通訊所 一年級 693430028 鄭筱親

A Survey of Quality of Service in IEEE 802.11 Networks

IEEE 802.11E: QoS Provisioning at the MAC Layer

通訊所 一年級 693430028 鄭筱親

Abstract

IEEE 802.11 experiences serious challenges in meeting the demands of multimedia services and applications.

IEEE 802.11e standard support quality of service at MAC layer.

The viewpoint 802.11 QoS schemes 802.11e

Outline

Introduction An Overview of IEEE 802.11 802.11MAC QoS Mechanisms IEEE 802.11E Conclusion and Future Work

Introduction(1/2)

WLANs are becoming ubiquitous and increasingly relied on 802.11 。

Wireless users can access real-time and Internet services virtually anytime, anywhere.

In wireless home and office networks, QoS and multimedia support are critical.

QoS and multimedia support are essential ingredients to offer VOD 、 audio on demand and high-speed Internet access.

Introduction(2/2)

The lack of a built-in mechanism for support of real time services makes it difficult to provide QoS guaranteed for throughput-sensitive and delay-sensitive multimedia applications.

IEEE 802.11e is being proposed as the upcoming standard for the enhancement of the vice differentiation.

An Overview of IEEE 802.11Task Group Responsibility

802.11a—OFDM 5GHz 54Mbs

802.11b—HR/DSSS 2.4GHz 22Mbs

802.11c—Bridge Operation Procedures Bridge

802.11d—Global Harmonization Additional regulatory domains

802.11e—MAC Enhancements for QoS EDCF 、 HCF

802.11f—Inter Access Point Protocol Interoperability

802.11g—OFDM 2.4GHz 36/54Mbs

802.11h—DFS Dynamic channel selection

802.11i—security WEP

802.11MAC (1/4)

MACExtent

免競爭式服務(具時限傳輸 )

Distributed Coordination Function (DCF)

競爭式服務(非同步傳輸 )

Point Coordination Function (PCF)

802.11MAC (2/4)

Distributed Coordination Function (DCF) Defines a basic access mechanism and optional RTS/

CTS mechanism. Shall be implemented in all stations and APs. Used within both ad hoc and infrastructure configuratio

ns. Point Coordination Function (PCF)

An alternative access method Shall be implemented on top of the DCF A point coordinator (polling master) is used to determin

e which station currently has the right to transmit. Shall be built up from the DCF through the use of an ac

cess priority mechanism

802.11MAC (3/4) Different accesses to medium can be defined through

the use of different values of IFS (inter-frame space). PCF IFS (PIFS) < DCF IFS (DIFS) PCF traffic should have higher priority to access the m

edium, to provide a contention-free access. This PIFS allows the PC (point coordinator) to seize co

ntrol of the medium away from the other stations. Coexistence of DCF and PCF

DCF and PCF can coexist through superframe. superframe: a contention-free period followed by a cont

ention period.

免競爭訊框

超級訊框

需競爭訊框

802.11MAC (4/4)

Figure ： Coexistence of DCF and PCF

Distributed Coordination Function (1/3)

Allows sharing of medium between PHYs through CSMA/CA random backoff following a busy medium.

All packets should be acknowledged (through ACK frame) immediately and positively. Retransmission should be scheduled immediat

ely if no ACK is received.

Distributed Coordination Function (2/3)

Carrier Sense shall be performed through 2 ways: physical carrier sensing: provided by the PHY virtual carrier sensing: provided by MAC

by sending medium reservation through RTS and CTS frames

duration field in these frames The use of RTS/CTS is under control of RTS_Thresho

ld. An NAV (Net Allocation Vector) is calculated to estima

te the amount of medium busy time in the future. Requirements on STAs:

can receive any frame transmitted on a given set of rates

can transmit in at least one of these rates This assures that the Virtual Carrier Sense mechanism

work on multiple-rate environments

Distributed Coordination Function (3/3)

MAC-Level ACKs Frames that should be ACKed:

Data Poll Request Response

An ACK shall be returned immediately following a successfully received frame.

After receiving a frame, an ACK shall be sent after SIFS (Short IFS).

SIFS < PIFS < DIFS So ACK has the highest priority

DCF: the Random Backoff Time (1/2) Before transmitting asynchronous MPDUs, a STA sh

all use the CS function to determine the medium state.

If idle, the STA defer a DIFS gap transmit MPDU

If busy, the STA defer a DIFS gap then generate a random backoff period (within the cont

ention window CW) for an additional deferral time to resolve contention.

DCF: the Random Backoff Time (2/2)Backoff time = CW* Random() * Slot timewhere CW = starts at CWmin, and doubles after each failure until reaching CWmax and remains there in all remaining retries

(e.g., CWmin = 7, CWmax = 255)

Random() = (0,1)Slot Time = Transmitter turn-on delay +

medium propagation delay + medium busy detect response time

8

CWmax

CWmin7

1531

第二次重送第一次重送

第三次重送初始值

63127127

255255 255255

Duration Reservation Strategy (1/2)

Each Fragment and ACK acts as a “virtual” RTS and CTS for the next fragment.

The duration field in the data and ACK specifies the total duration of the next fragment and ACK.

The last fragment and ACK will have the duration set to zero.

Duration Reservation Strategy (2/2)

Goal of fragmentation: shorter frames are less suspectable to transmission err

ors, especially under bad channel conditions

Point Coordination Function (1/6)

The PCF provides contention-free services. One STA will serve as the Point Coordinator (PC), wh

ich is responsible of generating the Superframe (SF). The SF starts with a beacon and consists of a Contenti

on Free period and a Contention Period. The length of a SF is a manageable parameter and tha

t of the CF period may be variable on a per SF basis. There is one PC per BSS.

This is an option; it is not necessary that all stations are capable of transmitting PCF data frames

Point Coordination Function (2/6)

The PC first waits for a PIFS period. PC sends a data frame (CF-Down) with the C

F-Poll Subtype bit = 1, to the next station on the polling list.

When a STA is polled, if there is a data frame (CF-Up) in its queue, the frame is sent after SIFS with CF-Poll bit = 1.

Then after another SIFS, the CF polls the next STA.

This results in a burst of CF traffic. To end the CF period, a CF-End frame is sent.

Point Coordination Function (3/6)

If a polled STA has nothing to send, after PIFS the PC will poll the next STA.

NAV setup: Each STA should preset it’s NAV to the maxim

um CF-Period Length at the beginning of every SF.

On receiving the PC’s CF-End frame, the NAV can be reset (thus may terminate the CF period earlier).

Point Coordination Function (4/6)

NAV

SIFS

SIFS

媒介忙碌中 CF-D1

CF-U1

CF-D2

SIFS

CF-U2

CF-D3

SIFS

PIFSSIFS

CF-D4

CF-U4

PIFS

SIFS

CF-End

免競爭週期

超級訊框

競爭週期

重設 NAV

CF- 邊界

Dx = Down TrafficUx = Up Traffic

Point Coordination Function (5/6)

When the PC is neither a transmitter nor a recipient: When the polled STA hears the CF-Down:

It may send a Data frame to any STA in the BSS after an SIFS period.

The recipient (.neq. PC) of the Data frame returns an ACK after SIFS.

Then PC transmits the next CF-Down after an SIFS period after the ACK frame.

If no ACK is heard, the next poll will start after a PIFS period

Point Coordination Function (6/6)

NAV

SIFS

媒介忙碌中 CF-D1

S-To-S

SIFS

ACK

CF-D2

SIFS

PIFS SIFS

CF-U2

SIFS

CF-End

免競爭週期

超級訊框

競爭週期

重設 NAV

CF- 邊界

Dx = Down TrafficUx = Up Traffic

QoS Mechanisms

QoS mechanisms for 802.11 can be classified into three categories: Service differentiation Admission control and bandwidth reservation Link adaptation

BETTER THAN BEST EFFORT SCHEMES:SERVICE DIFFERENTIATION (1/3) Enhanced DCF (EDCF)

prioritizes traffic categories by different contention parameters, including

arbitrary interframe space (AIFS), maximum and minimum backoff window size (CWmax/min), and a multiplication factor for expanding

the backoff window.

Persistent Factor DCF (P-DCF) each traffic class is associated with a persistent factor P a uniformly distributed random number r is generated in

every slot time Each flow stops the backoff and starts transmission only

if (r > P)

BETTER THAN BEST EFFORT SCHEMES:SERVICE DIFFERENTIATION (2/3) Distributed Weighted Fair Queue (DWFQ)

the backoff window size CW of any traffic flow is adjusted based on the difference between the actual and expected throughputs.

a ratio (Li′ = Ri/Wi) is calculated, where Ri is the actual throughput and Wi the corresponding weight of the ith station.

Distributed Fair Scheduling (DFS) differentiate thebackoff interval (BI) based on the pack

et length and traffic class For the ith flow, BIi = ρi × scaling × factor × Li/ϕi,

Distributed Deficit Round Robin (DDRR) the ith throughput class at the jth station is assigned wit

h a service quantum rate (Qi,j) equal to the throughput it requires

BETTER THAN BEST EFFORT SCHEMES:SERVICE DIFFERENTIATION (3/3)

QOS MECHANISMS FOR ADMISSION CONTROL AND BANDWIDTH RESERVATION (1/2)

Measurement-based approaches Calculation-based approaches Scheduling and reservation-based approache

s

QOS MECHANISMS FOR ADMISSION CONTROL AND BANDWIDTH RESERVATION (2/2)

QOS MECHANISM FOR LINK ADAPTATION (1/2) Received signal strength (RSS) PER-prediction MPDU-based link adaptation Link adaptation with success/fail (S/F) thresh

olds Code Adapts To Enhance Reliability (CATER)

QOS MECHANISM FOR LINK ADAPTATION (2/2)

IEEE 802.11E

Main new features of 802.11e: The Enhanced DCF THE CONTROLLED HCF

The Enhanced DCF (1/2)

The Enhanced DCF (2/2)

DISTRIBUTED ADMISSION CONTROL FOR EDCF TXOPBudget[i]

=Max(ATL[i] – TxTime[i]*SurplusFactor[i],0) If TXOPBudget[i] = 0

–TxMemory[i] shall be set to zero

all other QSTAs TxMemory[i] remains unchanged If the TXOPBudget[i] >0

–TxMemory[i] = f*TxMemory[i] + (1 – f)*

(TxCounter[i]*SurplusFactor[i] + TXOPBudget[i])

–TxCounter[i] = 0

–TxLimit[i] = TxMemory[i] + TxRemainder[i]

THE CONTROLLED HCF

Controlled channel access function allows reservation of transmission opportuniti

es (TXOPs) with a hybrid coordinator (HC) a type of PC handling rules defined by the HC

F

ADMISSION CONTROL ANDSCHEDULING FOR THE CONTROLLED HCF

The behavior of the scheduler is as follows: The scheduler shall be implemented if a traffic stream is admitted by the HC, the sc

heduler shall send polls anywhere between the minimum service interval and the maximum service interval within the specification interval.

PRELIMINARY RESULTS

Conclusion and Future Work

Traffic Differentiation 將資料分類並規範其優先順序 HCF 結合 DCF 和 PCF

Block Acknowledgement 無線環境下，資料發送出去，需接收端發出回應 資料傳送達一定量之後，接收端再統一送出一個 Summar

y 的確認訊息 Direct Link Protocol (DLP)

當兩個 Station 要直接互通時，必需確定雙方均有此項功能，即可直接互通

但同時擁有 Infrastructure 的模式。

看法及感想 QoS 是網路的一種功能，能夠對特定的網路流量，提供較佳的服務

由於一個傳輸頻率在一個時間與一個特定地點內僅能被一個站點所使用，因此須透過 CSMA程序將資源儘可能公平地分配給所有站點使用

每個站點可透過網路接收到同等的資料量，但資料量等級不僅沒有可靠的保證，而且傳輸流量在短期也不平均

QoS 希望能達到增加 Throughput 、降低 latency 的目的