Controlling contention window to ensure QoS for multimedia data in wireless network

Download

Research and Development on Information and Communication Technology

Controlling Contention Window to Ensure QoS

for Multimedia Data in Wireless Network

Ngo Hai Anh, Pham Thanh Giang

Institute of Information Technology, Vietnam Academy of Science and Technology

Correspondence: Ngo Hai Anh. Email: [email protected]

Communication: received 8 November 2019, revised 11 August 2020, accepted 1 September 2020

Digital Object Identifier: 10.32913/mic-ict-research.v2020.n1.896

Abstract: The IEEE-802.11e standard was published with

the goal of ensuring quality of service, especially with mul-

timedia data. However, this standard only assigns different

priorities for different types of data but does not control

the sharing of bandwidth among different data flows. In this

paper, we will propose a method of sharing bandwidth for

proportional data flows and controlling Contention Window

of each priority flow to achieve that ratio.

the smallest. Therefore, in some cases, for example real–

time data, which provides service differentiation for best

effort and variable-rate real-time traffic, 802.11e cannot

provide adequate QoS [5]. And hence it is necessary to

have a more flexible division mechanism. We will solve the

problem by measuring actual data each node received over

a period of time, and then comparing it to theoretical data

to determine whether to increase or decrease the Contention

Window (CW) value. Our proposed algorithm will control

the increase or decrease of this CW value to achieve flexible

ratio to adapt user’s on-demand with various types of data

such as voice, video and background.

Keywords: Ad hoc network, IEEE 802.11e EDCA, service

differentiation, traffic differentiation.

I. INTRODUCTION

Next parts of this paper include the following sections:

Section 2 summarizes some related research around the

issue we are going to solve; In Section 3, we analyze

the lack of 802.11e in adapting different throughput ratio

for different data types; Section 4 and 5 will describe our

proposed solution, the methods used to accomplish our goal

and the simulation scenario evaluating the results achieved

according to the outlined criteria.

For the purpose of providing QoS for multimedia data,

IEEE-802.11e was introduced in 2005 [1] and approved in

2012 [2]. The work was done with the objective of ensuring

quality of service for multimedia data. However, this stan-

dard has not yet ensured on-demand differentiation between

flow’s traffics. For example, the ratio between data types

such as voice, video, and best-effort is always kept at the

highest level for voice, and the lowest level for best-effort.

But this fixed ratio does not meet the flexible proportional

division, in saturation state, high priority Traffic Class (TC)

flows often gets all bandwidth, and low priority TC flows

are hard to access channels.

II. RELATED WORKS

Some research papers focus on designing a Medium

Access Control (MAC) and channel usage algorithm in

multi-rate Wireless Local Area Networks (WLANs) for

improving efficiency and sharing fairly channel resources

among contending nodes [6]. Aiming to address problems

that the high collision is often caused by Binary Expo-

nential Backoff (BEB) mechanism in the legacy IEEE-

802.11 and the shared channel can be overused by low bit-

rate nodes, authors proposed a Differentiated Reservation

(DR) algorithm to reduce the collision among contending

nodes by setting their back-off counter as a deterministic

value once successfully accessing the channel. Furthermore,

to eliminate the performance anomaly, some nodes are

permitted to send multiple packets in one transmission

opportunity according to their feature. Moreover, the al-

Some studies have shown that although IEEE-802.11e

has focused on improving QoS, there are still many issues

to be resolved [3, 4]. The research [3] investigates EDCA

performance in presence of voice and data integrated traffic,

and observes the inefficiencies that arise when access

differentiation is performed on the basis of traffic type

only. In [4], authors investigate the use of the 802.11e MAC

EDCF to address transport layer unfairness in WLAN.

In this paper, we will perform the evaluation by sim-

ulation to show that although IEEE 802.11e can provide

proportional bandwidth distribution for different types of

multimedia data, that division is fixed. For example, voice

data is always the largest ratio and background data is

Vol. 2020, No. 1, September

gorithm presented the implementation of the DR algorithm

that is readily applied to both the existing IEEE 802.11

DCF and 802.11e EDCA networks with the minimum

modification. In addition, we also investigate the limitation

of DR algorithm and propose a Group-Based Differenti-

ated Reservation (GDR) algorithm applied to high density

scenarios. The results of theoretical analysis and simulation

validate that the proposed algorithms (DR and GDR) can

obtain high throughput, good airtime fairness and low

collision rate.

classes of data and each class has its own queue, then

virtual collisions can occur when each Traffic Class (TC)

queue must contend for access within the station’s queue

manager, while real collisions occur when the winning

queue is allowed to transmit. The results in [12] show

that throughput in 802.11e is decreased and latency is

increased when compared to 802.11a networks due to this

extra contention. Hence, dynamic and aggressive tuning of

the network parameters is required to achieve benefit from

prioritization.

In the paper [13], authors extend the approach for the

IEEE-802.11e network, where different QoS classes are

defined and show how to find the class-specific optimal

Contention Window sizes that yield the maximum aggre-

gate throughput while maintaining the target throughput dif-

ference between classes. However, this approach is mainly

based on mathematical models with assumptions that are

difficult to realize by simulation or in practice.

In the research about a fair access mechanism in IEEE-

802.11e networks [7], some proposed mechanisms are

evaluated. According to their evaluated strengths and weak-

nesses, authors proposed a new mechanism for TXOP

determination. This algorithm considers data rate, channel

error rate and data packet lengths to calculate adaptive

TXOPs for the stations. The simulation results show that the

proposed algorithm leads to better fairness. It also achieves

higher throughput and lower delays in the network.

Another method uses a two-level protection and guar-

antees a mechanism for voice and video traffic in IEEE

802.11e wireless LANs [14]. In first-level protection, exist-

ing voice and video flows are protected from new and other

existing voice and video flows. In second-level protection,

the voice and video flows are protected from best-effort

data traffic. For each protection level, a couple of protection

mechanisms are proposed. Extensive simulation results

show that the proposed two-level protection and guaranteed

mechanism are very effective in terms of protecting and

guaranteeing existing voice and video flows as well as fully

utilizing the channel capacity

Some previous paper on IEEE-802.11 QoS has proved

that 802.11 parameters can be adjusted to get better network

performance by maximizing throughput based on current

network situations [8] [9]. 802.11 also allows parameter

tuning such that some services can be accommodated and

provided some QoS guarantees [10]. But these approaches

do not provide different priority levels then lead to the de-

velopment of 802.11e which has been done to dynamically

adjust 802.11e parameters to provide better performance

and QoS.

In research [11] the effects of different 퐶푊 sizes are

explored. This work shows that 퐶푊_푚푖푛default in 802.11

leads to worse network performance, and finds that in sit-

uations with small numbers of stations, 퐶푊_푚푖푛with small

values greatly increased the collision probability, reduced

overall throughput, and increased delays. However, when

the network is fully utilized, the probability of transmission

when the medium was free of a station also increases

because the number of stations was increased. This research

also finds out that when a single station is transmitting,

throughput is greatly increased when 퐶푊_푚푖푛values are

small since a physical transmission medium is less idle

time.

IEEE-802.11e parameters can be tuned based on network

conditions to allow better performance than a single set-

ting [15]. Although these settings are not changed dynam-

ically in this study, they do show that changing network

conditions require changing parameters to use the channel

efficiently. These optimizations are evaluation in a test-

bed under realistic conditions. The authors investigate two

methods of choosing 퐶푊_푚푖푛in IEEE-802.11e networks

based on proportional fairness and time-based fairness.

They conclude that proportional fairness in a network based

on weights provides higher throughput than time based

fairness. Their work shows that 퐶푊_푚푖푛is an important pa-

rameter and can be tailored to a network in the case of nodes

with or without ensuring fairness or traffic differentiation.

The effect of controlling dynamically to QoS prioritiza-

tion was presented in [12]. The authors show that IEEE-

802.11e designation increases the possibility of collisions

and increases delay by adding an extra contention layer. In

normal IEEE-802.11 networks the only collisions which can

occur are those between stations as they try to gain access

to the medium to transmit the packet at the head of the

single send queue. In contrast with 802.11, there is only

one type of data, because 802.11e divides into multiple

From the above research studied, we find that sharing of

traffic by different data types without permanently assign-

ing QoS parameters will be more effective for multimedia

data, which is the type of data that always varies according

to users’ needs. The next sections of the paper will present

our approach to achieving that goal.

Research and Development on Information and Communication Technology

III. ANALYZING OF EDCA THROUGHPUT

be sent with highest priority, so AC_VO is selected for this

type of data. Default values of ACs are described in Table

III. Note that although in this Table III, 802.11e EDCA

defined eight default priorities for different data types [2],

however, eight is not fixed and 802.11e EDCA allows less

or more than the number of eight.

TABLE I

USER PRIORITY AND ACCESS CATEGORY [2].

Priority

Designation

lowest

AC_BK

AC_BE

AC_VI

AC_VO

Background

Best effort

Video

Voice flow

Video flow

Video

Sender node

Receiver node

Voice

Background flow

highest

Figure 1. Two nodes with three data flows scenario.

EDCA parameters for each AC are shown in the Table

III.

The simulation parameters was shown in Table III.

TABLE II

CALCULATION OF CONTENTION WINDOW BOUNDARIES.

TABLE III

SIMULATION PARAMETERS.

CWmin

CWmax

AIFSN

TXOP limit (ms)

AC_BK

AC_BE

AC_VI

AC_VO

1023

Parameter

Value

3.008

1.504

Channel data rate

Antenna type

Radio Propagation

Transmission range

Carrier Sensing range

MAC protocol

Connection type

Packet size

Send rate

Simulation time

11 Mbps

Omni direction

Two-ray ground

250 m

550 m

IEEE 802.11e introduced some enhancements to existing

DCF and PCF of IEEE 802.11 with Hybrid Coordination

Function (HCF) [1] [2]. The HCF includes two mechanism:

Enhanced Distributed Channel Access (EDCA), which is

an enhanced DCF, and HCF Controlled Channel Access

(HCCA), which is an enhanced PCF. Both EDCA and

HCCA could work separately or together, in contract of

802.11, where DCF is mandatory and PCF is optional.

While the fundamentals of the original functions were not

changed, augmented information allows HCF to provide

QoS to specific flows and/or stations. In EDCA, MAC-layer

parameters provide priority to each Traffic Class (TC) in a

contention access manner similar to DCF. The parameters

that can be manipulated are the Arbitration Inter-Frame

Space (AIFS), the Transmission Opportunity (TXOP), the

퐶푊푚푖푛, and the 퐶푊푚푎푥. These parameters are given

default values at each station for each TC, or they can be

overridden by an Access Point using special coordination

frames.

IEEE 802.11e (EDCA)

UDP/CBR

1024 bytes

incremental from 1 to 1000 (packets/second)

400 s

With proper tuning of EDCA parameters set, traffic per-

formance from different access categories can be optimized

and prioritization of traffic can be achieved. It means that,

these parameters are simply intended to define a fixed

priority for each type of data: voice is always of the highest

priority, then lower for video, and background.

IEEE 802.11e EDCA uses different parameter sets lead-

ing to different priorities for data flows [2]. Therefore,

although there is no contention of accessing channel, be-

tween the flows there will be a contention of using channel.

This leads to different throughput of each flow. However,

multimedia data does not always have a fixed priority. In

many cases, we may need to change the priority flexibly, so

assigning fixed EDCA parameters to each data type will not

solve this flexible control priority problem. To clarify this

issue, we will focus on analyzing the effect of Contention

Window sizes to throughput in 802.11e EDCA.

EDCA uses differentiated, distributed and contention-

based medium access mechanism to enhance the original

DCF by providing prioritized QoS for each type of data

flow. EDCA mechanism defines four Access Categories

(ACs) to provide support for delivery of traffic with User

Priorities (UPs) at the stations. There are four traffic types

corresponding to the four ACs: AC_BK, AC_BE, AC_VI

and AC_VO stand for background, best effort, video, and

voice traffic, respectively. For example, voice data should

We evaluate the performance of 802.11e by using sim-

lulation tool. There is a lot of software which is used to

simulate wireless networks but is limited to the original

802.11. NS2 [16] is very popular for network simulations

but it does not support 802.11e. Therefore, we have applied

Vol. 2020, No. 1, September

the extension patch for 802.11e from [17] which was inte-

grated with NS2 to simulate and evaluate related 802.11e

problems in this paper.

observation of how the throughput ratio of three data types

changes corresponding to CW changes.

Looking at the simulation results in Figure 3. Here, the

throughput of background data is considered as the basis

(Th BE), the throughput ratios of Voice and Video com-

pared to Background (Th VO:BE and Th VI:BE) are used

to observe the effect of changing CW values to through-

put of three data types. When CW increases, throughput

decreases, whereas reduced CW increases throughput. As

we see, the values of CW corresponding ratio 3:2:1 is

17:20:32. We can see the change of throughput over dif-

ferent CW values, so in order to have flexible change ratio

as user’s demand, we will need many sets of CW values. It

means that when assigning fixed CW values for data types,

changing throughput ratio in 802.11e is very difficult to

achieve.

We consider a simple two-node model with the flows

of the three data types (voice, video, best effort) as shown

in Figure 1 to investigate the 802.11e parameters set for

contention only between data flows with different priorities,

there is no contention in accessing channel. This model will

be used to evaluate the influence of Contention Window

values to the throughput of different data types.

Multimedia data can use TCP or UDP, but in this paper,

we perform simulations with UDP data only as shown in

Table III because many video and multimedia applications,

such as VoIP use UDP. These applications can tolerate

some data loss with little or no noticeable effect. The

reliability mechanisms of TCP is not suitable for real-

time applications as the re-transmissions can lead to high

delay and cause delay jitter, which significantly degrades

the quality.

Figure 2 show the throughput of each Traffic Class

according to offered load, with default IEEE 802.11e pa-

rameters, the offered load is increased with high priority.

Figure 3. Estimation throughput by priority level (퐶푊_푚푖푛) between Voice

and Video with Background data.

In the next section, we will propose a method to achieve

this ratio with dynamic controlling CW size.

IV. PROPOSED METHOD TO ACHIEVE

THROUGHPUT BY RATIO

Figure 2. Simulation result of two nodes with three data flows scenario.

We demonstrate the relationship between Contention

Window size of three data types (Voice, Video, and Back-

ground). To do that, we fixed two CW-size of the highest

(voice data) and the lowest (background data) priorities and

step-by-step changed the CW-size of video data. The range

of CW values varies in many cases , for example {3 —

15} in default EDCA parameter as shown in Table III,

or {7 – 31} with QoS Access Point/Basic Service Sets

(BSS) [18], or {17 – 1023} for maximal fair throughput

allocation with a toy WLAN scenario consisting of one

Access Point and two associated stations [19]. In order

to show that our proposed method can adapt to variable

CW values, we will take CW values of three types of data

which also change from 17 to 33, and CW size of the video

data which will change in that range (17 to 33) to give an

1. Define the ratio share between traffic class

IEEE-802.11e provides priority level for each traffic

class, but it cannot guarantee throughput ratio suitable for

each traffic class [20]. We propose the weighted number

for each priority different from defined settings of IEEE-

802.11e, these numbers will correspond to throughput ratio

in our proposed solution. It means that, with a proposed

numbers 3:2:1:1 shown in Table I V .1, our proposed method

will get the ratio shared of throughput [of 2 Mbps total] in

Figure 5.

Our previous work shows that the fairness between dif-

ferent data flows in 802.11e can be achieved via changing

퐶푊 value with simple simulation scenario with two nodes

sender–receiver [21], and the achieved result of fairness

Research and Development on Information and Communication Technology

mainly limited aspects allocate resources between flows in

a traffic class (the same services).

In this paper, we improve the mechanism by flexibly

assigning 퐶푊 values to different data flows to achieve

flexible ratio with more complex simulation scenarios. It

means that users can simultaneously use different types of

data and still achieve a fair share ratio of all traffic classes.

TABLE IV

NUMBERS FOR THROUGHPUT RATIO WITH DIFFERENT DATA

TYPES.

Priority

Data type

Weighted number

Figure 5. Expected throughput sharing by ratio with weighted numbers.

lowest

AC_BK

AC_BE

AC_VI

AC_VO

Background

Best-effort

Video

Then the control mechanism at the MAC sub-layer is

proposed to control the corresponding Contention Window

value (퐶푊) for the throughput ratios corresponding to the

different Access Types (퐴퐶) according to the parameters

퐶푊[퐴퐶], since a large/small 퐶푊 value will result in

a short/long timeout (backoff) and consequently the cor-

responding larger/smaller throughput ratios, but will still

not exceed network bandwidth. An example of the result

would look like Figure 5, where bandwidth 2 [Mbps]

is divided among different data types (Video:Voice:Best-

effort:Background) in a ratio of 4:2:1:1).

Video

Voice

highest

Voice

Our proposed method will control 퐶푊 size to adapt

to different throughput ratios of {background, best effort,

video, voice} such as {10:10:30:40} or {15:15:30:40} per-

centage (%) of total bandwidth.

2. Control CW to achieve the ratio share between

traffic classes

Packet Reception

The basic feature of IEEE-802.11e is to ensure that fixed-

level priorities i.e. voice data are always assigned higher

priority than background data, which leads to low-level of

fairness in network. In order to achieve a more flexible

throughput sharing, our proposed mechanism will adjust

Contention Window (퐶푊) size, because the 퐶푊 size is

related to the back-off time, and the longer or shorter back-

off time leads to the opposite of access time being smaller

or larger. It means that network traffic will vary with 퐶푊

sizes.

Access Category Mapping

Lower Priority

Higher Priority

Background

Queue

Best effort

Queue

Video

Queue

Voice Queue

Adjusting MAC-layer value: Contention Window (Back-off time)

PROPOSE: Dynamic Control CW for different throughput ratio

Back-off

CW[AC1]

CW[AC2]

CW[AC3]

CW[AC4]

Based on the cross-layer scheme to ensure fairness in

IEEE 802.11 [22], we propose a number of MAC layer

improvements in IEEE-802.11e in order to achieve fairness

between different data flows (video, audio, text, for exam-

ple). In general, data flows should have different priorities

of bandwidth. For example, video data will need more

bandwidth than voice data, or when calling phone over

Internet (VoIP), voice data will be more prioritized than

video data. Therefore, we need to have the corresponding

way to assign bandwidth usage for data flows. We did this

work based on two modules named TX Flow Estimation and

Utilization Estimation. In fact, it is very difficult to have any

method to ensure fairness in different layers (application,

Background

ratio

Best effort

ratio

Video

ratio

Voice

ratio

Figure 4. Controlling CW size to adaptive different throughput ratio.

With the proposed mechanism as shown in the processing

module in Figure 4, after receiving input packets, the Ac-

cess Category Mapping will put these packets into different

priority queues, for example Background and Best Effort

data onto low priority queues, Voice and Video data into

high priority queues.

Vol. 2020, No. 1, September

network, MAC, etc.). Our proposed method will try to focus

on ensuring fairness so that high-priority flows will not take

all the bandwidth of low-priority flows.

below.

Algorithm IV.1: (퐴푐푡푖푣푒_푇푖푚푒[푖])

Initialization:

퐴푐푡푖푣푒_푇푖푚푒[푖] = 0

Module TX Flow Estimation works in MAC layer to

count the number of flows in the transmission range. We

call these flows TX flows. A TX flow is determined by MAC

addresses, IP addresses of source and destination and flow

type (e.g. Voice, Video, BestEffort), by decoding the header

of the packet. We define TX flows as 푛_{푇 푋 [푖]}where 푖 stands

for {Voice, Video, BestEffort}. Note that in this case, TX

flow is end-to-end flows data type, and it can include some

process-to-process flows.

푇_{퐴푐푡푖푣푒[푖]}= 0



for each each interval time EP



퐴푐푡푖푣푒_푇푖푚푒[푖] = 0.8 × 퐴푐푡푖푣푒_푇푖푚푒[푖] + 0.2 × 푇_{퐴푐푡푖푣푒[푖]}



푇_{퐴푐푡푖푣푒[푖]}= 0



for each each packet p





if 푝 → 푑푒푠푡퐼퐷 == 푙표푐푎푙퐼퐷



if 푝 → 푇푦푝푒 == 퐶푇푆



푇_{퐴푐푡푖푣푒[푖]}= 푇_{퐴푐푡푖푣푒[푖]}+ 푇_{푅푇 푆}+ 푇_{퐶푇 푆}

if 푝 → 푇푦푝푒 == 퐴퐶퐾



Assume that, there are 푛 data flows with 푘_푖as the

weighted number of the data types defined in IEEE-

802.11e. E.g. throughput ratio of Voice, Video and Best

Effort is 3:2:1, we have 푘_{푉 표푖푐푒}= 3, 푘_{푉 푖푑푒표}= 2,

푘_{퐵푒푠푡퐸 푓 푓 표푟푡}= 1.



푇_{퐴푐푡푖푣푒[푖]}= 푇_{퐴푐푡푖푣푒[푖]}+ 푇_{퐷퐴푇 퐴}+ 푇_퐴퐶퐾+ 푇_{퐵푎푐푘표 푓 푓}





The 푅푒푎푙_푆ℎ푎푟푒_푅푎푡푖표[푖] is defined as the ratio of the

퐴푐푡푖푣푒_푇푖푚푒[푖] to the estimation period 퐸푃 as the below

equation:

We calculate the total flow 푛_{푡표푡푎푙}in 푇 푋퐹푙표푤퐸푠푡푖푚푎푡푖표푛

module by the following formula:

퐴푐푡푖푣푒_푇푖푚푒[푖]

푅푒푎푙_푆ℎ푎푟푒_푅푎푡푖표[푖] =

(3)

퐸푃

In this equation 3, estimation period (퐸푃) time is a given

observation period, chosen so that it is neither too short

to ensure evaluation process, nor too long because it will

distort the simulation. In the simulation done in the next

part of this paper, the 퐸푃 value which we selected is two

seconds. 퐶푊 will be adjusted, and the throughput will be

evaluated during this 퐸푃 value.

푛

∑

푛_{푡표푡푎푙}

푘_푖× 푛_{푇 푋 [푖]}

(1)

푖=1

Next we define the fair ratio share of the bandwidth for

each flow by the formula:

Based on the 퐶푊 size which has been defined in Table

2, adjusted value of 퐶푊 will be determined by the formula:

푅푒푎푙_푆ℎ푎푟푒_푅푎푡푖표[푖]

퐶푊⁰[푖] =

× 퐶푊[푖]

(4)

퐹푎푖푟_푆ℎ푎푟푒_푅푎푡푖표[푖]

푘_푖

In the above formula, the fairness value

퐹푎푖푟_푆ℎ푎푟푒_푅푎푡푖표[푖] is used as threshold of

퐹푎푖푟_푆ℎ푎푟푒_푅푎푡푖표[푖] =

(2)

푛

푘_푖× 푛_{푇 푋 [푖]}

priority for accessing channel. If flows realized

that the real value 푅푒푎푙_푆ℎ푎푟푒_푅푎푡푖표[푖] is less

than its 퐹푎푖푟_푆ℎ푎푟푒_푅푎푡푖표[푖], it will use the 퐶푊

size which is smaller than in the back-off state.

Thus, the flow can increase its chance to access

channels and bandwidth allocation. Vice versa, if flow

realized that 푅푒푎푙_푆ℎ푎푟푒_푅푎푡푖표[푖] is greater than its

퐹푎푖푟_푆ℎ푎푟푒_푅푎푡푖표[푖], it will use a greater value than

퐶푊 in back-off state. Therefore, it will reduce the chance

to access channel, leading to other disadvantaged flows

to have more opportunities to access the channel. In

case some of the flows only have a small offered load,

they will access channel more easily, and the remaining

bandwidth will be shared by other flows. Thus, it allows

푖=1

Utilization Estimation module evaluates real link uti-

lization of the flow. The link utilization is determined

by analyzing period 퐴푐푡푖푣푒_푇푖푚푒[푖] of the flow in a

predefined estimation period 퐸푃. The 퐴푐푡푖푣푒_푇푖푚푒[푖] of

the flow is defined as time used to transmit packets in

푓 푙표푤푖. The algorithm I V .1 is used to estimate the value

of 퐴푐푡푖푣푒_푇푖푚푒[푖] of the data flow 푖 by sensing packets.

We compute real time for sending data in given EP time by

80 percent of current sending time plus 20 percent of the

past calculated time (before) as shown in Algorithm I V .1

Research and Development on Information and Communication Technology

more efficiently use of channel bandwidth and ensure fair

bandwidth allocation among flows.

a) Three nodes model

The first model includes a chain of three nodes with

six flows corresponding with three types of data (Voice,

Video, Background) in Fig. 6. This model was known

with large-EIFS problem [24]. According to IEEE 802.11

specifications, before a station starts the transmission, it

should defer for a distributed IFS (DIFS) or an EIFS delay

time and then selects a random contention window (퐶푊)

counter for backoff. Once the backoff counter decreases to

zero, the station starts its transmission. The choice between

DIFS and EIFS depends on the event of the last trans-

mission. If the last event is an unsuccessful transmission

(e.g., a collision), the station waits for an EIFS, otherwise

it waits for a DIFS. The station which waits for an EIFS

is called a retry station. The EIFS in the IEEE 802.11

standard is the largest Inter-Frame Space (IFS) and is used

to protect ongoing frame, especially acknowledge frame

(ACK) from collisions. When stations detect a transmission

but cannot decode it, they set their Network Allocation

Vectors (NAVs) for the EIFS duration. However sometimes

the EIFS duration is larger/smaller than ongoing frame

duration, leading to considerable unfairness and throughput

degradation [25].

V. ANALYSIS OF SIMULATION RESULTS

1. The throughput by ratio and measurement to

evaluate

For priorities shown in Table III, we will evaluate and

simulate the weighted numbers 3:2:1 as shown in Table V .1.

TABLE V

THE WEIGHTED NUMBERS OF THREE DATA TYPES (VOICE,

VIDEO, BEST-EFFORT) FOR SIMULATION.

Priority

Data type

Weighted number

lowest

highest

AC_BE

AC_VI

AC_VO

Best-effort

Video

Voice

To evaluate the effect of throughput sharing by ratio

according to our proposed method, we use the fairness

index, which is defined by R. Jain [23]. In this paper, the

Fairness Index is evaluated based on the goodput at the

destination station. We propose the equation for calculating

fairness by priorities as follows:

Voice flow

Video flow

Voice flow

Video flow

푛

푥_푖

Sender node

Sender node 2

Receiver node

(

)

푘_푖

푖=1

퐹푎푖푟푛푒푠푠퐼푛푑푒푥 =

(5)

Background flow

푛

푥_푖

푘_푖

푛 ×

(

)

푖=1

Figure 6. Three nodes model.

Here, 푛 is the number of flows, 푥_푖is the end-to-end

throughput of flow 푖, and 푘_푖is the weighted number with

the corresponding types of data. The result ranges from 1/푛

(worst case) to 1 (best case), and it is maximum when all

flows receive the same allocation. This fairness value will

be used to evaluate the sharing ratio of different bandwidth

data types, meaning that the closer the value is to one (1),

the more likely the ratio will be achieved.

We examine network performance in this model by

letting sender nodes 1 and 2 generate traffic at the same

offered load to receiver node. The performance metrics

Fairness Index and Total Throughput are evaluated with

offered load.

2. Simulation models

We evaluate proposed method with different simula-

tion models. The simulation parameters are described in

Table III with NS-2 simulator. The EDCA 퐶푊 used

values of three data {Voice, Video, Background} are

{17, 20, 32} as shown in Fig. 3 will ensure the ratio

Voice:Video:Background = 3:2:1. For more complex mod-

els, we use the simulations below, this ratio is only achieved

with our proposed mechanism but not with IEEE-802.11e,

as in sections: V .2.a(2 senders), V .2.b (3 senders), and V .2.c

(multiple senders with random model).

Figure 7. Fairness Index in Three nodes model.

Vol. 2020, No. 1, September

In Fig. 7 and 8, “EDCA” means the result by using

IEEE 802.11e and “Proposed Method” means our CW

size adjustment shown in the paper (Section IV). Fairness

Indexes are shown in Fig. 7. When offered load is small,

all flows get its requirement. When offered load becomes

larger, in EDCA, fixed priorities (corresponding fixed CW

size) are assigned to different data types and higher priority

flows can get more opportunity to transmit data, then

Fairness Index might be worse. In our method, flexible CW

size is controlled to different data types, abd the throughput

of each flow becomes fairer and the bandwidth allocation at

the MAC layer is also improved. Thus, our method achieves

good Fairness Index.

b) Three pairs model

Figure 9. Three pairs model.

Figure 9 shows three pairs of sender and receiver nodes.

The problem in this scenario is also known as three-pair

problem which was first investigated in [26]. In this sce-

nario, sender nodes 1-2 and 2-3 are out of the transmission

range but in the carrier sensing range. Senders nodes 1 and

3 are out of the carrier sensing range, hence the two external

pairs Sender 1 – Receiver 1 and Sender 3 – Receiver 3

are completely independent, i.e., they can send packets

simultaneously without interference with each other. Thus,

the two external pairs contend bandwidth only with the

central pair Sender 2 – Receiver 2, while the central pair

contends with both external pairs. In this topology, the

central pair cannot access the medium in the saturated state

in the original IEEE 802.11 [26].

We examine the network performance in this model by

letting the sender nodes 1, 2 and 3 generate traffic at

the same offered load to receiver nodes. The performance

metrics Fairness Index and Total Throughput are evaluated

with offered load.

Figure 8. Total throughput in Three nodes model.

The Total Throughput of all flows is shown in Fig. 8.

When the offered load is small, Total Throughput of all

methods is similar. When offered load becomes large, in

EDCA, bandwidth utilization is less efficient than others

because EDCA gives some advantages to traffics of higher

priority flows. In our method, by changing the flexibility

of CW size, the lower priority flows chance to access

channel can be increased by reducing CW size (back-off

time). Therefore, the throughput of different data flows

remains at a stable level for most of simulation time. It

means, the Fairness and Throughput are often opposite.

Figures 7 and 8 show that our proposed method achieves

better fairness and there is still has a slight difference in

throughput compared with EDCA.

Figure 10. Fairness Index in Three pairs model.

Fairness Index is shown in Fig. 10. When offered load

becomes large, EDCA cannot help the central pair access

Research and Development on Information and Communication Technology

the medium, because scheduling queue in EDCA only

works at the link layer, so it does not have information of

flows out of the transmission range. Therefore, it cannot

improve MAC layer fairness. In our Proposed Method,

the central pair finds out that its bandwidth is less than

fair bandwidth allocation. Then Proposed Method tries to

improve its chance to access channel by decreasing CW

Size and therefore reducing the back-off time of station

Sender 2. Thus, Proposed Method can achieve a better

fairness than EDCA.

Figure 12. Fairness Index in Random model.

Figure 11. Total throughput in Three pairs model.

The Total Throughput of all flows is shown in Fig. 11.

When offered load becomes large, EDCA achieves larger

Total Throughput than our methods. This phenomenon is

explained as follows. In EDCA, higher priority flows such

as voice flow and video flow can use most of channel

bandwidth with its CW value, so the throughput of the

left and right pairs (numbered one and three) will take up

most of the bandwidth, and the middle pair (numbered two)

has very little bandwidth. In our method, with flexible CW

size, we can give more opportunity for lower priority flows,

so the throughput of the middle pair will increase and the

throughput of the left and right pairs will decrease, leading

to the total throughput of our method being possibly lower

than EDCA.

Figure 13. Total throughput in Random model.

These terms of network performance are examined versus

the number of flows. The simulation results also prove that

our proposed method achieves good fairness performance as

in Fig. 12. Our throughput performance is slightly reduced

by the trade-off with fairness performance as described in

Fig. 13.

VI. CONCLUSION

Inheriting from IEEE 802.11, IEEE 802.11e standard

was developed and published, partly satisfying QoS for

multimedia data, but in terms of sharing fairness, this

standard is still limited because it gives relatively fixed

values to the QoS parameters. Our research has proposed

a mechanism to improve the fair sharing of bandwidth

between flows by adapting CW in a more reasonable way.

Our algorithm is simple to implement, and it provides

flexible CW value, corresponding to the value of priority of

throughput for different data types to achieve a reasonable

level of sharing between multimedia flows. The evaluated

and simulated results by NS-2 have verified that our method

is better than original IEEE-802.11e in some simulation

c) Random model

The third model is a random topology. We make a

topology with 50 stations at random positions in 1000[푚]×

1000[푚] area. Among those 50 stations, 푚 stations are

chosen randomly and these 푛 stations generate UDP traffic

flows to one destination station. Total offered load of source

stations is set equal to channel data rate 11[Mbps]. The

average of Fairness Index and total end-to-end throughput

are used as metrics to compare the throughput and fairness,

respectively. These terms of network performance are ex-

amined versus the number of flows. Each data point is the

average of over 50 simulations.

Vol. 2020, No. 1, September

models. Our future research is aimed at evaluating mixed

data UDP and TCP based on the wireless testbed system.

IEEE International Symposium on a World of Wireless, Mobile and

Multimedia Networks, June 2011, pp. 1–9.

[20] M. K. Alam, S. A. Latif, M. Akter, F. Anwar, and M. K.

Hasan, “Enhancements of the dynamic txop limit in edca

through a high-speed wireless campus network,” Wireless Personal

Communications, vol. 90, no. 4, pp. 1647–1672, Oct 2016. [Online].

Available: https://doi .org/10 .1007/s11277-016-3416-4

[21] N. H. Anh and P . T. G iang, “An enhanced mac-layer improving

to support qos for multimedia data in wireless networks,” Indian

Journal of Science and Technology, vol. 9, no. 20, 2016. [Online].

Available: http://www .indjst .org/index .php/indjst/article/view/92732

[22] P. Giang and K. Nakagawa, “Cross-Layer Scheme to control Con-

tention Window for per-flow Fairness in Asymmetric Multi-hop

Networks,” IEICE TRANSACTIONS on Communications, vol. E93-

B, no. 9, pp. 2326–2335, 2010.

[23] R. Jain, D. Chiu, , and W. Hawe, “A Quantitative Measure Of

Fairness And Discrimination For Resource Allocation In Shared

Computer Systems,” DEC Research Report TR-301, Tech. Rep.,

September 1984.

ACKNOWLEDGMENT

This work was financially supported by Institute of Infor-

mation Technology (IOIT), Vietnam Academy of Science

and Technology (VAST).

REFERENCES

[1] IEEE

802.11e,

https://standards.ieee.org/standard/802_11e-

2005.html.

[2] IEEE 802.11-2012, https://standards.ieee.org/standard/802_11-

2012.html.

[3] C. Casetti and C. Chiasserini, “Improving fairness and throughput for

voice traffic in 802.11e EDCA,” in Personal, Indoor and Mobile Ra-

dio Communications, 2004. PIMRC 2004. 15th IEEE International

Symposium on, vol. 1, September 2004, pp. 525–530.

[4] D. J. Leith and P. Clifford, “Using the 802.11 e edcf to achieve tcp

upload fairness over wlan links,” in Third International Symposium

on Modeling and Optimization in Mobile, Ad Hoc, and Wireless

Networks (WiOpt’05). IEEE, 2005, pp. 109–118.

[5] E. Park, N. Kwak, S. K. Lee, J. Kim, and H. Kim, “Provisioning

QoS for WiFi-enabled Portable Devices in Home Networks,”

TIIS, vol. 5, no. 4, pp. 720–740, 2011. [Online]. Available:

https://doi .org/10 .3837/tiis .2011 .04 .006

[6] J. Lei, J. Huang, J. Tao, and Y . X ia, “A Differentiated Reservation

MAC Protocol for Achieving Fairness and Efficiency in Multi-rate

IEEE 802.11 WLANs,” IEEE Access, vol. PP, pp. 1–1, 01 2019.

[7] M. Yazdani, M. Kamali, N. Moghim, and M. Ghazvini, “A fair

access mechanism based on txop in ieee 802.11 e wireless networks,”

International Journal of Communication Networks and Information

Security, vol. 8, no. 1, p. 11, 2016.

[24] Z. Li, S. Nandi, and A. K. Gupta, “Ecs: An enhanced carrier

sensing mechanism for wireless ad hoc networks,” Computer

Communications, vol. 28, no. 17, pp. 1970

– 1984, 2005.

[Online]. Available: http://www .sciencedirect .com/science/article/

pii/S0140366405001416

[25] F. Khan, “Fairness and throughput improvement in multihop wireless

ad hoc networks,” in 2014 IEEE 27th Canadian Conference on

Electrical and Computer Engineering (CCECE). IEEE, 2014, pp.

1–6.

[26] C. Chaudet, I. G. Lassous, E. Thierry, and B. Gaujal, “Study of the

impact of asymmetry and carrier sense mechanism in IEEE 802.11

multi-hops networks through a basic case,” in PE-WASUN ’04:

Proceedings of the 1st ACM international workshop on Performance

evaluation of wireless ad hoc, sensor, and ubiquitous networks. New

York, NY, USA: ACM Press, 2004, pp. 1–7.

[8] F. Cali, C. M, and G. E, “Dynamic Tuning of the IEEE 802.11

Protocol to Achieve a Theoretical Throughput Limit,” IEEE/ACM

Transactions On Networking, vol. 8, no. 6, pp. 785–799, December

2000.

[9] F.Cali, C. M, and G. E, “IEEE 802.11 Protocol: Design and

Performance Evaluation of an Adaptive Backoff Mechanism,” IEEE

Journal on Selected Areas in Communications, vol. 18, no. 9,

September 2000.

[10] D. Chen, S. Garg, M. Kappes, and K. Trivedi, “Supporting VBR

VoIP Traffic in IEEE 802.11 WLAN in PCF Mode,” IEEE Trans-

actions on Wireless Communications, 2002.

[11] S. Garg, M. Kappes, and A. Krishnakumar, “On the Effect of

Contention-Window Sizes in IEEE 802.11b Networks,” Avaya Labs

Research, June 2002.

[12] D. He and C. Shen, “Simulation Study of IEEE 802.11e EDCA,” in

Vehicular Technology Conference, vol. 1, Spring 2003, pp. 685–689.

[13] J. Yoon, S. Yun, H. Kim, and S. Bahk, “Maximizing Differentiated

Throughput in IEEE 802.11e Wireless LANs,” in Proceedings. 2006

31st IEEE Conference on Local Computer Networks, Nov 2006, pp.

411–417.

Ngo Hai Anh received his M.Tech degrees

from Vietnam National University and is a

PhD student at Graduate Univeristy of Sci-

ence and Technology, Vietnam Academy of

Science and Technology. He is currently

a researcher at Telemactis Department of

Institute of Information Technology, Viet-

nam Academy of Science and Technology.

His current research interests include wireless network, network

performance and network management.

[14] Y. Xiao, H. Li, and S. Choi, “Protection and guarantee for voice

and video traffic in IEEE 802.11e wireless LANs,” in INFOCOM

2004. Twenty-third AnnualJoint Conference of the IEEE Computer

and Communications Societies, vol. 3, March 2004, pp. 2152–2162.

[15] V. A. Siris and G. Stamatakis, “Optimal CWmin selection for

achieving proportional fairness in multi-rate 802.11e WLANs: test-

bed implementation and evaluation,” in WiNTECH ’06 Proceedings

of the 1st international workshop on Wireless network testbeds,

experimental evaluation & characterization, 2006, pp. 41–48.

[16] The Network Simulator: ns-2, http://www.isi.edu/nsnam/ns/.

[17] “IEEE 802.11e EDCA Simulation Model for ns-2 (TU Berlin),”

http://www2 .tkn .tu-berlin .de/research/802 .11e_ns2/, last accessed

on 02/11/18.

Pham Thanh Giang received his B.E.

degree from the Hanoi University of Tech-

nology, Vietnam, in 2002. He achieved

M.E. and D.E. degrees from the Nagaoka

University of Technology, Japan, in 2007

and 2010, respectively. He is currently a

researcher at Institute of Information Tech-

nology, Vietnam Academy of Science and

Technology. His interests are security, mobile ad hoc networks, the

Internet architecture in mobile environments and Internet traffic

measurement.

[18] “Wi-Fi Quality of Service,” http://www .revolutionwifi .net/

revolutionwifi/2010/08/wireless-qos-part-5-contention-

window .html, last accessed on 12/10/18.

[19] A. Garcia-Saavedra, P. Serrano, A. Banchs, and M. Hollick, “Energy-

efficient fair channel access for IEEE 802.11 WLANs,” in 2011

10 trang yennguyen 22680 Free

Download

Bạn đang xem tài liệu "Controlling contention window to ensure QoS for multimedia data in wireless network", để tải tài liệu gốc về máy hãy click vào nút Download ở trên

File đính kèm:

controlling_contention_window_to_ensure_qos_for_multimedia_d.pdf