Sie sind auf Seite 1von 6

Modeling and Simulation of MAC for QoS in IEEE 802.

11e Using
OPNET Modeler
Weihua Helen Xi, Toby Whitley, Alistair Munro, Michael Barton
Networks & Protocols Group, CCR, Department of Electrical & Electronic Engineering, University of Bristol
Bristol, UK BS8 1UB
email: helen.xi@bristol.ac.uk

Abstract enhancements in OPNET; the simulation results are


This paper presents an IEEE 802.11e model developed using the demonstrated in Section 5 and finally, the paper is concluded in
OPNET Modeler, which is employed for Medium Access Section 6.
Control (MAC) enhancements for Quality of Service (QoS).
IEEE 802.11e QoS is implemented using the concept of varying 2. 802.11 MAC Overview
levels of service for different traffic types. Four Access The IEEE 802.11 MAC specifies two access methods, the
Categories (ACs) are defined with each supporting a different fundamental Distributed Coordination function (DCF) and the
priority for accessing the radio channel. The QoS station optional Point Coordination Function (PCF). Most of the
(QSTA) based on the existing 802.11 model but using an AC commercial 802.11 products only employ contention based
priority queuing and access mechanism is presented. The DCF. Considering the doubts on the efficiency of PCF [5], it was
enhanced queue provides virtual contention for a transmission chosen to implement the 802.11e model supporting DCF mode
opportunity with higher ACs having a greater probability of only.
success. This work offers a flexible and more accurate model to
simulate the performance of QoS mechanisms within the DCF: The fundamental access method of the IEEE 802.11 MAC
802.11e draft. The simulation results prove this model performs is known as Carrier Sense Multiple Access with Collision
as expected and demonstrate the performance of different traffic Avoidance (CSMA/CA). All the stations share one radio
types under different network configurations for the selected channel, and before a station transmits data, the channel must be
workload. idle for a contention period. To determine whether the channel is
idle or not, the station (STA) needs to continuously sense it
1. Introduction throughout this period. If the STA detects a signal with power
Simple deployment and high transmission speed make the higher than a specific fixed power threshold, the radio channel is
Wireless Local Area Networks (WLAN) more accepted in assumed to be busy and thus unavailable for transmission. The
public areas like offices and airports. IEEE 802.11 [1] defines the contention period includes a DCF Interframe Space (DIFS) and a
WLAN MAC and Physical Layer (PHY) Specifications. 802.11b backoff period. The DIFS is 34s in 802.11a and the additional
extends the higher-speed PHY to 11Mbps in the 2.4 GHz band. backoff period is determined as a multiple of a 9s slot time.
The standard OPNET WLAN model is based on these two
specifications. 802.11a [2] provides eight PHY modes with IFS: The interval between frames is called the Interframe Space
transmission rates up to 54Mbps by operating in the less used (IFS). IFS has four types in 802.11 (three shown in Figure 2).
5GHz unlicensed frequency band. Recently, with the increasing Short IFS (SIFS) is used for an ACK frame and a CTS frame
demands in transmitting multimedia over radio medium, the responding to a data frame and a RTS frame, respectively. PCF
IEEE 802.11e draft [3] defines MAC enhancements to support IFS (PIFS) is used by STAs operating under PCF mode to gain
LAN applications with QoS requirements. priority access to the medium. In the DCF mode, DIFS is used
instead of PIFS in PCF mode. The shorter the contention period,
This paper describes an approach for modeling the MAC the higher the priority is to access the wireless medium.
enhancements for QoS in 802.11e, based on the OPNET 802.11a
contributed model [4]. The OPNET 802.11 WLAN standard CW: After sensing the channel idle for DIFS, to minimize the
model served as the foundation for the 802.11a contributed collision possibility, the STA generates a random backoff period
model, in which the MAC and especially the PHY are heavily for an additional deferral time before transmitting. The
modified. In the presented 802.11e model, a priority queue to Contention Window (CW) is used to determine the number of
provide virtual contention among different traffic types for a slot times in the backoff period. The initial value of CW, or the
transmission opportunity within one station is defined, together reset of CW after a successful transmission attempt, is CWmin.
with changes in MAC to adapt the 802.11e standard. The eight The CW size will double for every unsuccessful attempt, until
PHY modes with the Orthogonal Frequency Division the CW reaches the value of CWmax. The STAs deferred from
Multiplexing (OFDM) modulation schemes developed in the channel access during busy period holds the backoff value and
802.11a model are applied. continue to count down the CW.

This paper is organized as follows: Section 2 provides a brief RTS/CTS: Because of the limited radio range, the transmitting
introduction to the legacy MAC defined in 802.11 and 802.11b; station may be unaware of stations that can interfere with the
the MAC enhancements in the 802.11e draft follow in Section 3; receiving station. The Request to Send (RTS) and Clear to Send
Section 4 presents the implementation of the MAC (CTS) mechanism is used to mitigate this hidden station
1
problem. The RTS/CTS frames transmitted prior to the actual
data frame contain the Network Allocation Vector (NAV) field,
which defines the period of time that the medium is to be
reserved until the end of the Acknowledgement (ACK) to the
following data frame. All STAs within the reception range of
either the originating STA (which transmits the RTS) or the
destination STA (which transmits the CTS) will keep quiet. The
header of the data frame also contains the NAV.

PCF: The PCF provides the contention free frame transfer. This
mode only works in infrastructure networks with the Access
Point (AP) working as a Point Coordinator (PC), which performs
the role of the polling master. The PCF distributes information
within Beacon management frames to gain control of the
medium by setting the NAV in STAs.

3. MAC Enhancements for QoS in 802.11e


The QoS of 802.11e is achieved by providing different classes of Figure 1: Virtual Contention within a station
frames with different priorities when accessing the radio
channel. The access method in 802.11e is called Hybrid
Coordination Function (HCF) and combines functions from both The accessing priority goes from AC0, AC1, AC2, up to the
the DCF and the PCF. The EDCF working in the Contention highest priority AC3. As the priority of the AC increases, the
Period (CP) is the Enhanced version of DCF, which is values of the MAC parameters become smaller. Thus the AC
implemented in this model. The PCF in the Contention Free with the shorter contention period has more priority to occupy
Period (CFP) is beyond the scope of this paper. the channel, as shown in Figure 2.

Each QoS data packet in QSTA carries its priority value (0 to 7)


in the MAC frame header. The QSTA provides four physical
queues, or ACs, to map the eight Traffic Categories (TCs). The
priority mappings are described in Table 1, taken from the IEEE
802.1d bridge specification [6].
Table 1: Access Category Mapping
User Access Traffic Type
Priority Category
1 0 Background
2 0 Spare
0 1 Best Effort Figure 2: Interframe spaces and backoff windows contend
3 1 Excellent Effort
for channel access in EDCF
4 2 Controlled Load
5 2 Video < 100ms latency and jitter 4. 802.11e Model in OPNET
6 3 Voice < 10ms latency and jitter The node model for the 802.11e STA has an architecture like the
7 3 Network Control OPNET standard node model of wlan_station_adv, which
includes the MAC, PHY (comprised of transmitter and receiver),
Every QSTA includes four virtual stations with each one wlan_mac_inf, source and sink, as shown in Figure 3. The higher
representing an AC, as shown in Figure 1. Four ACs are layers (such as TCP/IP protocols and applications) are replaced
provided by four queues. The QoS data coming from the higher by a source (the dotted box) and a sink process. The MAC
layer will drop into one of the four queues. Each queue works as interface wlan_mac_intf is an equivalent of Address Resolution
a DCF station with its own accessing parameters like AIFS[AC], Protocol (ARP). In this way, the effects of MAC attributes on
CW[AC], CWmin[AC] and CWmax[AC] instead of the DIFS, the performance of WLAN can be evaluated, independent of the
CW, CWmin and CWmax in DCF, respectively. AIFS[AC] is higher layers.
determined by:

AIFS[AC] = SIFS + AIFSN[AC] aSlotTime

Where AIFSN[AC] is an integer greater than zero [7]. A failed


transmission will multiply the backoff period by a Persistence
Factor (PF) [AC].

2
CWmin, CWmax and PF are each replaced by a relative vector
of four elements.

Figure 4: Process Model of Priority Queue Module

INIT state: Acquire the MAC parameter values set by the user
Figure 3: 802.11e WLAN Node model and calculate the contention periods. The contention periods
including the backoff window size are not real timers which will
count down with every tick of the clock; they work as the
The PHY in the 802.11a model is used, in which the eight sets of weights of the subqueue and will be updated every time this
transmitters and receivers employ the OFDM modulation QSTA transmits the previous packet successfully.
scheme. In the MAC process, changes are made to the variables
representing the contention parameters such as AIFS and CW, IDLE state: The machine enters an Idle state and waits for an
and three existing functions corresponding with the virtual incoming event. The event can be either an incoming packet
contention in the higher layer are modified. Additionally the from the 4 bursty source modules, a feedback interrupt from the
RTS/CTS function is corrected to behave as required by the MAC process to inform it of a successful transmission, or ready
standard. Another important contribution is to implement to send the next packet to the MAC layer to contend with other
functions of the virtual contention as shown in Figure 1 and thus stations for the radio channel.
the simple bursty source as used in the standard model
wlan_station_adv is replaced. RECEIVE state: Packets arriving from any of the four bursty
sources will trigger the ARRIVAL event. The state machine then
In the dotted box, each of the four bursty sources named as ac0 goes into the RECEIVE state and will insert the packet into the
to ac3 has an attribute Traffic Category with an integer value corresponding subqueue with the same AC. If no packets stay in
from 0 to 3. The On-Off Process model of the bursty source is the MAC layer to contend with other stations, an interrupt will
easy to configure as both Constant Bit Rate (CBR) and Variable be sent to trigger the SEND state.
Bit Rate (VBR) traffic. Each source will generate packets with a
TC_Packet format that has a field also called Traffic Category, UPDATE state: After MAC transmits a packet successfully to
which inherits its process models Traffic Category attribute the destination or the transmission attempt retry counter reaches
value at the initial state. Packets generated with different AC its limit, the MAC process will send the queue module a remote
values flow into the queue module vc, then the queue module interrupt to inform it that MAC is ready to acquire the next
inserts the packets into its subqueues indexed from 0 to 3. The packet. This remote interrupt triggers the UPDATE_CP event in
queue module is responsible for extracting the head packet of the order to update the CW sizes of the four subqueues. The CW of
subqueue with the highest priority, this having the shortest the three deferred subqueues will deduct the immediate past
contention period. winning subqueues last CW, while the past winner is reset to its
CWmin value.
The Finite State Machine (FSM) of the queue module is depicted
in Figure 4. It works as the virtual contention within the QSTA SEND state: Updating of the CW triggers the SEND_PKT event
shown in Figure 1. It can insert and extract packets from the and the state machine enters the SEND state. It compares the
corresponding subqueue, calculate the weights of each subqueue non-empty subqueues to find the smallest weight of the
and send the packet from the subqueue with the smallest weight subqueue. The selected subqueue extracts its head packet and
to the MAC layer. We extend the MAC access parameters to the sends it to the MAC layer. In the case where there is light traffic
modules attributes, so the user of the 802.11e model can easily and when all the four subqueues are empty, a flag is set. Thus
set different values to evaluate the impact of MAC parameters when the next packet is generated, it is sent to MAC
on the network performance of different traffic types. The AIFS, immediately.

3
5. Simulation Results statistics of the whole network and catalogued according to the
The goal of the simulation is to verify the expected operation of packet's AC.
this model. The simulation environments are described as
follows: Two sets of simulation are executed. The first scenario is to test
the performance of the different priorities in an Ad hoc network
Each network configuration in the simulations has the same with 10 QSTAs. The simulation is run over each of the eight
WLAN parameter settings (Table 2). The MAC address of each PHY modes of 802.11a. The second scenario runs at 24Mbps
station is automatically assigned by the OPNET Modeler. The speed in an Ad hoc network with 3, 5 and 10 QSTAs,
randomly selected destination address makes it possible for each respectively. The network structure is as shown in Figure 5, with
station to receive traffic equally. The RTS/CTS mechanism is each scale representing 12.5 meters.
used to mitigate the hidden station problem, although it adds
overhead and decreases throughout. Since the packet size is
below the 2034-byte limit, fragmentation is not needed.
Table 2: WLAN Parameters Setting
Date Rate (Mbps) 6, 9, 12, 18, 24, 36, 48, 54
RTS threshold (bytes) 256
Fragmentation Threshold None
Modulation Scheme OFDM
MAC Address Auto Assigned Figure 5: Network Structures
Destination Address Random
5.1. Simulations in an Ad hoc network with 10 QSTAs over 8
Each QSTA in our simulation has the same traffic generation PHY modes
pattern (Table 3). Packets of each AC have a size of 1500 bytes Figure 6 and Figure 7 illustrate the different Throughput and
and arrive every 5ms. Each station has four ACs and the data Delay of each AC in the Ad hoc network with 10 QSTAs. The
rate is 9.6Mbps. A simple and high traffic mode [8] [9] is used to figure of Load is very similar with that of Throughput, and the
make sure that the throughput and delay performance of each trend of Medium Access Delay is similar to Delay. Throughput
AC will be independent of the characteristics of the traffic and Delay statistics are used in this paper. The Throughput of
streams. For example, the starvation of the traffic or the AC3 is observed to be higher than the others at every PHY mode
correlated packet arrival of realistic voice traffic would influence and the Delay is lower as expected, this also agrees with
the throughput results [10]. simulation results in [7] [10]. AC0 suffers a significantly larger
Table 3: Packet Generation for every AC delay than AC1, AC2 and AC3, and the throughput is
significantly impacted owing to the low priority of achieving
Packet Size (bytes) constant (1500) transmission opportunities. The delay of the audio and video
Interarrival Time (seconds) constant (0.005) services (AC3 and AC2) is lower than 0.2ms, which is negligible
Data Rate (Kbps) 2400 for users in real audio and video transmission. From the PHY
mode point of view, the higher the PHY modes used, the shorter
Table 4 lists the MAC access parameters for each traffic type. the transmission time, so the shorter the delay and the higher the
Each station generates traffic for all the four ACs equally. throughput [11].
Table 4: MAC Access Parameters used for Simulations
Type AIFSD CWmin CWmax PF
AC3 1 7 15 1.2
AC2 2 10 31 2.0
AC1 3 15 255 2.6
AC0 4 31 1023 3.0

All simulations last 60 seconds of simulated time and the first 2


seconds of the data are discarded to ensure stable system
statistics. Performance measurements logged include
Throughput, Load, Medium Access Delay and Delay.
Throughput is the rate of bits being sent from MAC to the higher
layer. Load statistics are collected at the MAC layer when the
data frame arrives from higher layer at MAC and is ready to
transmit. Medium Access Delay is the period from the time
when a packet arrives at MAC until it is removed for
transmission. Delay is the time from when a packet arrives at the Figure 6: Throughput per AC over 8 PHY modes in an Ad hoc
transmission stations MAC until it reaches the destination Network with 10 QSTAs
stations MAC. Each of the four measurements is the aggregated
4
Figure 9: Delay per AC with Different Numbers of QSTAs at
Figure 7: Delay per AC over 8 PHY modes in an Ad hoc 24Mbps
Network with 10 QSTAs
We can also observe the network with 3 QSTAs experiences a
5.2. Simulations with Different Numbers of QSTAs at better performance than the 5 QSTAs and the 10 QSTAs
24Mbps Rate networks, i.e. higher Throughput and lower Delay. Figure 10
In this scenario, the PHY speed is fixed at 24Mbps, while we tells us the aggregated Data Dropped in the 10 QSTAs network
investigate the performance differences in three Ad hoc is 20 times as much as that of 3 QSTAs. The 10 QSTAs network
networks with different numbers of stations. Figure 8 and Figure has more than twice the number of transmission attempts, thus
9 demonstrate the Throughput and Delay of each AC run under more collisions occur, that cause more retransmissions. If the
networks with 3, 5 and 10 QSTAs, respectively. Three retransmission count reaches four, the packet will be dropped;
simulations indicate the similar results trend as concluded above. too many users can deteriorate network performance.
Higher AC traffic gain advantages from having QoS. The
highest priority packets which are delay sensitive like
conversations enjoy ten times less delay than the lowest priority
packets like large file downloading.

Figure 10: Data Dropped with Different Numbers of QSTAs


6. Conclusion
Figure 8: Throughput per AC with Different Numbers of In this paper, an 802.11e model developed using the OPNET
QSTAs at 24Mbps Modeler was presented. QoS of the model was implemented
using the concept of varying levels of services for different
traffic types, i.e. voice, video and data. We defined four Access
Categories, each of which supports a different priority to access
the radio channel. The QoS station was modeled with a priority
queue and access mechanism enhanced from the existing 802.11
model. Simulation results prove the validity of this model and
demonstrate its fidelity and flexibility for the study of IEEE
802.11e QoS mechanisms.
5
References [6] IEEE 802.1d-1998, Part 3: Media Access Control (MAC) bridges,
[1] IEEE Std. 802.11-1999, Part 11: Wireless LAN Medium Access ANSI/IEEE Std. 802.1D, 1998 edition, 1998.
Control (MAC) and Physical Layer (PHY) Specifications, Reference
number ISO/IEC 8802-11:1999(E), IEEE Std. 802.11, 1999 [7] S. Choi, J.Prado, S. Shankar N, and S. Mangold, IEEE 802.11e
edition, 1999. Contention-Based Channel Access (EDCF) Performance Evaluation,
in Proc. IEEE ICC03, Anchorage, Alaska, USA, May 2003.
[2] IEEE Std. 802-11a, IEEE Standard for Wireless LAN Medium
[8] ITU-T Recommendation H.261: Video codec for audiovisual
Access Control (MAC) and Physical Layer (PHY) Specifications: High-
services at 64 kb/s. (1993).
Speed Physical Layer in the 5 GHz Band, September 1999.
[9] ISO/IEC JTC1/SC29/WG11: MPEG4 coding of audio visual
[3] IEEE 802.11e/D4.0, Draft Supplement to Part 11: Wireless Medium
objects: visual. (1998).
Access Control (MAC) and physical layer (PHY) specifications:
Medium Access Control (MAC) Enhancements for Quality of Service
[10] S. Mangold, S. Choi, "Analysis of IEEE 802.11e for QoS Support
(QoS), November 2002.
in Wireless LANs," IEEE Wireless Communications, Dec 2003.
[4] B. E. Braswell, J. C. McEachen, "A Baseline Model for the IEEEE
[11] D. Qiao, S. Choi, and K. G. Shin, "Goodput Analysis and Link
802.11A WLAN Protocol," OPNETWORK 2001 Proceedings, August
Adaptation for IEEE 802.11a Wireless LANs, IEEE Trans. on Mobile
2001.
Computing (TMC), vol. 1, no. 4, pp. 278-292, October-December 2002.
[5] S. Mangold, S. Choi, IEEE 802.11e Wireless LAN for Quality of
Service, in Proc. European Wireless 02, Florence, Italy, February
2002.

Das könnte Ihnen auch gefallen