The IEEE 802.11 standard defines the Point Coordination Function (PCF) in order to support time-bounded services, ex. voice frame and video frame. This priority access to the wireless medium is coordinated by a station called Point Coordinator (PC). Because the PCF may start transmissions after a shorter duration than DIFS but longer than SIFS that is called PCF Interframe Space (PIFS), the PCF has higher priority than the DCF. With PCF, a Contention Free Period (CFP) and a Contention Period (CP) alternate over time. Time is divided into repeated periods, called superframes. A superframe is combined by a CFP and a CP. DCF is used during the CP and PCF is used during the CFP.
A superframe starts with the beacon frame that is a management frame that maintains the synchronization of the local timers in the stations and delivers protocol related parameters. The PC, which is typically collocated with the AP, generates beacon frames at regular beacon frame intervals, thus every station knows when the next beacon will arrive. This time is called target beacon transition time (TBTT) and announced in every beacon frame. See Fig.2-5 for a typical sequence during CFP.
Because the PC itself has pending data for this station, it uses a combined data and poll frame by piggybacking the CF-Poll frame on the data frame. If the PC does not receive any response from a polled station after waiting for PIFS, it polls the next station, or ends the CFP. A specific control frame, called CF-End, is transmitted by the PC as the last frame within the CFP to signal the end of the CFP.
Chapter 2 Backgrounds
Fig.2-5 An example for the PCF operation [13]
2-3 Enhanced Distributed Coordination Function
IEEE 802.11 Task Group E currently defines enhancements to 802.11 MAC, called 802.11e, which introduces EDCF and HCF. With 802.11e, there may still be two phases of operation within the superframes, i.e., a CP and a CFP, which alternate over time continuously. The EDCF is used in the CP only, while the HCF is used in both phases.
In the CP , each Traffic Category (TC) with the stations contends for a TXOP and independently starts to count down their backoff counter after detecting the channel being idle for an Arbitration Interframe Space (AIFS). The AIFS is at least DIFS. See Fig.2-6 for illustration of the EDCF parameters.
Fig.2-6 Illustration of the EDCF parameters
A single station may implement up to eight transmission queues realized as virtual stations inside a station. If the counters of two or more parallel TCs in a single station reach zero at the same time, a scheduler will avoids the . The TC with highest priority will get the TXOP and other TCs will increase their CW and choose a value from the interval [0, CW]. There is then still a possibility that the transmitted frame collides at the medium with a frame transmitted by other stations.
virtual collision
Fig.2-7 shows the legacy 802.11 station and 802.11e station with four ACs with one station.
Chapter 2 Backgrounds
Fig.2-7 the legacy 802.11 station and 802.11e station with four ACs with one station [13]
The EDCA uses , , and instead of DIFS,
, and , of the DCF. The is determined by AIFS[AC] CWmin[AC] CWmax[AC]
CWmin CWmax AIFS[AC]
AIFS[AC]=SIFS + AIFSN[AC] SlotTime×
So the EDCA use the different parameter setting to prioritize different services, ex.
voice, video, and best effort.
2-4 Hybrid Coordination Function [12] [13]
The HCF extends the EDCF access rules. Only after detecting the channel as being idle for PIFS, the HC may allocate TXOPs to itself to initiate MSDU Deliveries whenever it wants. The QoS CF-Poll from the HC can be sent after a PIFS idle period without any backoff. So the HC can issue the polled TXOPs in the CP using its prioritized medium access.
During the CFP, the starting time and maximum duration of each TXOP is specified by the HC. Stations will not attempt to contend for the channel. Only the HC can grant TXOPs by sending QoS CF-Poll frames. See Fig.2-8 for an example of 802.11e superframe.
Fig.2-8: An example of 802.11e superframe where the HC grants TXOPs in CFP and CP
Chapter 2 Backgrounds
2-5 RTS/CTS-based access mechanism
The RTS/CTS-based access mechanism provides positive control over the medium in order to minimize the collisions caused by the hidden stations. Fig.2-9 shows the hidden terminal problem. For example, B is in the transmission range of C, but the others are not. B and D are in the transmission range of C, but A is not. A and C are unaware of each other since their signal do not carry that far. So their frames may collide with each other at B. But unlike an Ethernet, neither A nor C can be aware of the collision. A and C are called “hidden nodes” with respect to each other.
Fig.2-9 hidden node problem
RTS/CTS-based access mechanism can solve the hidden node problem. If A has a frame to B, A will send a RTS frame first. When B receives the RTS frame, B will return a CTS frame that contains a time value that alerts other stations to hold off from accessing the medium. After A receives the CTS frame from B, A will begin to send its data frame. Adding the RTS/CTS access mechanism will increase redundancy.
But if the data frame is always large like aggregated frames, the RTS/CTS access mechanism can reduce the collision cost instead. Fig.2-10 shows the RTS/CTS access mode and Basic access mode.
RTS CTS DATA ACK
SIFS SIFS SIFS DIFS
RTS
SIFS
CTS_Timeout DIFS 1) RTS/CTS access mode
2) Basic access mode
DATA
SIFS
ACK
DIFS
DATA
SIFS
ACK_Timeout
DIFS
Successful Transmission
Collision Transmission
Successful Transmission
Collision Transmission
Fig.2-10: RTS/CTS access mode and Basic access mode
Chapter 3 Related Work
Chapter 3 Related Work
With the rapid deployment of the IEEE 802.11 WLANs, there are many studies of contention-based DCF medium access function. In order to reduce the collision probability, the DCF applies a collision avoidance mechanism called backoff procedure. Most of the studies are supposed a saturated WLAN. The saturated throughput or channel utilization are the maximum load that the system can carry in a saturated condition. The definition of a saturated WLAN can be found in the papers [4]
[8] [9]. This basic performance figure indicates the limit throughput when the offered traffic load increases.
In the paper [4], it presents an analytic model for computing the capacity of an infrastructure IEEE 802.11 WLAN enhanced with the support of the bidirectional MAC frame aggregation. The analytic model helps us to understand the performance gain of bidirectional aggregation and serve as the foundation for the future aggregation scheduler development.
In the paper [5], this paper uses an analytical model to study the channel capacity when using the basic access (two-way handshaking) method in this analysis. The important contribution in this paper is that it provides closed-form approximations for collision probability p, the maximum throughput S and the limit on the number of stations in a wireless cell. p and S depend on the minimum window size W and the number of stations n only through a gap g = W/(n-1). Consequently, halving W is like
doubling n. The maximum contention window size has minimal effect on p and S. The choice of W that maximizes S is proportional to the square root of the packet length.
The results of this paper can suggest guidelines on when and how W can be adjusted to suit the measured traffic.
In the paper [14], the author first considers the enhanced DCF access method of IEEE 802.11e. The analytical model can be used to calculate the traffic priority and throughput corresponding to the configuration of multiple DCF contention parameters under the saturated WLAN.
In the paper [15], it provides an analytical model to evaluate the saturation throughput of the IEEE 802.11e EDCA. The analytical model is based on the use of the mean value analysis. It also models accurately the effects of the change of the contention window size and Arbitration Interframe Space (AIFS). This model is applicable to real-time system tuning and on-line admission control algorithms that need a low computation complexity.
In the paper [14], most features of the EDCA such as virtual collision, different arbitration interframe space (AIFS) and different contention window are considered.
The throughput and mean delay of differentiated service traffics are analyzed with using the Markov chain model Fig.3-1.
Chapter 3 Related Work
Fig.3-1 Transition diagram of discrete time Markov chain model for one AC per station
Chapter 4
Unidirectional Traffic Flow
To reduce the complexity of analysis, the saturated WLAN [4] [8] [9] is considered in the following sections. A saturated WLAN has the following properties:
(1) All stations and the AP always have a nonempty queue of data frames to transmit.
(2) The traffic distribution to all stations from the AP is uniform. Its meaning is that each station has the same probability to be the destination when the AP wants to send a frame.
(3) The AP always has at least one frame destined to each station waiting in its queue.
Suppose there are N-1 stations and one AP in the saturated WLAN. They have the equal probability 1
N to contend for the channel successfully. Fig.4-1 and Fig.4-2 show the scenario of the unidirectional traffic flow. If one station sends an aggregated data frame to the AP. The AP will return the Ack frame to the AP. This is called unidirectional traffic flow. In this thesis, we focus on the case there is at most one aggregated frame as shown in Fig.4-2. Multiple MAC frames carried in a single physical frame is called aggregation or an aggregate. RTS/CTS-based contention access can reduce the collision cost. The channel access in the saturated WLAN is contention-based, e.g. distributed coordination function (DCF) and enhanced DCF (EDCF) as in [10] [11]. Centrally-controlled channel access, e.g. Point Coordination Function (PCF) and HCF Controlled Channel Access (HCF) as in [12] [13] is not considered.
Chapter 4 Unidirectional Traffic Flow
Fig.4-1: WLAN Scenario – Unidirectional Traffic flow
Fig.4-2: Unidirectional Traffic flow with aggregation
4-1 Our Method for Calculating the Channel Utilization
If a collision happens in the channel, the channel will pay for this collision as shown in
channel collision EmptySlots OurAnalysis RTS CTS
T = W × +T + T +SIFS+ τ +DIFS (4.1)
RTS_timeout CTS 2
where T =T +SIFS+ τ
Fig. 4-3 Tchannel_collision
If a successful transmission happens in the channel, the channel will pay for this successful transmission as shown in
_ channel success
T Fig. 4-4.
_ _ 20 _ 3 4
channel success EmptySlots OurAnalysis RTS CTS phy data ACK
T =W × +T +T +T +T + SIFS+ τ DIFS
Chapter 4 Unidirectional Traffic Flow
Fig. 4-4 Tchannel_successfor unidirectional traffic flow
channel
p is the probability that a collision happens to the medium. is the probability that a collision happens to a station. The channel utilization can be represented as below:
channel channel collision channel channel success
p T following sections. From the result in [5] and the equation (4.21),
channel
In the paper [5], the rsuccess is the rate of successful packet transmissions and the is the rate of packet transmissions (including packet collisions). Then the average number of transmission per packet is
rxmit
Suppose one channel_collision_num contains two packet_collision_num . So the
rate of channel collisions rcollision is given by
(4.6) rxmit −rsuccess =2rcollision
cycle
T is the time between two payload transmissions – and consist of successful and collided transmissions.
(4.7) 1
From equations (4.4)-(4.6), we can get a conclusion
(4.8)
Next we will explain how to estimate the average value ( ) of the backoff time between two transmissions in the medium in the chapter 4-2.
EmptySlots
Chapter 4 Unidirectional Traffic Flow
4-2 The Proposed Method of Paper [4] for Calculating the Channel Utilization
In this paper [4], the channel utilization can be expressed as below:
(4.12) 1
v paper
µ = t
Where is the average period between two successful transmissions, which is defined in [9]. It is also called and is the average data frame size (MAC data frame size in this thesis) successfully transmitted during
.
tv
virtual transmission time mv
tv Fig.4-5 shows the concept of a virtual transmission time.
Fig.4-5 a virtual transmission time
One STA contends for the channel successfully
_ _ 1
TTXOP is the time duration for a TXOP in a virtual transmission time.
(4.14) TTXOP =TRTS +TCTS +Tphy_data +TACK + ×3 SIFS+ ×4 τ
where Tphy_data is the average transmission time for a physical data frame
i
T T SIFS T
TColl is the duration of the i-th collision in a virtual transmission time.
(4.15) T T where
Chapter 4 Unidirectional Traffic Flow
4-3 How to Calculate WEmptySlots
A station would select the number of slots at random out of an interval between 0 and contention window (CW). The backoff time would be selected from a larger range when a transmission fails. Each time the retry counter increases, the CW moves to the next greatest power of two and the size of CW is like the equation as below:
(4.18) 2 CW = m CWmin+2m−1
where CWmin is the minimum contention window
Define be the average number of backoff slots experienced by a packet until it is transmitted successfully or discarded in the saturated WLAN with unidirectional traffic flow
1 1
where K is the maximal retransmission number
Define be the average transmissions experienced by a packet until it is transmitted successfully or discarded in the saturated WLAN with unidirectional traffic flow.
W = x , we get the average time that a station makes a transmission.
(4.21) 1 1
Chapter 4 Unidirectional Traffic Flow
Next we will discuss three types of WEmptySlots as below:
4-3-1 The Proposed Method of Paper [4]
In the long term, the AP and each STA will get a fair share of channel accesses for transmitting their frames, and in particular, the average period of virtual transmission time for the AP and each STA is N × . In this paper [4], it concludes a conclusion that the total empty slots in a virtual transmission time as below:
tv
Random variable is the number of collisions in a virtual transmission time and it is a geometrical distribution. So its mean can be calculated as below:
Nc
The backoff slots between two transmissions in the medium can be calculated from the equation as below:
(4.26)
4-3-2 The Proposed Method of Paper [5] [6]
At a saturated WLAN, most transmission are preceded by a minimum backoff of
; when N stations uniformly choose a time in , the separation between choices has mean
CWmin CWmin
min
1 CW
N+ . In particular, the station that picks the earliest slots breaks the channel silence after the time min
1 CW
N + . So we can express the equation as below:
(4.27)
min
_ 2
W = N
EmptySlots paper 1
CW +
Chapter 4 Unidirectional Traffic Flow
4-3-3 Our Analytic Method
Every station will make a transmission for every in average. After the DIFS time interval, some station must count down its backoff counter from . Please see the
unith W
Wuni
Fig.4-6. Suppose other stations will count down their backoff counter form r.v.
{
X X1, 2,",XN−1}
. We model{
X X1, 2,",XN−1}
by Uniform distribution from [0,Wuni]. Let Y = min{
X X1, 2,",XN−1}
and Y means the backoff slots between two transmissions in the medium. Its cumulative distribution function, , can be calculated as below:Its probability density function, fY( )y , can be calculated as below:
(4.29)
The mean of Y can be calculated as below:
Therefore, the backoff slots between two transmissions in the medium as below:
(4.31) EmptySlots OurAnalysis_ uni
Chapter 5 Bidirectional Traffic Flow
Chapter 5
Bidirectional Traffic Flow
Fig.5-1 and Fig.5-2 show the scenarios of the bidirectional traffic flow. In the scenario of the unidirectional traffic flow, if a frame is transmitted successfully, the backoff counter will be reset. But in the scenario of the bidirectional traffic flow, when a frame is transmitted successfully, the backoff counter may not be reset. The key point is if the frame in the front of the buffer or not. There are examples as below.
(1) One station sends an aggregated data frame to the AP. The AP will return the Ack+AggregatedData frame to the station. If the AggregatedData frame is at the head of buffer in the AP, the backoff counter of AP will be reset by the piggyback. If not, the counter will not be reset.
(2) The AP sends an aggregated data frame to some station. The station will return the Ack+AggregatedData frame to the AP. The backoff counter of the station must be reset by the piggyback.
Fig.5-1: WLAN Scenario – Bidirectional traffic flow
Fig.5-2: Bidirectional traffic flow with aggregation
Chapter 5 Bidirectional Traffic Flow
5-1 Our Method for Calculating the Channel Utilization
If a collision happens in the channel, the channel will pay for this collision. The is the same as the
If a successful transmission happens in the channel, the channel will pay for this successful transmission as shown in
_
channel success EmptySlots OurAnalysis RTS CTS phy up data
phy down data ACK
T W T T T
T T SIFS τ DIFS
= × + + +
+ + +
Fig.5-3 Successful transmission state for bidirectional traffic flow
channel
p is the probability that a collision happens to the medium. is the probability that a collision happens to a station. The channel utilization can be represented as below:
channel mac up data mac down data OurAnalysis
channel channel collision channel channel success
mac up data mac down data
p T T
TRTS,TCTS,TACK,SIFS DIFS and , τ are deterministic values. Please reference the table 1 in the chapter 6. Next and will be calculated as the following sections. From the result in [5] and the equation (5.17),
channel
p WEmptySlots
p satisfies
(5.2) 1 1
1 (1 )
p N
Wbi
= − − −
Similar to the evaluation of (4.11), pchannel is
(5.3) 2
collision channel
success collision
r p
p = =
r +r − p
Chapter 5 Bidirectional Traffic Flow
5-2 The Proposed Method of Paper [4] for Calculating the Channel Utilization
Fig.5-4 shows a virtual transmission time. The evaluation is the same as the chapter 4.
Fig.5-4 A virtual transmission time
So tv can be expressed as below:
TTXOP is the time duration for a TXOP in a virtual transmission time.
(5.5)
TXOP RTS CTS up down ACK
T T
Colli
T is the duration of the i-th collision in a virtual transmission time.
(5.6)
From the result of [4], tv can be expressed as below:
where and are the average transmission time for a physical data frame in the uplink and downlink directions respectively
Tup Tdown
(5.8) where and are the average transmission time for a MAC data frame in the uplink and downlink directions respectively
_ mac up
T Tmac_down
The utilization in [4] can be expressed as:
where p satisfies p N in the
p CW N
Chapter 5 Bidirectional Traffic Flow
5-3 How to Calculate WEmptySlots
5-3-1 Our Analytic Method
Suppose there are stations (including the AP) in the saturated WLAN. They have the equal probability
N 1
N to contend for the channel successfully. Fig.5-5shows a possible transmission pair in the saturated WLAN that supports bidirectional traffic flow. represents that the AP contends for the channel successfully to send a frame to and will return a Ack+AggregatedData frame to the AP. So the backoff counter of the can be reset by the piggyback.
represents that contends for the channel successfully to send a frame to the AP and the AP will return a Ack+AggregatedData frame to the . If this frame is in the head of the buffer, the backoff counter of the AP can be reset by the piggyback..
Fig.5-5 A possible transmission pair in the saturated WLAN that supports bidirectional traffic flow for the AP
Next we will extend the concept from the unidirectional traffic flow to the bidirectional traffic flow. is the average number of backoff slots experienced by a packet until it is transmitted successfully or discarded in the saturated WLAN with unidirectional traffic flow. From the equation (4.19) and
Wx
Fig. 5-5, we approximate the
time interval that the AP contends for the channel successfully by . Next we will discuss the average time interval after which the backoff counter of the AP can be reset through the piggyback.
Wx
So the average time interval after which the backoff counter of the AP is reset by
So the average time interval after which the backoff counter of the AP is reset by