You are on page 1of 6

2010 Sixth Advanced International Conference on Telecommunications

Differentiation Between Different Traffic Categories Using Multi-level of Priority in


DCF-WLAN
Mehaseb Ahmed Gamal Abdel Fadeel, Ibrahim Ismail Ibrahim
Electronics and Communication Department Electronics and Communications Department
Faculty of Engineering, Misr International University (MIU) Faculty of Engineering, Helwan University
Cairo, Egypt Cairo, Egypt
mehaseb_ahmed@hotmail.com gam_hel@yahoo.com, iiibrahim@softhome.net

II. IEEE 802.11 QoS OVERVIEW


Abstract‫ ــــ‬The IEEE802.11 protocol defines access techniques to
wireless medium using CSMA/CA, and defines two modes of This section will go thought different QoS mechanisms used
operation in WLAN DCF and PCF. DCF does not provide any in WLAN; generally, there are four parameters of QoS in
QoS, so that throughput and delay of real-time traffic, such as WLAN throughput, delay, jitter and packet loss rate, in
voice and video cannot be guaranteed over WLAN. IEEE 802.11e WLAN various ideas proposed to achieve QoS requirement,
introduces a service differentiation mechanism based on
which are classified into four methods, which will be detailed
prioritized contention access, which causes throughput
degradation if the network becomes high loaded by high priority separately as following
traffic. This paper proposes service differentiation for multiple
access categories by using multilevel priority mechanisms in A. Service differentiation
MAC layer, using multilevel priority controllability to network Service differentiation is based on the thought that different
performance according to throughput of each traffic category. requirements of different applications should be served
distinctively in different classes. There are two modes of
Keywords ‫ ـــ‬WLAN, QoS, DCF, MAC, Performance Analysis. differentiation service in IEEE 802.11, priority-based service
and fair scheduling service, the classification mechanism
I. INTRODUCTION identifies and separates traffics into different flow Therefore
each flow can be handled selectively. Priority based service
The IEEE 802.11 standard [1] defines the Medium Access
always serves flows with the highest priority. Fair scheduling
Control (MAC) and Physical Layer (PHY) specifications.
service fairly schedules the bandwidth on the basis of weight
Protocol defines two modes of communications for wireless
of each flow.
stations, Distributed Coordination Function (DCF) and Point
Coordination Function (PCF). There are many studies on DCF B. Resource reserving and admission control in MAC layer
protocol proposed in [2], [3] that presented analytical models Service differentiation is useful to support high QoS
of DCF throughput analysis under saturated conditions. The guarantee for real-time multi-media applications, but in the
DCF scheme is designed with no QoS mechanisms being situation of heavy loading, it cannot work efficiently. So, it
considered. needs to reserve resource and control accessing in MAC to
Recently IEEE 802.11e [4] was proposed to enhance QoS, ensure QoS of running applications.
Resource reserving: in IEEE 802.11 WLAN, it is hard to
IEEE 802.11 Working Group E [5] has defined a new
support resource reserving in WLAN based on CSMA/CA.
mechanism, Hybrid Coordination Function (HCF). HCF is
Many projects of resource reserving advise that IEEE 802.11
composed of two access methods, Enhanced Distributed
should do some modifications. For example, [6] defines a
Channel Access (EDCA) mechanism for contention based data
series of accessing systems, which can support resource
transmission and HCF Controlled Channel Access (HCCA)
reserving in MAC layer by enable multi-media channel
for contention free data transmission. IEEE802.11 guarantees
accessing.
QoS by using different values of CW min and CW max for all
Accessing control: it is hard to know the exact situation of
traffic classes, this causes throughput degredation when the loading at present because of wireless environment, so it is
network become high loaded with high priority traffic due to difficult to decide, which kind of accessing controlling. There
increase of collision. are two foundational methods, one is to pick up parameters of
In this paper, the proposed method is used to support present network situation, and the other is to conclude network
multiple traffic category service differentiation to guaranteed situation based on the calculation. In [7], there are some ideas
throughput and delay of specific access categories, by using to passively check channels by virtual frames in MAC,
multilevel of priority mechanisms in MAC layer. estimate the present service level by virtual frame, and adjust
The paper is organized as follows. Section II provides an applicable parameters dynamically basing on the change of
overview of QOS in IEEE 802.11 (DCF). In Section III, the channels situation by means of virtual MAC and Virtual
proposed scheme is explained. Section IV contains analytical Source algorithm. In [8], based on the calculation, the station
analysis of the proposed model. In Section V, simulation estimates the situation of network and makes access control
results are given to the proposed scheme. Finally, summary considerations.
and conclusion are drawn in Section VI.

978-0-7695-4021-4/10 $26.00 © 2010 IEEE 262


263
DOI 10.1109/AICT.2010.50
C. Information extracting system based on PHY layer
Packet from Upper Layers
Some parameters of MAC layer can be tuned to achieve
maximum throughput and minimum delay in another words, it
needs an information extracting system based on physical NO
Randomization Mechanism
layer of dynamic wireless environment to guarantee QoS. NO

The most direct way to do this is using SNR (Signal to


Noise Ratio) of channel, because IEEE 802.11 supports to get
SNR parameter from physical layer. In [9], assuming that RSS Po R
(Received Signal Strength) and SNR have linear relationship,
each station is assigned different RSS threshold so that they NO YES

use different transmission rate to make whole WLAN system


efficient. In [10], the foundation of the algorithms is that
combining SNR parameters, payload length and
Collision YES
Pr R YES

retransmission counters to estimate the situation of the


wireless channel. In [11], estimating quality of wireless
channel from SNR, every station passes its SNR information Forward Packet to MAC Layer
to access point by SNR frames periodically, access point
updates arranges bandwidth for each station according to SNR
to maximize WLAN throughput. Fig. 1 Randomization Mechanism used by Best-effort Traffic.

unsuccessful transmission w is doubled up to its maximum


D. System tuning parameters
Tuning parameters defined in IEEE 802.11, like RTS value CW max = 2 m CW min and for successful transmission w
threshold [12] and fragmentation threshold, has a big impact decremented to its minimum value CW min as shown in Fig. 2
on the performance of WLANs. [14]. The backoff time counter is “decremented” as long as the
channel is sensed idle for a period equal to a Distributed Inter-
III. SCHEME DESCRIPTION Frame Space (DIFS), and “suspend” when a transmission is
detected on the channel, and reactivated when the channel is
IEEE802.11 DCF is capable of supporting best effort traffic. sensed idle again for more than a DIFS. The station transmits
This section introduces mechanism used to differentiate when the backoff time reaches zero.
between multiple traffic categories using two level of The second level of priority is differentiation between
differentiation, the first one is randomization mechanism, different traffic categories in the same traffic type by using
which used to randomize access of best-effort traffic to MAC contention window, where small values of CW min assigned
layer, and the second is using different values of minimum for high priority traffic, and large values of CW min assigned
contention window to determine priority of traffic category. for low priority traffic so priority mapping classifier used to
Randomization mechanism used by best-effort traffic [13] map each traffic category to its contention window based on
illustrated in Fig. 1, where the new generated best-effort its priority.
packet accesses the DCF MAC, at beginning of any contention
window, with probability Po , on the other hand if collision is
detected, this (backlogged) packet schedules for
retransmission (i.e., re-access) at beginning of any contention
window (CW) with probability Pr until it is successfully
transmitted. By other words a newly generated best-effort
packet accesses the DCF MAC if and only if the condition
Po R is valid, if the packet suffer from collision the station
become in backlogged state and backlogged packet try to
access DCF MAC if Pr R . Note that R is a uniformly
distributed random number in range from 0 to 1 selected by
the station each time try to access medium.
DCF uses a binary exponential backoff (BEB) scheme,
which enables stations to generate a random backoff interval
before transmission of each packet [1].
Backoff time = rand( 0,w − 1 )*σ
where σ is duration of time slot, and w is contention
window, it depends on the number of transmissions failed for
the packet. At first transmission attempt, w is set equal to a
Fig. 2 Binary Exponential Backoff Algorithm.
value CW min , called minimum contention window; after each

264
263
Fig. 3 shows the proposed MAC scheme for multiple access
categories service differentiation. At the first all generated
packets from upper layer is classified in to real-time traffic and
best effort traffic using traffic classifier, where real-time
packets can access MAC layer directly but best-effort traffic
access MAC layer through randomization mechanism, so best-
effort traffic suffers from lager delay than real-time traffic. So
randomization mechanism is considered first level of priority,
and then contention window value used to assign different
priorities for each traffic category.
The proposed scheme considers two main traffic flows
(Real-time traffic and Best-effort traffic), and each traffic flow
consists of multiple traffic categories, assumes that different
types of traffic categories, X ( X 1 ) different traffic flows
for real-time traffic with types x ( x = 1,2,3..........., X ); also
Y ( Y 1 ) different traffic category for Best-effort traffic with
types y ( y = 1,2,3..........., Y ).
Fig. 3 MAC Scheme to Differentiate between Multiple Access Categories.

IV. ANALYTICAL MODEL


In case of successful transmission only one station
Consider a WLAN consists of N competing stations use transmits N = 1 , in this case the expected packet flow from
proposed MAC scheme, Let τreal , x and τbest , y is probability randomization mechanism is τ m = π 0 Fm (0) + π 1 Fm (1) .
that station carrying type x or y traffic transmit in randomly Also note that τreal , x and τbest , y represent probability of
chosen time slot, and τm is expected packet flow from transmission for each station to transmit real-time traffic or
randomization mechanism calculated as (5) in [13]. best-effort traffic respectively, wx represent contention
window for each traffic category in real-time traffic and wy
N represent contention window for each traffic category in best-
τm = ∑ π i Fm (i ) (1) effort traffic. Each packet collides with constant and
i=0
independent probability Px , real and Py , best , so collision
probability for each traffic type real-time traffic and Best-
2(1 − 2 Px, real ).(1 − τm)
τreal , x =
( )
effort traffic respectively is
(1 − 2Px, real )(wx + 1) + Px, realwx 1 − (2 Px, real )mx
X Y

x = 1,2,3..........., X Px , real = 1 − (1 − τreal , x) N −1 .∏ (1 − τreal , i ) N .∏ (1 − τbest , i ) N


i =1 i =1
i≠x

2(1 − 2 Py , best ).τm x = 1,2,3..........., X


τ =
best , y
(
(1 − 2 Py , best )(wy + 1) + Py , bestwy 1 − (2 Py, best )my ) Y X

Py , best = 1 − (1 − τbest , y ) N −1 .∏ (1 − τbest , i ) N .∏ (1 − τreal , i ) N


y = 1,2,3..........., Y i =1 i =1
i≠ y

Since i is the number of backlogged stations, at state i the y = 1,2,3..........., Y


mean best-effort packet flow Fm per CW from randomization Ptr is the probability that there is at least one transmission
mechanism to the input of DCF MAC is
in the considered time slot.
i −1 N −i −1 i
Fm (i ) = iPr( 1 − Po)
N −i
(i − P ) + Po( N − i )(1 − Po ) (i − P ) X Y

Ptr = 1 − ∏ (1 − τreal , i ) N .∏ (1 − τbest , i) N


r r
(2)
i =1 i =1

πi is Probability of finding system in state i,


Ps the probability that a transmission occurring on the channel
i = 0,1, 2,3,......., N . Also π i refer to equilibrium probabilities
is successful is given by the probability that exactly one
that satisfying the following
station transmits on the channel. Let Ps , x be the probability
N
π j = ∑ π i Pi , j , j = 0,1,2,3,......., N that there is one and only one transmission of real-time traffic
i =0
occurring in the channel, and Ps , y for best-effort traffic.
N
and ∑ πi = 1
i=0

265
264
X X
channel conditions (i.e., no hidden terminals and no capture
Ps ,real , x = Nτreal , x (1 − τreal , x ) N −1 .∏ (1 −τreal , i ) N .∏ (1 −τbest , i ) N effect); also traffic classes are assumed to be packetized with
i =1 i =1
i≠ x equal packet lengths in order to be able to determine the
Y X
degree of service differentiation.
Ps ,best , y = Nτbest , y (1 −τbest , y ) N −1 .∏ (1 − τbest , i) N .∏ (1 −τreal , i ) N
i =1 i =1
i≠ y
We consider four access categories that will be supported by
X Y the network as shown in Table II; since differentiation is done
So Ps = ∑ P s ,real,i + ∑P s ,best, j (3) by assigning different values for minimum contention
i =1 j =1
window, note that CW 1 ≺ CW 2 ≺ CW 3 ≺ CW 4 to achieve
Normalized system throughput S is differentiation between traffic categories with required
X Y priorities.
S = ∑ Sreal , i + ∑ Sbest , j
i =1 j =1
Normalized throughput per AC with respect to variation of
Sreal , x Is the Throughput of traffic category x of real-time Po at Pr = 0.9 is shown in Fig. 4 and Pr = 0.7 is shown in Fig.
traffic, and Sbest , y is the Throughput of traffic category y of 5, using different values of minimum contention window for
each access category. Figures show that for low values of
best-effort traffic. Throughput of each traffic category can be CW min the throughput increases and decreases for large
calculated as flow
values of CW min . So to guarantee throughput of real-time
Ptr .Ps , x .Ereal ,x traffic AC1 and AC2 will use small values of CW min
Sreal , x = (4) ( CW 1 = 16 and CW 2 = 32 ) compared to values of CW min
(1 − Ptr )σ + PtrTs Ps + PtrTc (1 − Ps )
used by best-effort traffic, where AC3 and AC4 uses large
Ptr .Ps , y .Ebest , y values of CW min , ( CW 3 = 64 and CW 4 = 128 ). So
Sbest , y = (5) contention window considered first level of priority used in
(1 − Ptr )σ + PtrTs Ps + PtrTc (1 − Ps )
differentiation between different AC.
Where Ts Is the average time the channel is sensed busy in
case of successful transmission, Tc is the average time the Note that increasing Po causes increasing in throughput of
channel is sensed busy by each station during the collision. best-effort traffic and decreasing in real-time traffic, which is
Value of Ts and Tc depend on the access technique used, very clear in Fig. 4 and Fig. 5. Studying the effect of
calculated as follow for RTS/CTS mechanism. randomization mechanism parameter Pr achieved by
comparing throughput of each access category in Fig. 4
⎧Ts , real = RTS + SIFS + δ + CTS + SIFS + δ + H + Ereal where Pr = 0.9 and Fig. 5 where Pr = 0.7 , we note that
⎪ decreasing value of Pr will causes in increase throughput of
⎪ + SIFS + δ + ACK + DIFS + δ
⎪ real-time traffic and decrease throughput of best-effort traffic
⎨Ts , best = RTS + SIFS + δ + CTS + SIFS + δ + H + Ebest and this is considered second level of priority that used to
⎪ + SIFS + δ + ACK + DIFS + δ differentiate between real-time traffic and best effort-traffic.

⎪⎩ Tc = RTS + DIFS + δ
Delay per access category is shown in Fig. 6 and Fig. 7 we
Simply average delay calculated directly by finding the can note that delay of real-time access categories is almost
following relation between delay and throughput. Since constant and not effected by changing Po or Pr , as expected
Dreal , x is delay of access category x in real-time traffic, and only contention window value has a small effect on the delay,
where increasing CW min causes increasing in time delay
Dbest , y is delay of access category y in best-effort traffic. before packet transmission. On the other hand delay of best-
effort categories is not guaranteed.
N .Ereal , x
Dreal , x = (6) TABLE I
S real , x PARAMETERS USED TO OBTAIN SIMULATION AND NUMERICAL RESULTS

Packet payload 8184 bits


N .Ebest , y MAC header 272 bits
Dbest , y = (7)
Sbest , y PHY header 128 bits
ACK 112 bits + PHY header
RTS 160 bits + PHY header
V. SIMULATION and RESULTS CTS 112 bits + PHY header
Channel Bit Rate 11 Mbit / sec
In this section, the mathematical model is compared with Propagation Delay 1 μs
simulation results to evaluate performance of proposed Slot Time 9 μs
scheme using RTS/CTS access mechanism in MAC layer, SIFS 16 μs
using parameters in Table I the simulation considers ideal DIFS 43 μs
Backoff Stage 6

266
265
TABLE II By comparing delay of real-time traffic in Fig. 6 and Fig. 7
DIFFERENT TRAFFIC CATEGORIES ASSUMED IN SIMULATION
with voice over IP QoS requirements in Table III, we can note
Traffic that the proposed scheme gives good performance for voice
Real-time Traffic Best-effort Traffic
Type transmission over WLAN.
Access
AC1 AC2 AC3 AC4
Category TABLE III
Randomization Randomization VOICE OVER IP QOS REQUIRMENTS
Direct Direct
Mechanism Mechanism
Access
MAC Pr and Po Pr and Po Quality Delay( ms )
CW 1 CW 2 Good 0-150
CW 3 CW 4
Medium 150-400
Priority High Low High Low Poor >400

Fig. 4 Throughput per AC versus Po at N = 10 at P r = 0 .9 Fig. 6 Throughput per AC versus Po at N = 10 at P r = 0 .9

Fig. 5 Throughput per AC versus Po at N = 10 at Pr = 0.7 Fig. 7 Throughput per AC versus Po at N = 10 at P r = 0 .7

267
266
To evaluate the performance of the proposed scheme, the [6] K. Liu “A Reservation-based Multiple Access Protocol with Collision
total throughput of the proposed scheme is compared with Avoidance for Wireless Multihop Ad Hoc Networks,” IEEE ICC ’03,
vol. 2, May 2003, pp. 1119–23.
throughput of IEEE 802.11 standard protocol using RTS/CTS [7] N. H. Vaidya, P. Bahl, and S. Gupta, “Distributed Fair Scheduling in a
(Fig. 8) it is clear that the proposed scheme for service Wireless LAN,” Proc. ACM MOBICOM 2000, Aug. 2000, pp. 167–78.
differentiation has the same throughput of IEEE 802.11 [8] M. Barry, A. T. Campell, and A. Veres, “Distributed Control Algorithms
protocol, which means that adding the randomization for Service Differentiation in Wireless Packet Networks,” Proc. IEEE
INFOCOM 2001
mechanism and scaling CW min according to traffic categories [9] J. D. Pavon and S. Choi, “Link Adaptation Strategy for IEEE 802.11
priority did not have effect on the throughput of standard WLAN via Received Signal Strength Measurement,” IEEE ICC ’03,
MAC layer but add concept of service differentiation to MAC vol. 2, May 2003, pp. 1108–13.
[10] D. Qiao, S. Choi, and K. G. Shin, “Goodput Analysis and Link
layer. Adaptation for IEEE 802.11a Wireless LANs,” IEEE Trans. Mobile
Comp., vol. 1, no. 4, 2002, pp. 278–92.
VI. SUMMARY AND CONCLUSION [11] H. Aida, “Wireless Packet Scheduling with Signal-to-Noise Ratio
Monitoring,” IEEE Annual Conference on Local Computer Networks
This paper introduced multiple access category service (LCN 2000), Tampa, FL, Nov. 2000.
[12] T. Sheu and T. Chen, “The Impact of RTS Threshold on IEEE 802.11
differentiation algorithms in DCF-WLAN. The differentiation MAC Protocol,” Tamkang Journal of Science and Engineering, vol. 6,
achieved by using two level of priority; the first level is no. 1, pp. 57-63, March 2003. (EI)
randomization mechanism, and second is the value of [13] M. A. Mehaseb, G. AbdelFadeel, and I. I. Ibrahim “Modified MAC
minimum contention window, to control throughput and delay Protocol for Service Differentiation in DCF-WLAN,” The Sixth IEEE
and IFIP International Conference on Wireless and Optical
traffic categories according to each traffic requirements. The Communications Networks (WOCN2009).
mathematical model is validated by simulation, which shows [14] Q. Pang, S. C. Liew, J. Y. B. Lee, and V. C. M. Leung, “Performance
that how control parameters can differentiate between evaluation of an adaptive backoff scheme for WLAN,” Wireless
different traffic categories. Communications and Mobile Computing, vol. 4, no. 8, pp. 867–879,
2004.
References

[1] IEEE Std. 802.11-1999. “Part II: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications”. Reference number
ISO/IEC 8802-11:1999(E). IEEE Std. 802.11, 1999 edition.
[2] G. Bianchi, ''Performance analysis of the IEEE802.11 distributed
coordination function" IEEE Journal in Selected Areas: Communication,
vol. 18, pp. 535{547, March 2000.
[3] F. Cali, M. Conti, and E. Gregori, “Dynamic Tuning of the IEEE 802.11
Protocol to Achieve a Theoretical Throughput Limit”. Networking,
IEEE/ACM Transactions on Vol. 8, Issue 6, Dec. 2000 Page(s):785 -
799.
[4] S. Xu, “Advances in WLAN QoS for 802.11: an Overview”. Personal,
Indoor and Mobile Radio Communications, 2003. 14th IEEE
Proceedings, Vol. 3, 7-10 Sept. 2003 Page(s):2297 - 2301.
[5] IEEE 802.11e/D4.0, “Draft Supplement to Part II: Wireless Medium
Access Control (MAC) and physical layer (PHY) specifications:
Medium Access Control (MAC) Enhancement for Quality of Service
(QoS)”. Nov. 2002.

Fig. 8 Throughput versus Number of Station of Proposed Scheme and


IEEE 802.11 Standards using RTS/CTS.

268
267

You might also like