constant bit rate, TFRC adjusts its data rate continuously according to the packet

loss rate and RTT experienced at the user end. Any increase in packet loss or RTT implies network congestion which

results in throughput reduction of TFRC. TFRC increases its data rate in low packet loss and RTT conditions.

Lets, p,R and tRTO be the segment size, packet loss rate, round trip time and TCP retransmission timeout value,respectively. Let b be the packets acknowledged by a single TCP acknowledgement. Using above parameters, through-

put of TFRC is modelled by the Eq. (14). It is pertinent to mention that we define throughput as the physical layer

data rate sent by the transmitter on the physical link. Throughout this paper, we add the various standard header fields

into the transport layer header to estimate the physical layer throughput.

X= s

R

2bp

3 + tRTO 3

3bp

8 p(1 + 32p2)

(14)

Analytical results of throughput for a range of packet loss probabilities and RTT are shown in Fig. 2. RTT trends

show that RTT is inversely related to the data rate. Packet loss trends show that packet loss is inversely related to the

throughput governed by TFRC. Moreover, trends of TFRC are more drastic for change in packet loss than change in

RTT.

0.04 0.045 0.05 0.055

5

5.5

6

6.5

7

7.5

x 107

Round Trip Time (RTT) (ms)

Throughput(bps)

0.5 1 1.5 2

x 104

2

3

4

5

6

7

8

x 107

Packet Loss Probability (p)

Throughput(bps)

p=0.000021

p=0.000023

p=0.000019

RTT=41ms

RTT=45ms

RTT=49ms

Figure 2: TFRC round trip time and packet loss probability versus throughput

Table-4shows the capacity of IPTV and VoIP over IEEE 802.11n using TFRC for both applications. Results show

that TFRC can accommodate more IPTV users than UDP. TFRC adjusts its data rate by decreasing its data rate in

congested situations. Coping with the network situations makes TFRC a better suitable candidate for transmission of

IPTV and VoIP with increased capacity. Analytical and experimental results show that IPTV and VoIP with TFRC

can accommodate at least 5 IPTV users with 0 VoIP users. This suggests that TFRC provides 1 more HDTV user than

UDP. Comparison of Table-3and Table-4shows that UDP and TFRC provide same VoIP capacity for small number

of IPTV users. On contrary, large number of IPTV users can be accommodated with the use of TFRC. This suggests

that TFRC is more suitable for IPTV than VoIP. VoIP has small packet size with different packetization interval than

9


10/23

IPTV. This suggests that VoIP suits well to UDP than TFRC. In next subsection, we aim to figure out IPTV and VoIP

running different protocols either UDP or TFRC.

Table 4: IPTV and VoIP capacity over IEEE 802.11n using TFRC.



1 35 1 39 1 32

2 22 2 26 2 20

3 11 3 17 3 8

4 2 4 6 4 1

5 1 5 4 5 0

4.2. Cross-protocol performance of IPTV and VoIP using UDP and TFRC

In this section, we analyze the performance of IPTV and VoIP over the protocols UDP and TFRC. Based on thecurrent network architecture, both IPTV and VoIP use UDP at the transport layer. Aim of this section is to investigate:

How about changing only IPTV or VoIP transport layer protocol without influencing other network traffic?

We evaluate the performance of TFRC based IPTV in presence of UDP based VoIP. UDP is a constant bit rate

protocol. Available capacity in network decreases with UDP because TFRC decreases its data rate based on the

increase in packet loss rate and RTT. On contrary, UDP keeps sending packets with constant bit rate. Our analysis

shows that data rate of TFRC is highly dependent upon the bit rate of UDP. To increase the TFRC based IPTV

connections, UDP connections must be reduced. A comparison of analytical and experimental performance versus

simulation results is shown in Table-5.

Table 5: IPTV(TFRC) and VoIP(UDP) capacity over IEEE 802.11n using TFRC.


IPTV VoIP IPTV VoIP IPTV VoIP1 7 1 9 1 5

2 5 2 6 2 4

3 3 3 4 3 1

4 0 4 2 4 0

Similarly, we evaluate the performance of UDP based IPTV with TFRC based VoIP. Table-6shows the analytical

and experimental results versus simulation results. Results show that TFRC follows same trends in presence of UDP.

Number of UDP connections must be reduced in order to avoid network congestion. Results show UDP and TFRC

cannot co-exist fairly with each other because UDP occupies all the network bandwidth irrespective of the packet loss

and delay encountered by the system.

Table 6: IPTV(UDP) and VoIP(TFRC) capacity over IEEE 802.11n



1 8 1 10 1 6

2 6 2 7 2 4

3 4 3 5 3 3

4 1 4 2 4 0

10


11/23

4.3. Fairness Analysis of IPTV and VoIP with TCP traffic

Studies show that 80% of network traffic is non-real time traffic[12]. This suggests that fairness of IPTV and VoIP

traffic must be studied with Transmission Control Protocol (TCP) traffic in the network. To implement TCP traffic,we use a packet size of 1000 bytes with File Transfer Protocol (FTP) at application layer. TCP is used at the transport

layer of FTP which makes FTP a suitable candidate to test non-real time reliable traffic. Multiple TCP connections

are established because probability of achieving minimum window size is highest with a single TCP source.

Capacity and fairness results from analytical and experimental evaluations of IPTV and VoIP traffic using UDP

with FTP traffic are shown in Table-7. Results show that TCP decreases its window size in presence of constant bit

rate protocol UDP. This suggests that UDP provides not only less capacity but also provides unfair share in bandwidth

to TCP traffic in the network.

We test the performance of IPTV and VoIP using TFRC with FTP traffic in the network as shown in Table-7.

Results show that TCP competes fairly with TFRC traffic in the network. Bandwidth analysis shows that TFRC

provides more share in bandwidth to TCP traffic than UDP. TCP gets 4.3 Mbps throughput in presence of UDP while

11.5 Mbps throughput in presence of TFRC experimentally. Comparable performance gains with similar trends are

observed for both SDTV and HDTV. TFRC provides at least 167.4% more throughput to TCP than UDP.

Our study over various transport layer protocols suggests that TFRC provides better capacity than UDP for trans-mission of IPTV and VoIP over IEEE 802.11n. TFRC adjusts its data rate according to the network conditions which

makes it a better candidate for IPTV and VoIP. Our investigation suggests that TFRC must be adopted for all real time

applications. Performance of TFRC deteriorates severely in presence of UDP due to congestion less mechanism of

UDP. TFRC provides much more fair share in bandwidth to TCP than UDP. In the next section, we aim to investigate

the optimum physical layer parameters for combined IPTV and VoIP over IEEE 802.11n.

Table 7: PERFORMANCE STATISTICS FOR COMBINED IPTV AND VoIP ALONG WITH TCP TRAFFIC

IPTV Combined Average Throughput (Mbps)

Flows Simulation Analytical Experimental

HDTV UDP- UDP: 50.7 UDP: 51 UDP: 47.2

TCP TCP: 6.8 TCP: 7 TCP: 4.3

HDTV TFRC- TFRC: 66.5 TFRC: 68.2 TFRC:62.3TCP TCP: 14.2 TCP: 16.3 TCP: 11.5

SDTV UDP- UDP: 51.3 UDP: 52.9 UDP: 48.4

TCP TCP: 6.1 TCP: 8.4 TCP: 4.2

SDTV TFRC- TFRC:64.7 TFRC: 67.3 TFRC:61.2

TCP TCP: 14.4 TCP: 16.2 TCP: 12.3

5. Optimal Physical layer parameters of IEEE 802.11n for IPTV and VoIP

IEEE 802.11n is equipped with a number of parameters and enhanced features including queue size, SIFS, DIFS,

contention window size and physical layer header time. Simulation results of our previous investigations [3]-[4] for

transmission of IPTV and VoIP are shown in Table-8. In this section we present the analytical and experimentalevaluation of all the parameters.

Table 8: Parameters for IEEE 802.11n

Parameters Default Parameters Proposed Parameters [4]

Queue Size 50 pkts 70 pkts

SIFS 16s 14.4s

DIFS 34s 30.6s

Physical Header 20s 18s

Contention Window 15 11

11


12/23

5.1. Trends of SIFS, DIFS and Physical header

Layers of computer network protocols are independent of each other. So, all layers can be modelled independently.

Parameters like SIFS, DIFS, contention window and physical header time belong to the physical layer. It means thatany change in physical layer parameters needs to be modelled at only physical layer level.

Let TS I F S be the time duration required to wait for SIFS time, TDIF Sbe the time duration required to wait for

DIFS time,TPHYbe the time duration required for physical header transmission and Tpayloadbe the time required to

transmit payload data. Various time periods as observed at the physical layer are shown in Table-1. Throughput at

physical layer given byXcan be modelled by Eq. (15) assuming RTS/CTS is disabled. Eq. (16)shows the throughput

equation if RTS/CTS is enabled.

X= s

TS I F S + TDIF S+ Tpayload+ Theader+ Tack(15)

X= s

3TS I F S + TDIF S+ Tpayload+ Theader+ Tack(16)

Using default values of all time periods, various parameters are varied in Eq. (15) and Eq. (16). Fig. 3shows theresults of simulations and analytical observations by varying various parameters. Results show that SIFS and DIFS

are inversely proportional to the throughput. Moreover, SIFS has more drastic effect on throughput than DIFS because

SIFS is encountered more than DIFS. SIFS and DIFS have no optimal values but DIFS must have the value given by

Eq. (17). TProptime is the propagation of a packet from AP to user.

TDIF S =2TS I F S + TProptime (17)

0 1 2 3

x 105

0.5

1

1.5

2

x 108

SIFS (s)

Throughput(bps)

SIFS behaviour

2 3 4 5

x 105

4

5

6

7

8

9x 10

7 DIFS behaviour

DIFS (s)

Throughput(bps)

RTS/CTS disabled

RTS/CTS enabled

RTS/CTS enabled

RTS/CTS disabled

Figure 3: Trends SIFS and DIFS with RTS/CTS enabled and disabled.

If DIFS is less than SIFS then throughput would be zero effectively because all stations would transmit when any

station is waiting for SIFS. This results in wastage of capacity in form of collisions. Trends of physical header timewith RTS/CTS enabled and disabled are shown in Fig. 4. Decrease in physical header time provides more throughput

because remaining time is used for payload transmission. We suggest a decrease of only 10% which changes physical

header time to 18s. Large change in physical header is not possible due to small clocking frequency of devices.

Results show that decrease in time durations of physical layer parameters gives more time for data transmission.

We suggest a decrease of only 10% in physical header duration which can increase capacity by at least 1 VoIP user.

Our previous simulations [4]also proposed 10% decrease in parameters through simulations only. Large decrease in

physical layer parameters is not possible due to limitation of physical devices.

5.2. Trends of Contention Window

Contention window represents the physical layer waiting time encountered in CSMA/CA which tries to avoid

collisions and also tries its best to minimize the redundant waiting time. Our previous study [4] has shown through

12


13/23

1 1.5 2 2.5 3 3.5

x 105

6

7

8

9

10

11x 10

7

Physical header time (s)

Throughput(bps)

RTS/CTS enables

RTS/CTS disabled

Figure 4: Trends Physical header time with RTS/CTS enabled and disabled.

simulations that contention window size of 11 is the optimal size instead of 15. Large packet size is the major cause

for small contention window size. IPTV has got a large data rate requirement but the packet size is 1366 bytes. Large

data rate limits the support for large number of IPTV users. This suggests that only few IPTV users are competing

with each other at any particular instant. Few competing users can accommodate in a situation of small window size

as compared to large window size. Large contention window size results in redundant waiting times of all users which

decreases throughput and capacity.

Considering the CSMA/CA backoffmechanism, various transmission probabilities have been shown in [15]. An-

alytical results for IPTV and VoIP from CSMA/CA equations are shown in Fig. 5. Trends show that throughput

increases upto a contention window size of 11. This behaviour suggests that small contention window size results

in collision of packets between various users. Beyond contention window size of 11, throughput decreases. This

suggests that large contention window size results in redundant waiting time of stations which decreases throughput

slightly. Our analytical and simulation results confirm that contention window size of 11 is the optimal size. Results

suggest that only optimal contention window size should be used for practical applications running IPTV and VoIP.

Moreover, for large scenarios, contention window size can be estimated practically by computing the throughput ob-tained by all devices at varying contention window sizes. Contention window size providing maximum throughput

is the optimal size for that particular configuration of various devices. It is pertinent to mention that slight tuning of

contention window is required for applications running TCP traffic simultaneously with IPTV and VoIP traffic based

upon the number of competing devices.

0 5 10 150

10

20

30

40

50

60

70

Contention Window (CW)

Throughput(Mbps)

Analytical

Simulation

Figure 5: Contention window trends for IPTV and VoIP.

13


14/23

5.3. Trends of Queue Size

Queue size represents the physical layer buffering capacity for the wireless AP. Queue size for IPTV and VoIP

over IEEE 802.11n can be modelled through queue utilization ratio. Let traffic arrival rate be Aiand departure rate beDi, then queue utilization ratio is given by eq (18).

= ArrivalRate

DepartureRate=

Ai

Di(18)

Actual queue size for any system is a function of the queue utilization ratio. Queue utilization ratio of less than 1

indicates small arrival rate as compared to serving rate. On contrary, queue utilization ratio greater than 1 indicates

less serving rate than the utilization rate. Queue size, given by B, can be represented as a function of queue utilization

ratio by Eq. (19).

P(QueueS ize= B) = 1

1 B1B (19)

Eq. (19) can be simplified arithmetically into Eq. (20).

B = 1

1 + P(20)

Queue size depends upon the traffic arrival rate. This suggests that it needs to be modelled with different traffic

categories. Traffic arrival rate of TCP and TCP-like resembles Additive Increase and Multiplicative Decrease (AIMD)

model. A simplified model for estimating TCP rate is given by Eq. (21).

R= 1

RT T

3

2P(21)

Traffic arrival rate of TFRC is modelled by the TFRC rate equation. Eq. (22)shows the rate equation of TFRC.

X=

s

R

2bp

3 + tRTO 3

3bp

8 p(1 + 32p2)

(22)

Our analysis shows that the bottleneck link in IPTV transmission is the wireless AP. TFRC increases its data rate

until it gets some packet loss or increased RTT from the AP. On contrary to TFRC, UDP is a congestion-less protocol

which provides constant bit rate. Fixed data rate of UDP provides a fixed arrival rate at the queues of AP and queue

size depends upon the transmission rate from queues of AP to the wireless user.

To model the queue size of UDP, we use a constant bit rate model having uniform distribution. IEEE 802.11n

provides a theoretical data rate of 600 Mbps while a practical data rate of 300 Mbps at physical layer level using 33

MIMO technology.

Fig. 6presents the analytical and experimental results for optimal queue size of IPTV and VoIP for transmission

over IEEE 802.11n. Results show that initially queue size increases throughput sharply. After 100 packets, increase in

throughput is nearly negligible. Results show that maximum capacity is obtained at a queue size of 70 packets. Very

large queue size results only in wastage of resources because wireless channel becomes the bottleneck link for largequeue size. Experimental results show less throughput than analytical because of varying network conditions and large

packet loss in a real network scenario. Delay is maximum for experimental results because of the wireless conditions.

Analytical results display minimum delay because they incorporate limited network conditions. Simulation results

display delay and throughput in between experimental and analytical.

5.4. Trends of Aggregation

IEEE 802.11n is equipped with the aggregation mechanisms which use aggregation at two levels. At level 1,

multiple MAC layer frames called MAC Service Data Units (MSDUs) are aggregated together to form an Aggregated

MAC Service Data Unit (A-MSDU). At level 2, multiple physical layer frames composed of physical layer header

and A-MSDUs are aggregated together to transmit an Aggregated MAC Physical Data Unit (A-MPDU).

14


15/23

0 200 40010

20

30

40

50

60

70

Queue Size (Pkts)

Throughput(Mbps)

Analytical

Simulation

Experimental

0 200 40038

40

42

44

46

48

50

52

Queue Size (Pkts)

Delay(msec)

Analytical

Simulation

Experimental

Figure 6: Queue size trends for IPTV and VoIP.

LetTMPDUbe the transmission time of a single MPDU and Nbe the number of MPDUs inside a single A-MPDU.

LetTS I F S andTDIF Sbe the time required to wait for SIFS and DIFS before transmission. Then transmission time T

of an MPDU (without aggregation) is given by Eq. (23). Eq. (24) represents the transmission time TAMPDUwith an

aggregation ofNMPDUs inside a single A-MPDU.

T=TDIF S+ TMPDU+ TS I F S + TACK (23)

TAMPDU=TDIF S+ NTMPDU+ TS I F S + TACK (24)

Fig. 7shows the aggregation performance of IPTV and VoIP over IEEE 802.11n. Results show that capacity in-

creases upto 4 times aggregation due to reduction in collision time and header overhead. Beyond 4-times aggregation,

large delay disrupts the support for large number of users.

2 4 6 8

20

25

30

35

40

45

50

55

60

65

70

Aggregation (Pkts)

Throughput(Mbps)

Analytical

Simulation

Experimental

2 4 6 8

40

45

50

55

60

65

70

75

80

85

90

Aggregation (Pkts)

Delay(msec)

Analytical

Simulation

Experimental

Figure 7: Aggregation trends for IPTV and VoIP.

Comparison of physical layer parameters for IPTV and VoIP over IEEE 802.11n shows that maximum capacity

can be achieved at optimal values of all the parameters. Table-9presents the comparison of optimal parameters of

IPTV and VoIP over IEEE 802.11n. As layers are independent so changes in physical layer parameters do not affect

transport layer protocol. Optimal parameters remain same with the use of UDP or TFRC. SIFS, DIFS and physical

15


16/23

header have no optimal values because decrease in physical layer parameters gives more time for payload data to

transmit. This shows that SIFS, DIFS and physical header duration must be reduced as much as possible. We suggest

a decrease of 10% in these parameters which can increase capacity by at least 1 VoIP user without effecting current

network.

Table 9: Optimal Parameters of IEEE 802.11n

Parameters Standard Values Simulations [4] Analytical Experimental

Queue Size 50 pkts 70 pkts 75 pkts 80 pkts

Aggregation 1 pkt 4 pkts 4 pkts 4 pkts

Contention Window 15 11 11 11

6. Improving performance through multicast mechanism

In this section, we aim to analyse and improve the performance of combined IPTV and VoIP over IEEE 802.11n.In this regard, we present the limitations of current IPTV architecture using TFRC. We analyse the performance of

IPTV in presence of VoIP using multicast TFMCC protocol. Finally, we present our proposed protocol W-TFMCC

by mitigating the limitations of all previous protocols. We present the performance of W-TFMCC with a test run real

network scenario and explain its performance gains.

6.1. Using Multicast for IPTV transmission

Studies reveal that capacity of IPTV is dependent upon the channel popularity [25]. Zipfs law states that the

channel viewership ofxth channel would be x-times less than the first channel [25]. Fig. 8shows the histograms of

channel viewership versus channel popularity.

2 4 6 8 10 12 14 16 18 200

0.05

0.1

0.15

0.2

0.25

Channels Popularity

Viewership

Figure 8: Channel viewership with popularity (Zipfs Law)

Trends of histogram show that channel viewership decreases exponentially as its popularity decreases. So, unicast

mechanism for transmission of IPTV results in wastage of capacity. Our results for IPTV capacity using UDP and

TFRC show that multiple receptions of same channel require multiple transmissions. This suggests that multicast

transmissions must be used for all users viewing same channel. TFRC and TCP-like employ unicast mechanism to

observe the packet loss and RTT of the receiver. Our study shows that congestion control mechanism of TFRC can

enhance capacity of IPTV. For multicast mechanisms, TCP Friendly Multicast Congestion Control Protocol (TFMCC)

has been designed which has congestion control mechanisms similar to TFRC. TFMCC selects the worst case receiver

based on the packet loss rate and makes all transmissions based on the worst case receiver. It was believed in TFMCC

16


17/23

that all receivers would get the reception if the worst case receiver gets the reception. Our analysis shows that per-

formance of TFMCC is limited in wireless medium. In wireless medium, TFMCC selects the worst case receiver and

adjusts its data rate according to the packet loss rate and RTT experienced by the worst case receiver. This suggests

that any user experiencing worst-case receiver conditions deteriorates the performance of all other users viewing same

channel. Fig. 9presents the scenario in which few IPTV users lie inside first range of AP having small packet losses

while other users lying outside first range experiencing large packet loss conditions.

Figure 9: Network Scenario for IPTV and VoIP.

In this paper, we propose wireless enhancement of TCP Friendly Multicast Congestion Control Protocol (W-

TFMCC). Basic idea of W-TFMCC is to limit the capacity of only those users experiencing worst case conditions.

Throughput of all users lying in better environment must not be deteriorated. W-TFMCC estimates the packet loss

rate of all users lying in its range and makes an estimate of all users which can be provided with a reliable QoS either

HDTV streams or SDTV streams. We show that capacity of W-TFMCC provides 61% coverage area in comparison to

TFMCC and TFRC providing only 30% and 5% coverage areas respectively. Moreover, W-TFMCC results not only

in an increase in capacity but also saves resources by avoiding any redundant transmissions of channels which makes

it more efficient.

6.2. TFMCC for IPTV transmission

TFMCC is a single rate multicast congestion control protocol. It competes fairly with TCP traffic in the network

because it uses congestion control mechanism similar to TFRC. TFMCC uses the packet loss rate of the receivers

to determine the sending rate continuously. TFMCC protocol uses an equation to determine the sending rate contin-uously. Let p be the packet loss rate and s be the segment size, then TFMCC throughput X is given by Eq. (25).

X= 8s

R

2p

3 + 12

3p

8 p(1 + 32p2)

(25)

Simulations for TFMCC are performed over ns2 by extending support of [4] for IPTV and VoIP over IEEE

802.11n. Performance of TFMCC has been tested in varying environments as shown in Table-10. Different regions

have different range of packet loss and RTT. Our results for performance analysis of TFMCC over IEEE 802.11n

for transmission of IPTV and VoIP are shown in Table-11. Coverage area shows the percentage of supported users

according to Zipfs law. Results show that TFMCC gives a high coverage area 61.65% in Terrain-A with ideal con-

ditions. However, coverage area reduces to 30% in worst conditions which suggests that only 30 users are supported

17


18/23

out of 100 (approximately for a certain bandwidth). Performance of TFMCC is worst in varying network conditions.

TFMCC gives increased performance over IEEE 802.11n in reliable network conditions where all users have limited

packet loss and RTT. Performance of TFMCC deteriorates in large packet loss and RTT conditions. Different users

lying in different conditions demanding same channel get worst QoS because TFMCC selects worst case receiver.

This necessitates the demand of W-TFMCC to provide better QoS to all users.

Table 10: Terrains for IPTV and VoIP over IEEE 802.11n.

Terrain Environment

Type-A HDTV Support : Low Packet Loss & Low RTT

Type-B HDTV Support : Low Packet Loss & Large RTT

Type-C SDTV Support : Large Packet Loss & Low RTT

Type-D SDTV Support : Large Packet Loss & Large RTT

Type-E SDTV Support : Very Large Packet Loss & RTT

Table 11: IPTV and VoIP capacity over IEEE 802.11n using TFRC and TFMCC respectively.

Terrain TFRC TFMCC Coverage

Type Simulation Simulation Analytical Experimental %

TV VoIP TV VoIP TV VoIP TV VoIP

A 5 1 1 35 1 39 1 32 61.65

B 4 15 1 33 1 37 1 29 54

C 4 8 1 25 1 29 1 23 49

D 4 3 1 14 1 19 1 12 44

E 3 5 1 6 1 9 1 4 30

6.3. Wireless Enhancement of TFMCC (W-TFMCC)

Our proposed protocol W-TFMCC is an enhancement of current protocol TFMCC. Basic limitation in current

mechanism arises from the varying channel conditions in wireless environment.

TFMCC transmits all streams using a single multicast transmission depending upon the worst case user reception.

Transmissions to worst case user deteriorate the data rate and effectively throughput is worstly effected for all users.

On contrary, W-TFMCC selects the worst case user depending upon the packet loss conditions and RTT experienced

by worst case user. If HDTV throughput is not possible for any user then W-TFMCC transmits a separate channel

group for users having large packet loss.

Based upon our previous investigations on TFRC, we prefer the same data rate equation of TFRC to be used for

W-TFMCC, as shown in Eq. (26).

X= 8s

R2p3 + 123p

8 p(1 + 32p2)

(26)

If packet loss or RTT or both parameters increase beyond a limit then W-TFMCC separates those users from

current multicast transmission. Fig. 10presents the scenarios of wireless environment with varying loss and RTT

conditions. Limits of packet loss and RTT for provision of HDTV and SDTV streams with reliable QoS are marked

in the figure.

Our simulation and analytical results for W-TFMCC performance in comparison to TFRC are shown in Table-12.

Results show that performance of TFRC deteriorates in worst conditions because supported users are limited. On

contrary, W-TFMCC transmits streams to group of users. For ideal channel conditions, like Terrain-A, W-TFMCC

transmits only 1 stream instead of 5 for 5 users watching same channel. To estimate the performance of W-TFMCC,

we use the users distribution as given by Zipf s law. Considering a set of 20 channels, probability of watching first five

channels is the sum of individual channels probabilities. To support 5 users watching same channel, TFRC transmits

5 streams. On contrary W-TFMCC transmits only 1 stream for 5 users watching same channel. Using Zipfs law,

18


19/23

0 0.002 0.004 0.006 0.008 0.010

0.05

0.1

0.15

0.2

0.25

Packet Loss Probability

RoundTripTime(s)

HDTV

SDTV

Figure 10: Packet loss and RTT limits for SDTV and HDTV channels.

coverage area of W-TFMCC appear to be 61.65%. This suggests that 61 IPTV users can be supported out of 100 users

which is a remarkable gain. In worst terrains, coverage area decreases to 56.25% users because SDTV and HDTV

streams have to be transmitted simultaneously to support different group of users watching same channel. Results

show that W-TFMCC transforms the concept of capacity from number of users to number of channel streams. In case

of excess packet loss and RTT, W-TFMCC stops transmissions until it finds the ideal situation for SDTV or HDTV

transmissions. Capacity of W-TFMCC is always greater or equal to TFRC and TFMCC. Worst case of W-TFMCC

can occur if all users are present in different conditions and viewing different channels. In such a situation, W-TFMCC

would converge to unicast mechanism. This suggests that W-TFMCC can perform better in all environments. Fig.

11 presents the coverage comparison for 100 users using UDP/TFRC/TFMCC/W-TFMCC. Results show that W-

TFMCC performs best in worst terrains. Large number of users can be supported easily using W-TFMCC. Advantagegain increases with more users in comparison to TFRC.

Table 12: IPTV and VoIP capacity over IEEE 802.11n using TFRC and W-TFMCC respectively.

Terrain TFRC W-TFMCC Coverage

Type Simulation Simulation Analytical Experimental %

TV VoIP TV VoIP TV VoIP TV VoIP

A 5 1 1 35 1 39 1 32 61.65

B 4 15 2 30 2 34 2 26 61.65

C 4 8 2 22 2 26 2 20 61.65

D 4 3 2 12 2 18 2 15 58.7

E 3 5 2 3 2 7 2 6 56.25

Algorithm-1presents the pseudocode for performance of W-TFMCC. Xrepresents the throughput which is esti-

mated using the RTT and packet loss probability. XhdandXsddenote the reference throughputs required to transmit

HDTV and SDTV streams respectively as given by Table-2. Repeated measurements are performed after every finite

amount of time. Sampling time must be as small as possible to detect the continuously varying wireless conditions.

Users groups are formed based upon RTT and packet loss probability.

6.4. Test Run for W-TFMCC

Performance of W-TFMCC can be analysed with a test run having challenging wireless conditions. It is pertinent

to mention that Fig.12presents a real home network scenario. Wireless AP is located in one room depending upon the

available sockets and users convenience. There are four regions depending upon the signal strength in various regions.

19


20/23

0

10

20

30

40

50

60

70

A B C D E

Coverage(%)

Terrain Type

UDP

TFRC

TFMCC

W-TFMCC

Figure 11: Coverage comparison of UDP/TFRC/TFMCC/W-TFMCC

Algorithm 1:W-TFMCC Protocol

1 pusr(i) ; /* Packet loss probability of user i */2 RT T(i) ; /* RTT of user i */

3 HDch=[ ] ; /* Transmitted HDTV channels list */

4 S Dch=[ ] ; /* Transmitted SDTV channels list */

5 HD(i) ; /* HDTV channel for user i */

6 S D(i) ; /* SDTV channel for user i */

7 X(i)= s

R

2bp

3 +tRTO 3

3bp

8 p(1+32p2)

;

8 ifX(i)> Xhdthen

9 HDch = [HDch H D(i)] ; /* Start HDTV transmission for i */

10 else ifX(i)> Xsdthen

11 S Dch = [S Dch S D(i)] ; /* Start SD transmission for i */

12 else

13 HDch = H Dch - pusr(i) ; /* Remove user from channel list */14 S Dch = S Dch - pusr(i);

15 end

Various regions have been marked with different patterns. We analyse the performance of all protocols in the given

test run environment. Depending upon the packet loss and RTT, our experimental receiver device splits coverage into

four different regions from Level-1 to Level-4. We consider users demand in each region such that Level-1 region

demands an HDTV and a FTP device (computer). There are three regions (rooms) having Level-2 coverage. Level-2

demands two HDTV, three VoIP phones and an FTP device (laptop). Level-3 region demands two HDTV and one

VoIP phone. Level-4 coverage area has a single HDTV demand.

UDP and TFRC use unicast transmission mechanisms. Demand of each user is fulfilled by individual and unique

transmission with UDP/TFRC. Our investigation reveals that 4 IPTV users can work using UDP while 5 HDTVdevices can work with TFRC. Unicast transmissions of UDP/TFRC are independent of each other. All users are

supplied with individual streams even if all users are viewing same channel. This suggests that UDP and TFRC waste

significant resources because of unicast mechanism.

TFMCC uses the multicast mechanism by supplying single channel stream to all users viewing same channel.

Major limitation of TFMCC arises when user in level-4 and level-1 are viewing same channel. TFMCC selects the

level-4 user because it is in worst condition as compared to user in level-1 region. All transmissions are performed

according to level-4 user. As a result, users lying in level-1 suffer poor image quality because TFMCC adjusts its

data rate according to level-4 user. Worst conditions of level-4 user deteriorate the performance of level-1 user who is

unable to watch HDTV stream.

Our proposed protocol W-TFMCC gives best performance in this environment. Packet loss and RTT of users

lying in level-1 and level-4 regions watching same channel is identified. If packet loss and RTT of level-4 user is

20


21/23

greater than the possible limit for HDTV stream then level-4 user is supplied with an SDTV stream. Enhanced group-

based mechanism has sorted the problem of level-1 user by supplying it with an HDTV stream. Channels of all users

lying in different regions are identified with W-TFMCC. Packet loss and RTT of all users watching same channel are

identified. Various channel streams are supplied to different groups based upon their environment conditions. Worst

case of W-TFMCC is the scenario in which all users are lying in different conditions and watching different channels.

In such a case, performance of W-TFMCC would be similar to TFRC because it would use unicast streams for all

viewers.

TV

TV

TV

TV

PC

TV

VoIP

VoIP

VoIP

VoIP

16'-0"*14'-0"

16'-0"*13'-0"

18'-0"*13'-0"

16'-0"*7'-0" 10'-0"*6'-0"

11'-0"*7'-0"

22'-0"*29'-0"

63'0"

40'0"

WLAN

COVERAGE AREAS

Level-1

Coverage

(Maximum)

Level-2

Coverage

Level-3

Coverage

Level-4

Coverage

(Minimum)

Laptop

TV

TV

Figure 12: Test Run scenario for IPTV and VoIP using W-TFMCC.

6.5. Comparison of TFMCC vs W-TFMCC

Comparison of TFMCC with W-TFMCC shows that our proposed W-TFMCC protocol enhances capacity in

tough terrains. Among the five tested terrain types (A, B, C, D, E), which varied in packet loss and round trip time

conditions, significant performance gains are achieved for W-TFMCC. Table-11 shows that coverage of TFMCC drops

from 61.65% to 30% by moving from terrain A to E. On contrary, our proposed protocol W-TFMCC drops coverage

from 61.65% to only 56.25% by moving from terrain A to E, as shown in Table-12. Moreover, W-TFMCC provides

61.65% coverage in slight tough terrains (B, C) too. Major reason for performance enhancement of W-TFMCC isattributed to group based channel distribution.

7. Prior State-of-the-art Approaches

In this section, we present the performance results of previous state of the art approaches. In [2], authors enhance

the quality of experience (QoE) by investigating the most appropriate technology for IPTV. Authors conclude that

IEEE 802.11a provides the best performance preceded by WIMAX and IEEE 802.11g network. In [6], authors show

that 6 7 video sources over 1-hop and 2 3 video sources over 3-hop wireless network are supportable. In[7],

authors show through experiments that IPTV over IEEE 802.11ncan support 12 and 10 users in indoor and outdoor

environments, respectively. In[5], authors show that IEEE 802.11band IEEE 802.11gnetworks can support 2 and 6

streams while IEEE 802.11ncan support dozens of IPTV streams. In [8], authors perform IPTV experiments and show

21


22/23

the commercial aspects of wireless IPTV. In [14], authors show through experiments that three SDTV channels with at

least one VoIP stream can be supplied over two hops in IEEE 802.11 b/g access points. In [20], authors show through

ns2 simulations that multimedia applications can be provided with less delay, less jitter and less packet loss with their

proposed protocol. In [9], authors verify through simulations that a protocol in between TCP and UDP at transport

layer can provide better streaming performance. Comparison of various studies shows that performance of previous

studies was limited owing to less throughput of predecessor IEEE 802.11 standards. Although TFRC has provided

better performance but it becomes less useful for IPTV which requires multicast wireless transmissions. Our proposed

protocol W-TFMCC has shown to support 32 VoIP users with 1 multicast IPTV user (experimentally) which is an

unprecedented effort. Major novel improvements in our analysis include the analytical and experimental investigations

of optimal transport layer protocols with optimal physical layer parameters for transmission of combined IPTV and

VoIP over IEEE 802.11n, which has not been covered in previous literature.

8. Conclusion

In this paper, we evaluate the capacity of IPTV and VoIP over IEEE 802.11n experimentally and analytically. Our

investigation over transport layer protocols shows that use of TFRC instead of UDP at transport layer can enhanceIPTV capacity by 25%. TFRC adjusts its data rate according to the congestion situations in the network which makes

it highly suitable for IPTV. Results show that TFRC provides fair share in bandwidth to TCP traffic than UDP. Our

investigation over physical layer parameters of IEEE 802.11n shows that queue size and contention window have

optimal size of 70 pkts and 11 respectively for IPTV and VoIP users. Study shows that optimal values of SIFS, DIFS,

Physical header can increase capacity by 10% at least. Trends of aggregation show an optimal aggregation of 4 pack-

ets for IPTV and VoIP beyond which large delay disrupts the support for many users. Our major contribution is the

proposition of W-TFMCC protocol with extensive simulations and experiments. Performance of TFMCC protocol

deteriorates in severe wireless conditions due to absence of group based transmission mechanism. W-TFMCC pro-

tocol provides multicast HDTV and SDTV streams to users based upon the delay and RTT of packets of all users.

W-TFMCC incorporates not only number of users but deals with the channel viewership too. Number of streams trans-

mitted by W-TFMCC are directly related to the number of channels watched by users. Results show that performance

of W-TFMCC is greater than TFRC/TFMCC if at least two users are watching the same channel. W-TFMCC adjustsits data rate according to the network conditions by making user groups depending upon the packet loss conditions.

Our study concludes that use of W-TFMCC with optimal physical layer parameters can increase network capacity at

least by 44% in comparison to UDP and TFRC respectively.

9. Bibliography

[1] Cisco, Cisco Visual Networking Index Forecast 2011-2016, available athttp://goo.gl/hSJjX

[2] M. Garcia, J. Lloret, M. Edo and R. Lacuesta, IPTV distribution network access system using WiMAX and WLAN technologies, In

workshop on Use of P2P, GRID and agents for the development of content networks (UPGRADE-CN), Germany, June 2009.

[3] S. Saleh, Z. Shah, A. Baig, IPTV Capacity Analysis using DCCP over IEEE 802.11n, Published in the 78th IEEE Vehicular Technology

Conference (VTC Fall), Las Vegas, USA, Sep. 2013.

[4] S. Saleh, Z. Shah, A. Baig, Capacity Analysis of Combined IPTV and VoIP Over IEEE 802.11n, Published in the 38th IEEE Conference

on Local Computer Networks (LCN), Sydney, Australia, Oct. 2013.

[5] T. Guo, C. H. Foh, J. Cai, D. Niyati and E. W. M. Wong, Performance Evaluation of IPTV Over Wireless Home Networks, in IEEEtransactions on multimedia, vol. 13, no. 5, Oct. 2011.

[6] E. Shihab, F. Wan, L. Cai, A. Gulliver and N. Tin, Performance analysis of IPTV traffic in home networks, in IEEE Global Telecommuni-

cations Conference (GLOBECOM), pp. 5341-5345, Washington, D.C., Nov. 2007.

[7] M. Atenas, S. Sendra, M. Garcia and J. Lloret, IPTV Performance in IEEE 802.11 n WLANs, in IEEE Global Telecommunications

Workshop (GLOBECOM), pp. 929-933, Miami, USA, Dec. 2010.

[8] R. Kilik and K. Amadou, Wireless IPTV in practice, In IEEE International Carpathian Control Conference (ICCC), pp. 187-190, Kyoto,

Japan, May 2011.

[9] K. Piamrat, P. Fontaine, and C. Viho. Managing wireless IPTV in multimedia home networking, In IEEE International Conference on

Advanced Communication Technology (ICACT), pp. 352-356, PyeongChang, Korea, Jan. 2013.

[10] F. Chaparro, C. D. Guerrero and F. Fraile, Available bandwidth estimation for high quality television content, In IEEE Colombian Confer-

ence on Communications and Computing (COLCOM), pp. 1-5, Medellin, Colombia, May 2013.

[11] S. Garg and M. Kappes, Can I add a VoIP call?, in IEEE International Conference on Communications (ICC), Alaska, USA, May 2003.

[12] I. Ullah, Z. Shah, M. Owais and A. Baig, VoIP and Tracking Capacity over WiFi Networks, in 73rd IEEE Vehicular Technology Conference

(VTC Spring), Budapest, Hungary, May 2011.

22
http://goo.gl/hSJjXhttp://goo.gl/hSJjX


23/23

[13] S. Shin and H. Schulzrinne, Experimental Measurement of the Capacity for VoIP Traffic in IEEE 802.11 WLANs, in IEEE International

Conference on Computer Communications (INFOCOM), Alaska, USA, May 2007.

[14] M. Gidlund and J. Ekling, VoIP and IPTV Distribution over Wireless Mesh Networks in Indoor Environment, in IEEE transactions on

Consumer Electronics, Nov. 2008.[15] G. Bianchi, Performance analysis of the IEEE 802.11 distributed coordination function, in IEEE Journal on Selected Areas in Communi-

cations, vol.18, no.3, Mar. 2000.

[16] C. Yang and H. Wei, IEEE 802.11n MAC Enhancement and Performance Evaluation, in Springer Journal of Mobile Networks and Appli-

cations, vol. 14, issue 6, Dec. 2009.

[17] B. Ginzburg and A. Kesselman, Performance analysis of A-MPDU and A-MSDU aggregation in IEEE 802.11n, in IEEE SarnoffSympo-

sium, New Jersey, USA, April-May 2007.

[18] T. Selvam and S. Srikanth, Performance study of IEEE 802.11n WLANs, in proc. of the International Conference on Communication

Systems And Networks (COMSNETS), pp. 637-642, Bangalore, India, Jan. 2009.

[19] L.M. de Sales, R. de A. Silva, H.O. Almeida, A. Perkusich, Multi(Uni)cast DCCP for live content distribution with P2P support, in IEEE

Wireless Communications and Networking Conference (WCNC), Shanghai, China, Apr. 2012.

[20] S. Yue, Y. Cao, An improved TFMCC protocol based on end-to-end unidirectional delay jitter, in IEEE International Conference on

Communication Technology (ICCT), Jinan, China, Sept. 2011.

[21] F. Hou, Z. Chen, J. Huang, Z. Li, A.K. Katsaggelos, Multimedia multicast service provisioning in cognitive radio networks, in International

Wireless Communications and Mobile Computing Conference (IWCMC), Sardinia, Italy, July 2013.

[22] A. Botta, A. Dainotti, A. Pescap, A tool for the generation of realistic network workload for emerging networking scenarios, in Elsevier

Computer Networks, Volume 56, Issue 15, pp 3531-3547, 2012.[23] Iain E. Richardson,H.264 and MPEG-4 Video Compression: Video Coding for Next-generation Multimedia, John Wiley & Sons, Feb.

2004.

[24] B. Dekeris and L. Narbutaite, IPTV Channel Zap Time Analysis, in Proceedings of International Conference on Ubiquitous and Future

Networks, Hong Kong, China, June 2009.

[25] D. E. Smith,IPTV bandwidth demand: Multicast and channel surfing, in IEEE International Conference on Computer Communications

(INFOCOM), Alaska, USA, May 2007.

Saad Salehreceived his MS and BS degrees in Electrical Engineering from National University of Sciences and Tech-

nology (NUST), Islamabad in 2013 and 2011, respectively. Currently, he is working as a Team Lead (Researcher) at

AN-DASH Group, SEECS, NUST, Islamabad. His research interests include computer networks, emerging issues in

IEEE 802.11 WLANs, machine learning and social networks.

Zawar Shahcompleted his PhD degree in Electrical Engineering from the University of New South Wales (UNSW),

Sydney, Australia in 2009. Currently, he is a Senior Lecturer in Information Technology (IT) at Whitireia Commu-

nity Polytechnic, Auckland, New Zealand. His research interests include QoS issues in Wireless Networks, Vertical

Handover issues between 3G/4G networks, Cloud Computing, Network Architectures and Protocols.

Adeel Baigreceived Ph.D. and M.Eng.Sc. degree in computer engineering from the University of New South Wales,

Sydney, Australia, in 2007 and 2001, respectively. Currently, he is an Assistant Professor at School of Electrical

Engineering and Computer Science (SEECS), Islamabad. His research interests are in the protocols and applications

for on-board mobile communication networks, network optimization, and QoS provisioning.

23

Date post:	02-Jun-2018
Category:	Documents
Upload:	sayed-alhussaini
View:	217 times
Download:	0 times

Elsarticle Template Libre

Documents