Documente Academic
Documente Profesional
Documente Cultură
Abstract— IEEE 802.11e Enhanced Distributed Channel Access ACK Skipping (DACKS) technique that includes a
(EDCA) is developed to provide Wireless Local Area Networks modification in the IEEE 802.11e Access Point (AP).
(WLANs) with Quality of Service (QoS) support. Several However, none of the previous works studies fairness
solutions have been proposed in order to provide a fair channel combined with performance taking into account the QoS
access for all competing stations. This paper first studies the QoS support. Work in [7] and [8] investigates the issue of fairness
capabilities of the Adaptive Weighted and Prioritized Polling in IEEE 802.11-based multi-rate WLANs but does not
(AWPP) protocol that adopts the frame structure and the basic consider QoS support or the main EDCA differentiation
polling scheme of the Priority Oriented Adaptive Polling (POAP) mechanisms specified by the IEEE 802.11e standard.
protocol. Our analytical approach is validated by plotting
analytical results against simulation outcome. We then explore Our paper analyzes the performance and fairness provision
the fairness provision of the AWPP protocol utilizing Jain’s of a novel resource distribution mechanism for centralized
fairness index. Finally, we provide a comparative performance wireless networks, which does not demand resource requests
analysis between EDCA, POAP and AWPP protocols and we and it is able to provide predictable QoS to mixed type network
demonstrate that AWPP outperforms the other two. traffic. The considered Adaptive Weighted and Prioritized
Keywords-QoS; fairness; medium access control; WLANs; Polling (AWPP) protocol [9] adopts the frame structure and the
IEEE 802.11e; EDCA; adaptive polling basic polling scheme of the efficient Priority Oriented Adaptive
Polling (POAP) protocol [10]. AWPP introduces a new
technique for efficient bandwidth allocation among wireless
I. INTRODUCTION stations and packet buffers. More specifically, it provides a
IEEE 802.11 Wireless Local Area Networks (WLANs) deterministic traffic differentiation engine that operates
have been very popular and deployed widely in public and proportionally to priorities as well as data rate and takes
private areas. The original IEEE 802.11 WLAN standard advantage of the existing infrastructure since it utilizes the AP
defines the Distributed Coordination Function (DCF) that for coordinating access on the shared medium. We first study
provides best-effort service at the Medium Access Control the QoS capabilities of AWPP by plotting throughput and
(MAC) layer. However, DCF cannot satisfy the demand for delay performance results for multiple traffic classes. We then
better Quality of Service (QoS) support for multimedia investigate for the first time the fairness of the AWPP protocol
applications that have attracted a lot of attention lately in by employing the Jain’s fairness index for various values of the
WLANs. The ratified IEEE 802.11e supplement [1] defines priority factor and the traffic rate. Finally, we compare the
Hybrid Coordination Function (HCF) in order to provide QoS performance of AWPP against EDCA and POAP protocols and
and differentiated services among contending stations. HCF show that AWPP outperforms both of them.
specifies two medium access functions, HCF Controlled
The remainder of this paper is organized as follows:
Channel Access (HCCA) and Enhanced Distributed
Section II provides a thorough background on related medium
Contention Channel Access (EDCA) for controlled and
access control schemes that provide QoS and focuses on the
contention-based channel access, respectively. This paper
three potential polling cases of the AWPP protocol. Section III
considers only the mandatory EDCA medium access function.
describes in detail the operation of AWPP and in particular the
In the literature, much research effort is focused on MAC buffer and station polling selection algorithms. A throughput,
solutions to support different QoS requirements for WLANs delay and fairness analysis (by employing the well known
[2]-[8]. In particular, a resource reservation based protocol Jain’s fairness index) is carried out in Section IV. Section V
capable of guaranteeing QoS is introduced in [2]. Authors in provides various simulation and analytical results that
[3] and [4] enhance and fine tune the parameters of EDCA. An demonstrate the validity of our analytical approach as well as
analytical model for the performance analysis of EDCA that the outperformance of AWPP against the EDCA and POAP
incorporates the main QoS features of the IEEE 802.11e protocols. Finally, Section VI concludes the paper.
standard is proposed in [5]. Work in [6] addresses the problem
of providing throughput guarantees to EDCA stations in a
WLAN with legacy DCF stations by proposing the Dynamic
j 0
The estimation of the throughput enables the computation
of the average delay based on Little’s law, which states that the
average system queue size equals the jobs’ arrival rate V. PERFOMANCE RESULTS
multiplied by the average waiting time. Thus, in order to get an The QoS capabilities of the AWPP protocol are studied via
indication of the delay, we first need to estimate the Average a network scenario that involves multiple traffic classes.
Buffered Traffic (ABT) as follows: Specifically, four traffic classes are considered: Best Effort
1 1 (l T ) (BE) traffic of priority 0, Excellent Effort (EE) traffic of
ABT V (t )dt (lt Tt )dt (17) priority 3, Streaming Multimedia (SM) traffic of priority 4, and
o o 2 Interactive Voice (IV) traffic of priority 6. Unless otherwise
where τ is the observation interval, V(t) is the buffered traffic at specified, the traffic generation rate of each class is equal to
time t, l is the traffic load rate and T is the average throughput 500kbps per flow. The AP is the source of four different class
(in terms of bit rate). Thus, the average delay is estimated to be flows destined to each wireless station, while another identical
equal to ABT/l. It should be clarified that this approach ignores four flows originate at each wireless station and destine to the
any packet drops that may occur due to buffer overflow or AP. We ensure traffic symmetry and keep the analysis practical
packet lifetime expiration. Hence, it is expected to have rather by setting AP_ExtraPriority equal to 0.
high delay values that may keep rising during the observation The main objective of this study is to examine the behavior
interval. of the different traffic class flows in a comparative way. Hence,
The last step in our analysis is the estimation of Jain’s following the presented approach we get results referring to the
Fairness Index, which is quite straightforward when already different traffic classes. Initially, we attain throughput and
having the throughput results. Specifically, it equals: delay results for the AWPP protocol and validate the presented
analytical approach using simulation. Some of the considered
network parameters are summarized in Table I.
Regarding the simulation process, we employed a unified
25000 30
BE (A) BE (S) BE (A) BE (S)
EE (A) EE (S) EE (A) EE (S)
SM (A) SM (S) 25 SM (A) SM (S)
20000
IV (A) IV (S) IV (A) IV (S)
Throughput (kbps)
20
15000
Delay (s)
15
10000
10
5000
5
0 0
0 25000 50000 75000 100000 0 25000 50000 75000 100000
Network Load (kbps) Network Load (kbps)
Figure 1. Simulation (S) and analytical (A) values of throughput versus Figure 2. Simulation (S) and analytical (A) values of delay versus
network load for the AWPP protocol network load for the AWPP protocol
0.85 10
0.8 8
0.75
Delay (s)
6
0.7
4
0.65 PF = 1 PF = 2
2 15000
0.6 PF = 3 PF = 4
0 10000
0.55
60000
50000 5000
0 25000 50000 75000 100000 40000
30000
Network Load (kbps) 20000 0
10000 Throughput (kbps)
Figure 3. Jain’s Fairness Index vs network load for the AWPP protocol Network Load (kbps)
event-based simulator developed in C++ to model the studied Figure 5. IV traffic: Delay and throughput versus network load
protocols, focusing on the accuracy of MAC. It is validated via
the comparison of its output with the analytical results. according to their priority, thus, they are all equally served and
the resulted fairness index is equal to 1.
Figures 1 and 2 provide simulative and analytical results for
the traffic throughput and average delay performance of Since AWPP distributes resources by also taking into
AWPP. As far as the analytical calculations, Eq. (16) was first account the traffic rate, we have examined how this factor
computed with the values presented in Table I and the result for affects the provided fairness. The 3D curve in Figure 4
BU was 0.876. Moreover, the network load is regulated by represents Jain’s Fairness Index as a function of varying BE
changing the number of stations. Our analytical model is and IV traffic rates. EE and SM traffic rates are considered
validated since an almost exact match between analytical and constant (equal to 500kbps). The figure illustrates that the
simulation results is observed. The figures demonstrate that variation of the high priority IV class traffic rate affects the
when low load is considered all traffic is fully served. provided fairness in a much greater degree than BE traffic.
However, when load increases the lower priority flows receive We next study the AWPP performance against the EDCA
limited bandwidth due to the fact that the higher priority traffic and POAP protocols in a specialized developed simulation
should be sufficiently served. framework, adapted to the operational characteristics of each
Figure 3 plots Jain’s Fairness Index versus network load for protocol. The condition of any considered link was modeled
different values of the Priority Factor (PF). It should be pointed using a finite-state machine and we employed the default
out that our objective is to measure fairness among traffic parameter values for all protocols. The comparative
classes but not per-station; hence, we have adjusted (18) performance of the three protocols is plotted in Figures 5 to 8
accordingly. Since PF actually defines how much a traffic class and the simulation results are derived by utilizing a statistical
is favored against a lower priority class, it is expected to have analysis based on the “sequential simulation” method. The
lower fairness among classes when PF is high. Furthermore, average delay and the traffic throughput have been depicted in
under high load conditions it is not possible to adequately serve a 3D curve against network load for each separate traffic class.
all flows and fairness decreases due to the fact that higher Figure 5 shows that in our scenario AWPP fully serves the
priority traffic gains more resources. As it expected, when PF
SM Traffic
equals 1, then based on (2) classes are not differentiated AWPP
POAP
EDCA
30
25
1
20
Jain's Fairness Index
Delay (s)
0.8 15
0.6 10
5
0.4
0
30000
0.2 8000
4000 4000 6000
3000 20000
2000 2000 4000
Network Load (kbps) 2000
IV DataRate (kbps) 1000 BE DataRate (kbps) 10000 0
0 0 Throughput (kbps)
Figure 4. Jain’s Fairness Index versus BE and IV traffic data rates Figure 6. SM traffic: Delay and throughput versus network load
10 [13] IEEE Std 802.1d-2004, “IEEE Standard for Local and Metropolitan
Area Networks Media Access Control (MAC) Bridges,” 2004.
[14] J. D. C. Little, “A proof for the queuing formula: L= λ w,” Operations
5 Research, vol. 9, no. 3, pp. 383-387, 1961.
[15] R. Jain, D. Chiu, and W. Hawe, “A Quantitative Measure Of Fairness
0 And Discrimination For Resource Allocation In Shared Computer
30000 Systems,” DEC Research Report TR-301, Sep. 1984.
15000
20000 10000
5000 Throughput (kbps)
Network Load (kbps) 10000 0