A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY

SENSITIVE SERVICES OVER TDMA/TDD WIRELESS NETWORKS

∗

Francisco M. Delicado, Pedro Cuenca and Luis Orozco-Barbosa

Department of Computer Engineering

University of Castilla la Mancha

Campus Universitario s/n, 02071 Albacete, SPAIN

Keywords:

Resource Request, QoS, Wireless LAN, HIPERLAN/2, Delay Sensitive Trafﬁc.

Abstract:

Recent advances on wireless technology are enabling the design and deployment of multiservice wireless

networks. In order to be able to meet the QoS requirements of various applications, it is essential to deploy QoS

provisioning mechanisms. In this paper, we propose a QoS provisioning mechanism for providing support to

time constrained applications over wireless networks. The proposed mechanism is developed using a signaling

procedure. Our simulation results show the effectiveness of the proposed mechanism when supporting time

constrained services, such as MPEG-4 video communications.

1 INTRODUCTION

Nowadays wireless local networks represent an alter-

native to wired LAN’s. Current wireless LAN’s oper-

ate at transmission rates able to support various types

of applications, such as, data, voice and video. How-

ever, one of the main open issues is the deﬁnition of

QoS mechanisms capable of meeting the QoS require-

ments of the various applications. Within a broadcast

environment, such as a wireless LAN, it is important

to properly allocate the bandwidth to the various ap-

plications. Even though, recent standards have the un-

derlying elements to carry this important task, they

fall short on the deﬁnition of all the mechanisms re-

quired to implement a whole structured set of mech-

anisms. In this paper, taken as a basis the HIPER-

LAN/2 standard (H/2, 2000), we deﬁne a resource re-

quest mechanism capable of meeting the QoS require-

ments of time-sensitive applications.

The article is organized as follows. Section 2 pro-

vides a short overview of the HIPERLAN/2 standard.

In Section 3, we review the related bibliography on

the area of MAC algorithms for TDMA/TDD wireless

networks. Our proposed resource allocation mecha-

nism is described in Section 4. The results of a per-

formance evaluation study via simulation are given in

∗

This work was supported by the Ministry of Science

and Technology of Spain under CICYT project TIC2003-

08154-C06-02 and the Council of Science and Technology

of Castilla-La Mancha under project PBC-03-001.

Section 5. Finally, Section 6 concludes the paper.

2 HIPERLAN/2 MAC PROTOCOL

The HIPERLAN/2 MAC protocol (H/2, 2000) is

based on a dynamic TDMA/TDD scheme with cen-

tralized control, using as logical transmission unit

frames of 2 ms. Given that the allocation of the frame

resources to each Mobile Terminal (MT) is made by

the Access Point (AP), the requirements of the appli-

cation resources have to be known of these entities,

which are responsible of allocating the available re-

sources according to the user needs. Towards this

end, each MT has to request to the AP the required re-

sources by issuing a Resource Request (RR) message,

while the AP informs the MT of the positive outcome

by using a Resource Grant (RG) message. The con-

nections are identiﬁed by two identiﬁers: MAC ID

and DLCC ID. A MAC ID is assigned to each MT

at association time, while a DLCC ID is assigned to

each connection when this latter is setup; both identi-

ﬁers are assigned by the AP.

Figure 1 shows the format of the transmission

frame deﬁned by the HIPERLAN/2 standard for a

system operating under the infrastructure mode. The

frame is divided into four phases, each phase being

composed by a group of transport channels. A trans-

port channel is a logical entity and its classiﬁcation

depends on the type of data that it conveys.

402

M. Delicado F., Cuenca P. and Orozco-Barbosa L. (2004).

A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY SENSITIVE SERVICES OVER TDMA/TDD WIRELESS NETWORKS.

In Proceedings of the First International Conference on E-Business and Telecommunication Networks, pages 402-409

DOI: 10.5220/0001403604020409

 SciTePress

The phases of a frame are:

1. Broadcast phase: this phase is used for the commu-

nications taking place on the downlink. It is formed

by a preamble to determine the beginning of the

frame and three transport channels:

• BCH: contains the conﬁguration parameters of

the frame, such as the number of RCH channels,

the size of the preambles of the uplink phase and

RA, among others.

• FCH: its size is variable and includes the re-

source grant (RG) messages for each connection.

These messages specify the number of granted

transport channels and their position within the

frame.

• ACH: contains the information regarding the

number of collisions having occurred in the RA

phase of the previous frame.

2. Downlink phase: this phase is formed by a group

of downlink PDU trains, which are formed by a

preamble and a variable number of SCH and LCH

channels dedicated to each one of those connec-

tions with resources granted in the FCH. The SCH

channels convey error control codes as well as in-

formation on any change on the connection param-

eters. The LCH channels transport user data, repre-

sented as PDU trains in Figure 1. Each PDU train

conveys the data pertaining to various connections

associated to a single MAC ID.

3. Uplink phase: similar to the downlink phase, this

phase is also formed by Uplink PDU trains, sep-

arate one from another by a guard time, and com-

posed by a preamble and a variable number of SCH

and LCH channels dedicated to each one of the

connections. The SCH channels convey resource

request messages control codes or other control

messages. The LCH channels transport user data,

represented as PDU trains in Figure 1. Each PDU

train conveys the data pertaining to various connec-

tions associated to a single MAC ID.

4. Random Access (RA) phase: consists of the num-

ber of RCH channels determined in the BCH chan-

nel, the minimum number of RCH channels in each

frame should be one. Each RCH channel is inserted

in a RCH PDU which is formed by a preamble

and a RCH channel. It is used for the transmis-

sion of control information, provided that the MT

does not have granted resources in the UL phase,

or during the beginning of the association process.

Each RCH PDU is separated from the previous one

by a guard time. A contention process based on a

Slotted-ALOHA scheme is used to access the RCH

channels.

It is important to note that not all of the transport

channels have the same size and that this one depends

on the channel type. This is also true for the pream-

bles and guard times whose sizes are speciﬁed, at the

beginning of the frame, in the BCH channel (H/2,

2000).

3 RELATED WORK

Given the central role played by the MAC algorithms

on enabling the provisioning of the QoS requirements

to the various applications, there has been a large

number of studies focusing on the design and eval-

uation of QoS-aware MAC protocols. In (Karol et al.,

1995), the authors describe a resource reservation pro-

tocol. In this protocol, the MTs initially issue a re-

source request message by using a contention-based

protocol, similar to Slotted-ALOHA. Once having

been allocated a number of slots, the data packets

convey, via “piggybacking’, the following MT’s re-

source requests. As soon as the MT becomes idle, the

resources are freed. At a latter time, when the MT

becomes once again active, this one has to start the

reservation process by issuing a ﬁrst request via the

contention process.

The MASCARA algorithm introduced in (Passas

et al., 1997) makes use of a resource request mech-

anism similar to the one described in (Karol et al.,

1995). The mechanism makes use of a “scheduler”

based on a “token bucket” scheme to distribute the re-

sources among the active connections.

More recent works focusing on the HIPERLAN/2

standards (Kadelka and Masella, 2001), (Politis and

Tafazolli, 2002), have proposed several algorithms to

distribute the bandwidth resources to different types

of connections. Within a given type of connection,

the bandwidth is allocated by taking into account the

status of the buffers or the connection parameters

provided at the time of establishing the connections.

However, the authors do not specify the resource re-

quest (signaling) protocol.

In (Lenzini and Mingozzi, 2001) the authors intro-

duce an allocation algorithm based on a FIFO pol-

icy and a signaling protocol. The proposed allocation

algorithm allocates a ﬁxed number of control chan-

nels, whose associated connection can use to further

request more resources (bandwidth).

One of the main drawbacks of the request mecha-

nisms proposed by the authors in (Karol et al., 1995)

and (Passas et al., 1997) is that they have been built

around a contention-based MAC. It is well known that

the performance of these mechanisms severely de-

grades as the number of active connection increases.

This makes the allocation mechanism prone to delay

and losses; an undesirable condition when developing

QoS mechanisms. Furthermore, the “piggybacking”

mechanism can not be implemented in HIPERLAN/2.

A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY SENSITIVE SERVICES OVER TDMA/TDD

WIRELESS NETWORKS

403

One option is to replace it by a “polling” mechanism,

or by making use of a ﬁxed number of control chan-

nels as proposed in (Lenzini and Mingozzi, 2001).

The drawback of the latter approach is the excessive

overhead introduced as the number of uplink connec-

tions increases.

4 A NOVEL RESOURCE

REQUEST MECHANISM

One of our main design objectives has been the deﬁni-

tion of an effective signaling protocol. The effective-

ness of such mechanism will depend on the amount

of overhead: number of control packets and retrans-

missions required by the unsuccessful transmission

attempts due to the use of the contention-based mech-

anism.

The algorithm works in the following way. The AP

starts by polling the MT as soon as a connection is set

up. The MT replies to the AP’s polling by specifying

the amount of required resources. Based on the num-

ber of resource requests received, the AP will grant

the requested resources to the requesting MTs. De-

pending on the network operating conditions, it may

take several frames for the AP to completely grant all

the requested resources (LCHs). As soon as the AP

ﬁnishes granting the requested LCHs to a given MT,

the AP reserves a SCH for that MT. The MT can then

use this SCH channel to reply to the AP’s polling.

Figure 2 depicts the process used by this type of

connection. Figure 2.(a) shows the operation of the

polling mechanism implemented by the AP. The AP

starts by issuing a RG message, RG:#SCH=1, con-

veyed via an SCH channel to poll the MT. The MT

replies by a RR message requesting n LCH channels,

RR:#LCH=n. It is then assumed that the AP grants

i slots in a ﬁrst reply (RG) and the remaining n-i re-

quested resources in a second reply (RG:#SCH = 1,

#LCH = n-i). In this second RG, the AP also allocates

an SCH to the MT, the MT can use this SCH to place

its next resource request (RR) message.

Figure 2.(b) shows the case when the AP hav-

ing polled the MT issues a second polling message,

RG:#SCH = 1, after its timer has expired out. Fur-

thermore, in the case depicted, it is assumed that the

MT has been unable to grant the resources requested

in the ﬁrst request issued by the MT. This condition

may arise when due to a long backlog, the AP is un-

able to serve the request due to the lack of available

resources. In this case, after a timeout period, the AP

polls once again the MT.

By basing the mode of operation of the request

mechanism on a “polling” scheme we look to address

two main issues: 1) to minimize the number of con-

trol packets (overhead) required to convey the MT’s

requests to the AP; 2) to avoid the use of a contention

procedure, which under heavy load conditions may

result on excessive delays or even on an unstable op-

eration of the allocation scheme. It is evident that the

effectiveness of the overall control structure to meet

the QoS requirements will depend very much on how

the different elements cooperate. Under the proposed

architecture, it is required that the entity responsible

of allocating the resources counts with the most up-

to-date information on the status of the all the active

MTs. The use of a proper tuned-up polling mecha-

nism should prove more effective on distributing the

bandwidth among the active MTs by taking into ac-

count their level of activity. In fact, the polling mech-

anism can be further enhanced to fairly distribute the

bandwidth according to the policies on place in a

given setup.

Under the proposed scenario, we assume that the

AP will serve the MTs requests on a strictly ﬁrst-in

ﬁrst-out (FIFO) discipline. This will allow us to carry

out a comparative study of our proposed scheme with

some other schemes proposed in the literature. Our

future plans include the study of different scheduling

disciplines to supplement the signalling mechanism.

5 PERFORMANCE EVALUATION

The main objective of this performance evaluation is

to assess the behavior of the proposed resource re-

quest mechanism with variable bit rate trafﬁc and de-

lay sensitive applications. In our study we use one

HIPERLAN/2 cell operating in centralized mode. We

suppose that the connections have already been estab-

lished, i.e., it is assumed that all MTs and associated

connections are already engaged in the transmission

of video data. The only control messages are those

required to request and grant resources, RR and RG

messages, respectively. Furthermore, we assume the

use of use short preambles, times of guard of 2 ms,

and that the physical modes for the SCH and LCH

channels are QPSK3/4 and 16QAM3/4, respectively.

Regarding the applications each MT supports two

applications of MPEG-4 video ((MPEG4, 1999)) one

in the downlink and another in the uplink. Each video

application begins to transmit within a random pe-

riod given by the expression t = unif orm(0,

)

being f the frame rate. In this way, the peak peri-

ods of the source rates are randomly distributed along

a GOP period. The transmission of a video frame is

uniformly distributed along the interval of duration of

a frame (

). We use a sequence in QCIF format at

25 frames/sec of the movie Mr. BEAN (Fitzek and

Reisslein, 2001). Figure 3 shows a snapshot of the

trafﬁc generated by the MPEG-4 video sources. It is

clear from this ﬁgure that these sources exhibit a high

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

404

degree of burstiness characterized by a periodic trafﬁc

pattern and a high variance on the rates at which the

data is generated.

The model of HIPERLAN/2 network has been im-

plemented in OPNET 9.0 (OPNET, 1987-2002). We

have measured the jitter among video frames, i.e.,

the time between the last packets of two consecutive

frames and the end-to-end delay for each packet. In

order to limit the delay experienced by the video ap-

plication, an essential condition to guarantee the QoS

required by the video application, the maximum time

that a piece of video data (referred form now on as

packet) can remain in the transmission buffer has been

set to 100 ms. This time limit is on-line with the val-

ues speciﬁed in (Karam and Tobagi, 2000).

In order to evaluate the proposed mechanism, re-

ferred from now on as PROP, we carry out a compar-

ative study of our proposed scheme with two schemes

proposed in (Lenzini and Mingozzi, 2001). The ﬁrst

method therein, referred from now on as MING1 re-

serves a given number of LCH channels and one SCH

channel every given number of frames. In this way,

the MT counts at regular intervals with a given num-

ber of LCH channels for data transmission. The MT

uses the SCH channel to request for more resources

on a timely basis. The second method in (Lenzini and

Mingozzi, 2001), namely MING2, only reserves one

SCH channel every given number of frames. It is clear

that the performance of all three mechanism heavily

relies on parameters, such as the number of reserved

LCH and SCH channels, the frequency at which the

MT´s can place the resource request (RR) messages

and the timer used by the polling mechanism. We start

then by studying the performance of the MING1 and

MING2 schemes by tuning their system parameters

when supporting the video services under consider-

ation. This will set the basis for a fair comparative

study.

Figure 4.(a) shows the end-to-end delay as a func-

tion of the number for the MING1 scheme. We have

varied the number of reserved LCHs per frame as well

as the reservation period of the SCH channel. This

last parameter has been expressed as

#SCH

#frames

. From

the ﬁgure, it is clear that the system performance is

severely affected as the number of reserved LCHs is

increased. This is due to the fact that this scheme is

unable to statistically multiplex the trafﬁc. This is par-

ticular important in a system supporting video trafﬁc.

This trafﬁc is characterized by a high variability. This

is clearly shown by the fact that when LCH=4, the

maximum number of MTs is six while for LCH=2,

the maximum number MTs reaches 12. From the ﬁg-

ures, it is also clear that by varying the reservation

period of the SCH does not help when the system op-

erates under heavy load. On the contrary, under low

loads, the periodicity of the SCH has a clear impact on

the system performance. From the ﬁgure, a clear ex-

ample is provided for the system consisting of 6 MTs

and making use of 2 LCHs.

Figures 4.(a) and (b) show that the best results are

obtained for the system with the least number of re-

served LCHs and lowest ratio of SCH per frame. It is

important to note that the number of LCHs (LCHs=2)

is given by the average bitrate required by the video

sequence.

In the case of the MING2, no reservation of LCHs

whatsoever is made. However, the AP changes the

rate at which the SCH is assigned to the MTs. Fig-

ure 5.(a) shows the end-to-end delay. This scheme

exhibits similar performance to the one obtained for

the MING1 scheme. At low loads, the performance

on the system improves as the frequency at which the

SCH´s are provided to the MT´s is increased. This is

due to the fact that the MTs can more frequently place

their requests. It is clear that under these load condi-

tions, there are plenty of available resources to fulﬁll

the MT’s requirements. However at high load condi-

tions, (#MT ≥ 15), the system shows better results as

the SCHs are granted at a lower rate. Once again, this

can be explained to the decrease in the multiplexing

gain due to the resources granted to the connections.

As more resources are reserved, the frequency of the

control slots (to convey the MT’s requests) should be

made less frequent to free resources that can be effec-

tively used for data transmission.

Regarding the jitter, the MING2 exhibits the best

results for the cases when the SCHs are granted at the

lowest rate (longest period), see Figure 5.(b).

In the case of the PROP mechanism, Figure 6.(a)

shows the end-to-end delay. From the ﬁgure, it is clear

that the best results are obtained when the timer is set

to 0.04 s. This can be explained by the fact that for

a shorter timer period, more capacity is used to the

SCHs resulting in a reduction of available capacity

dedicated to data transmission (LCHs). In the case

of low loads, the performance is independent of the

value of the timer; since the system is able to promptly

fulﬁll the MT´s needs. In other words, the system

is able to promptly poll the stations, i.e., before the

expiration of the timer.

Figure 6.(b) shows the jitter for the PROP mecha-

nism. From the ﬁgure, it is clear that the best results

are obtained for the case when the timer has been set

to 0.04 s.

In order to make a fair comparison, we have taken

the system conﬁgurations exhibiting the best results

for all three schemes. For the case of MING1,

LCH=2, and a SCH every 15 frames. Similarly to

MING1, for MING2, one SCH channel every 15

frames. In the case of PROP, the timer has been set

to 0.04 s.

Figures 7.(a) and 7.(b) show the end-to-end delay

and jitter for all three schemes, respectively. From

A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY SENSITIVE SERVICES OVER TDMA/TDD

WIRELESS NETWORKS

405

the ﬁgures, it is clear that our proposed mechanism

exhibits the best results. As shown in the ﬁgures, the

MING1 schemes is unable to provide support to more

than twelve stations due to the fact that this scheme

is unable to statistically assign LCH to the active con-

nections. In particular, it is important to note that un-

der low loads, our mechanism ensures a shorter delay

on the transmissions taking place over the downlink.

Figures 8 show the packet losses as a function of

active MTs. The losses reported in this ﬁgure cor-

responds to those PDU having resided in the MT’s

transmission buffers for more than 100 ms. As seen

from the ﬁgures, our proposed scheme exhibits good

results. The ﬁgures also show that the inability of the

MING1 scheme to provide lossless transmission even

to a system consisting of as few as eight MTs.

6 CONCLUSIONS AND FUTURE

WORK

In this paper a novel resource request algorithm for

HIPERLAN/2 has been developed and evaluated. We

have evaluated the proposed scheme and conducted

a comparative study with two other schemes previ-

ously proposed in the literature. We have been partic-

ularly interested in examining the ability of the pro-

posed scheme in supporting MPEG-4-based services.

Our results have shown that the proposed scheme is

able to support a larger number of connections than

a scheme based on the static reservation of a mini-

mum number of slots. Furthermore, we have shown

that proposed scheme exhibits a shorter end-to-end

delay under all network loads. We plan to conduct

further studies under a multi-service scenario, i.e.,

when various services, such as video, voice and data

are multiplexed together over the same wireless chan-

nel. As already mentioned, we will also consider to

supplement the proposed signalling mechanism with

a scheduling policy.

REFERENCES

Fitzek, F. H. and Reisslein, M. (2001). MPEG-4 and

H.263 video traces for network performance evalua-

tion. IEEE Network, 15(6):40–54.

ETSI TS 100 761-1. Broadband radio access networks

(BRAN); HIPERLAN type 2; data link control (DLC)

layer; part1: Basic data transport functions.

Kadelka, A. and Masella, A. (2001). Serving IP quality

of service with HiperLAN/2. Computer Networks,

37(1):17–24.

Karam, M. and Tobagi, F. (2000). On the Trafﬁc and Ser-

vice Classes in the Internet. In Proc. of IEEE GLOBE-

COM’00, San Francisco,CA, US.

Karol, M., Liu, Z., and Eng, K. (1995). An efﬁcient

demand-assignment multiple access protocol for wire-

les packet (ATM) networks. Wireless Networks,

1:267–279.

Lenzini, L. and Mingozzi, E. (2001). Performance evalu-

ation of capacity request and allocation mechanisms

for HiperLAN2 wireless LANs. Computer Networks,

37:5–15.

ISO/IEC 14496-2. Information technolo- gy –Generic cod-

ing of audio-visual objects– Part 2: Visual.

OPNET Technogies, Inc. OPNET Modeler 9.0. http:

//www.opnet.com

Passas, N., Paskalis, S., Vali, D., and Merakos, L. (1997).

Quality-of-service-oriented medium access control

for wireless ATM networs. IEEE Communications

Magazine, 35(11):42–50.

Politis, C. and Tafazolli, R. (2002). MAC protocol for sup-

porting QoS in All-IP HiperLAN2. In Pro. 4th Euro-

pean Wireless Conference (EW’02), Florence, Italy.

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

406

PSfrag replacements

2 ms

MAC frameMAC frameMAC frame

Broadcast phase

RA PHASE

PreamblePreamblePreamblePreamble

PreamblePreamble

Preamble

BCH FCH

ACH

Downlink phase

Uplink phase

DOWNLINK PDU TRAINDOWNLINK PDU TRAIN

UPLINK PDU TRAINUPLINK PDU TRAIN

RCH PDU TRAIN

SCHsSCHsSCHs

SCHs SCHsSCHs

LCHsLCHsLCHs

RCHRCH

MAC

ID k, DLCC ID j

MAC

ID k, DLCC ID k MAC ID m, DLCC ID i

MAC

ID l, DLCC ID qMAC ID l, DLCC ID p

MAC

ID m, DLCC ID r

Guard Time

Figure 1: HIPERLAN/2 MAC frame format

PSfrag replacements

2 ms

RG: #SCH = 1 (polling)

RR: #LCH = n

RG: #LCH = i, i ≤ n

send #LCH = i

RG: #SCH = 1, #LCH = n-i

send #LCH = n-i; RR: #LCH = k

(a)

PSfrag replacements

2 ms

RG: #SCH = 1 (polling)

RR: #LCH = n

RG: #LCH = i, i ≤ n

send #LCH = i

RG: #SCH = 1, #LCH = n-i

send #LCH = n-i; RR: #LCH = k

expire temp.

RG: #SCH = 1 (polling)

RR: #LCH = q, q ≥ n

(b)

Figure 2: Resource request Mechanism

A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY SENSITIVE SERVICES OVER TDMA/TDD

WIRELESS NETWORKS

407

2000

4000

6000

8000

10000

12000

14000

16000

0 5000 10000 15000 20000 25000

bytes

#frame

Figure 3: MPEG-4 stream

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0 5 10 15 20 25 30

seconds

#MTs

SCH per FRAME = 1/5

SCH per FRAME = 1/10

SCH per FRAME = 1/15

LCH = 2

LCH = 4

(a) Delay vs. #MTs

0.042

0.0425

0.043

0.0435

0.044

0.0445

0.045

0.0455

0 5 10 15 20 25 30

seconds

#MTs

SCH per FRAME = 1/5

SCH per FRAME = 1/10

SCH per FRAME = 1/15

LCH = 2

LCH = 4

(b) Jitter vs. #MTs

Figure 4: Delay and Jitter - MING1 mechanism

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0 5 10 15 20 25 30

seconds

#MTs

SCH per FRAME = 5

SCH per FRAME = 10

SCH per FRAME = 15

(a) End-to-End delay vs. #MTs

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0 5 10 15 20 25 30

seconds

#MTs

SCH per FRAME = 1/5

SCH per FRAME = 1/10

SCH per FRAME = 1/15

(b) Jitter vs. #MTs

Figure 5: Delay and Jitter - MING2 mechanism

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

408

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0 5 10 15 20 25 30

seconds

#MTs

Polling Thr. = 0.01

Polling Thr. = 0.02

Polling Thr. = 0.04

(a) End-to-End delay vs. #MTs

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0 5 10 15 20 25 30

seconds

#MTs

Polling Thr. = 0.01

Polling Thr. = 0.02

Polling Thr. = 0.04

(b) Jitter vs. #MTs

Figure 6: Delay - PROP mechanism

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 5 10 15 20 25 30

seconds

#MTs

MING1 (2,1/15)

MING2 (1/15)

PROP (0.04)

(a) End-to-End delay vs. #MTs

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0 5 10 15 20 25 30

seconds

#MTs

MING1 (2,1/15)

MING2 (1/15)

PROP (0.04)

(b) Jitter vs. #MTs

Figure 7: Delay and Jitter- all mechanisms

500

1000

1500

2000

2500

3000

3500

0 5 10 15 20 25 30

#LCH

#MTs

MING1 (2,1/15)

MING2 (1/15)

PROP (0.04)

Figure 8: Packets Losses - all mechanisms

A QOS-AWARE RESOURCE REQUEST MECHANISM FOR DELAY SENSITIVE SERVICES OVER TDMA/TDD

WIRELESS NETWORKS

409