1

Intelligent Wireless Local Area Networking

Qualifying Exam

Mustafa Ergen

2

• Degrees

– BS: Middle East Technical University, 2000– MS: University of California Berkeley, 2002

• Selected Publications

1. Mustafa Ergen, Pravin Varaiya, “Admission Control and Throughput Analysis in IEEE 802.11,” ACM-Kluwer MONET Special Issue on WLAN Optimization at the MAC and Network Levels.

2. Mustafa Ergen, Sinem Coleri, Pravin Varaiya “QoS Aware Adaptive Resource Allocation Techniques for Fair Scheduling in OFDMA Based Broadband Wireless Access Systems,” IEEE Transactions on Broadcasting, Vol.:49: Dec. 2003

3. Mustafa Ergen, Duke Lee, Ruchira Datta, Jeff Ko, Anuj Puri, Raja Sengupta, Pravin Varaiya, “Comparison of Wireless Token Ring Protocol with IEEE 802.11,” Journal of Internet Technology, Vol. 4 No. 4.

4. Sinem Coleri, Mustafa Ergen, Anuj Puri, Ahmad Bahai, “Channel Estimation Techniques Based on Pilot Arrangement in OFDM Systems,” IEEE Transactions on Broadcasting VOL. 48, NO. 3 September 2002, pp 223-229.

5. Xuanming Dong, Mustafa Ergen, Pravin Varaiya, Anuj Puri “Improving the Aggregate Throughput of Access Points in IEEE 802.11 Wireless LANs”, IEEE WLN, Bonn, Germany, October, 2003.

6. Mustafa Ergen, Duke Lee, Raja Sengupta, Pravin Varaiya “Wireless Token Ring Protocol-performance comparison with IEEE 802.11,” IEEE ISCC, Antalya, Turkey, July 2003. *Received Best Student Paper Award*

7. Sinem Coleri, Mustafa Ergen, Tak-Kuen John Koo, “Lifetime Analysis of a Sensor Network with Hybrid Automata Modeling,” ACM WSNA Atlanta, September 2002.

8. Mustafa Ergen, Anuj Puri, “MEWLANA-Mobile IP Enriched Wireless Local Area Network Architecture,” IEEE VTC, Vancouver September, 2002.

9. Mustafa Ergen, Sinem Coleri, Baris Dundar, Rahul Jain, Anuj Puri, Pravin Varaiya, “Application of GPS to Mobile IP and Routing in Wireless Networks,” IEEE VTC, Vancouver, Canada, September, 2002.

10. Sinem Coleri, Mustafa Ergen, Anuj Puri, Ahmad Bahai, “A Study of Channel Estimation in OFDM Systems,” IEEE VTC, Vancouver, Canada, September, 2002.

11. Mustafa Ergen, Sinem Coleri, Baris Dundar, Anuj Puri, Jean Walrand, Pravin Varaiya, “Position Leverage Smooth Handover Algorithm For Mobile IP,” IEEE ICN Atlanta, August, 2002.

12. Duke Lee, Sinem Coleri, Xuanming Dong, Mustafa Ergen, “FLORAX- Flow-Rate Based Hop by Hop Back-pressure Control for IEEE 802.3x,” IEEE HSNMC Jeju Island Korea July, 2002.

3

Outline• Introduction to IEEE 802.11• 4 Markov models of DCF• Throughput Analysis• Different data rates• Unsaturated Traffic• Application: Admission Control• Application: Indoor Throughput• Next Generation WLANs

– Adaptive Antenna

– Multi-hop Networking

– Positioning

• Conclusion

4

Contribution

• Joint Markov Model• 802.11+ Model• Unsaturated Model• Individual Throughput with Different Data Rates• 802.11a Performance Analysis• Admission Control• Indoor Throughput

5

Introduction to IEEE 802.11

6

802.11 MAC Meta-States

BackoffIdle

TxSequence

& Retry

BusyDuring Tx

Medium not busyduring Tx attempt

Finish Tx

Still in sequenceand last step successful

Pre-Tx backoffsuccessful

Just TransmittedAck or CTS

All other transmitted frameswhether successful or not

Post-Tx backoff successful

PCSVCSWait

Idle forIFS time

Busy during backoff

7

Idle Procedure

PAV ("lastPCSBusyTime")NAV ("lastVCSBusyTime")currentTimepacketToSendNote: PAV = (lastPHY_CCA == IDLE) ? lastPHY_CCATime : currentTime

System Fields:

Queueempty?

LLC or MAC

MAC Packet Queue

PCSVCSWait

currentTime >MAX(PAV, NAV)

Tx

YES

NO

YES

NO

Packet Add Trigger

Packet size >RTSThreshold &&

FragNum == 0

packetToSend =RTS

packetToSend =dequeued data

packet

YES

NO

8

Backoff Procedure

Enter backoff BC =Rand() & CW

CW = 2N-1

Wait 1TS

PAVNAVcurrentTimeBC (Backoff counter)TS = 1 slot time = 20 (802.11b), 9 (802.11a)

System Fields:

MAX(PAV, NAV)< currentTime - TS

YES

YES

Leave backoff

NOPCSVCSWait

NOBC == 0? BC == 0?

BC--

NO

Idle for IFS TimeEnter backoff

YES

9

VCS: NAV Update Procedure

Packet isRTS?

currentIFSTimelastRxStartTimelastRxEndTimecurrentTimeNAVT = 2*aSIFSTIme + CTSTime + 2*aSlotTime

currentTime +Packet Duration >

NAV

UpdateNAV

Countdown on

T

lastRxEndTime >lastRxStartTime

PacketCorrect?

currentIFSTime= EIFS

NO

YES

currentIFSTime= DIFS

NAV = currentTime+ Packet Duration

Expired

currentTime -lastRxEndTime >= T

YES

YES

YES

PHY_RXEND.ind

NAV =currentTime

STA ispacket

addressee

NO

System Fields:

lastRxStartTime= currentTime

lastRxEndTime= currentTime

PHY_RXSTART.ind PHY_RXEND.ind

PHY_CCARESET.req

Packetneeds Ack? Tx

WaitSIFS

YES

YES

10

Frame Sequence and Retry Procedure

(RTS + CTS) is treated the same as (Data + Ack) with frame length < aRTSThreshold

PHY_TXEND.conf

Last frameneeds Ack?

AckTimer

YES ReceivePHY_RXSTART.ind

before timeout

Waitframeend

ReceivePHY_RXEND.ind Valid Ack?

Single-castdata or RTS

YES

CW = aCWminSRC = 0 (LRC = 0 if

frame len >aRTSThreshold)

CW =MAX(CW*2+1,

aCWmax)SRC++ (or LRC++)

Timeout andand didn't receive

PHY_RXSTART.ind

NO

SRC (or LRC)limit reached?

Backoff

Discard frameCW = aCWmin

SRC (or LRC) = 0

NO

YES

PHY_RXSTART.ind PHY_RXEND.ind

Further Txsequence

TxWaitSIFS

YES

Retransmission

NO

Packet Fragmentsor RTS+CTS+Data

YES

11

IEEE 802.11 DCF

Saturation Throughput

Time Scale of DCF Function

• Observation time • Determination of discrete events• Construction of Markov model• Saturation throughput

OPNET Simulation

•FHSS•1Mbps Channel•Saturation Throughput•Packet Size 1000bytes•Inter-arrival time 0.005•Load 1.6 MbpsDIFS

time

w/o RTS/CTS

RTS/CTS

EIFS

SIFS SLOT

12

4 Markov Models of DCF

13

Joint Model

All stations are dependent

14

Independent Model

Each station has its own independent channel, but with same parameter p(n)

15

Markov Model Analysis

0a 1a 2a 3a

0a 1a 2a 3a

1/4

1/4

1-p

p

1-p

p

1-p

p

802.11b

802.11+

Case I

Case II

Probability of Tx after/before Tx

Assumption:•Saturation Throughput•Limitless Retry •Everybody hears everybody•Case I: No consecutive Transmission

CWmin=16CWmax=1024 1/20

+b

16

Probability of Transmission of a STAp: Probability of Channel Busyn: Number of Stations

Ptr: Probability of Transmission in Medium

Ps: Probability of Successful Transmission in Medium

EP: Packet Size

Ts: Duration of Successful Transmission

Tc Duration of CollisionDuration of Slot TimeS: Throughput

0a 1a 2a 3a

0a 1a 2a 3a

1/4

1/4

1-p

p

1-p

p

1-p

p

802.11b

802.11+

cstrsstr

s

TPPTPP

EPPS

)()1(

][

][

eventvirtualoflengthE

payloadEThroughput

EP: Packet Size

Ts: Duration of Successful Transmission

c Duration of CollisionDuration of Slot Time

Ptr: Probability of Transmission

Ps: Probability of Successful Transmission

from model

given by the PHY layer

17

Independent Markov Model

Definition

b0a: Probability of being in state 0a

Transmission occurs if STA is at 0a

p: Channel Busy if there is one station at 0a but me

Ptr: Transmission if there is at least one STA at 0a

Ps: Successful Transmission if there is one STA at 0a

0a 1a 2a 3a

0a 1a 2a 3a

1/4

1/4

1-p

p

1-p

p

1-p

p

802.11b

802.11+

•No freeze in backoff

•Freeze in backoff

1

1

11

)1(

)1(1

0))1(1()()1(1

ns

ntr

nn

nP

P

ppp

4.0

p

pp

b a

25

22)(

0

Assumption:

constant and independent collision probability

18

Joint Markov Model

• n=2 • STA a and STA b• # states = 4n

• Dependent STAs in 802.11+

• Ptr=p0a0b+p0a1b+p0a2b+p0a3b+p1a0b+p2a0b+p3a0b :At least one Zero State

• Ps=p0a1b+p0a2b+p0a3b+p1a0b+p2a0b+p3a0b: Only one Zero State

2a0b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

2a1b 1

2a2b 1

2a3b 1

3a0b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

3a1b 1

3a2b 1

3a3b 1

0a0b 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16 1/16

0a1b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

0a2b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

0a3b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

1a0b 1/4 1/4 1/4 1/4 1/4 1/4 1/4 1/4

1a1b 1

1a2b 1

1a3b 1

0a0b 0a1b 0a2b 0a3b 1a0b 1a1b 1a2b 1a3b 2a0b 2a1b 2a2b 2a3b 3a0b 3a1b 3a2b 3a3b

802.11b802.11+

19

Throughput analysis

20

One level backoff

OPNET Simulation:

Ptr= ( #Total ACK rcvd + #Collision)/( #Back-off slot+ #Total ACK rcvd+ #Collision)

Ps= (#Total ACK rcvd)/(#Back-off slot + #Total ACK rcvd + #Collision)

Verification of the simulation

Simulation Time = SLOT * #Back-off slot+ Ts* #Total ACK rcvd+ Tc #Collision)

•Independent of access mechanism•Independent of PHY layer•FHSS used

•802.11+ Joint Markov Model exactly approximates Simulation

21

Throughput

T+s=Tdata+SIFSTack+DIFS+

T+c=TdataEIFS

Tbs=Tdata+SIFSTack+DIFSSLOT

Tbc=TdataDIFS+SLOT

T+s=Trts+SIFSTcts+SIFSTdata+SIFSTack+DIFS+

T+c=TrtsEIFS

Tbs=Trts+SIFSTcts+SIFSTdata+SIFSTack+DIFS+SLOT

Tbc=TdataDIFS+SLOT

Basic Access Mechanism

RTS/CTS Access Mechanism

)1()( trs

cstrss PS

EPPTPPTP

Verification of Duration Values from Simulation

Ts=0.0088secTc=0.0088sec

Ts=0.0087secTc=0.0007sec

802.11+ Joint Markov Model exactly approximatesSimulation

FHSSData Rate 1MbpsSaturation Throughput

for FHSSn

Thr

ough

put M

bps

22

Multi Level Back-off

))2(1()1)(21(

)1)(21(2),,(

))2(1()1)(21(

)21(2),,(

m

mb

ppWWp

ppmWp

ppWWp

pmWp

•FHSS•Data Rate 1Mbps•Saturation Throughput•W=16•m=7•CWmin=16•CWmax=1024•Retry Count = 255

Basic

Ts=0.0088sTc=0.0088s

RTS/CTS

Ts=0.0090sTc=0.0007s

m

ii p

bb

0

0,00, )1(

b

+

23

Different (Mixed) data rates

24

Individual Throughput with Different Data Rates

• Throughput distributes evenly among STAs

• ni is the number of stations with data rate i

• D is the total number of data rate choices

• E[Ts] is Average• E[Tc] is highest of the STA in

collision][][)1(

1

][][)1(

)1()(][

][

1

1 1 1

111

1

1

cstr

si

cstr

s

n

i

D

j

n

k

inijc

j

l

l

c

D

i

is

iss

TETEP

EPP

nS

TETEP

EPPS

Ti

nknTE

Tnn

PTE

j

NDStation ID R1 R2 R1 R3 R4 R1 R4 R1 ] :Data Rates[Ts

1 Ts2 Ts

1 Ts3 Ts

4 Ts1 Ts

4 Ts1] :Succ. Dur.

[Tc1 Tc

2 Tc1 Tc3 Tc

4 Tc1 Tc

4 Tc1] :Coll. Dur.

n1=4, n2=1, n3=1, n4=2

25

Verification in 802.11bSimulation Scenario

Start: 5 Stations with 1 Mbps Data RateStep: 1 Station shift to 11 MbpsStop: 5 Stations with 11 Mbps Data Rate

Throughput Individual Throughput

Throughput of all stations isthe same!

26

Unsaturated case

27

New Model: Unsaturated Traffic

• Modifications

• Operation in non-saturated load

• Different Data Rates,

• Modified in IEEE 802.11a

)11

)(1()1)(21(2

))2(1()1)(21(

1

2

p

ppppWWp m

Traffic Intensity

28

802.11a OFDM Packet Format

sBpS

sT

sBpS

sT

sBpS

sT

sBpS

EPsT

ACK

CTS

RTS

DATA

4.8/)616(14

20

4.8/)616(14

20

4.8/)616(20

20

4.8/)616(28

20

29

Probability of CollisionProbability of Transmission

Throughput with RTS/CTS DR=54MbpsThroughput w/o RTS/CTS DR=54Mbps

30

AnalysisThroughput with Different Data Rates not mixed

Throughput with Constant total load n

Throughput with Offered Load DR=54Mbps

31

Admission Control

32

Throughput fixed station same SNR Fairness Constraint

0

)(

min)(max)(

.

},...,1{)())((max

)()(

01)(

],1[

],,1[

)(

K

selectedSTAsofIsubsetofthroughputIS

where

xtxtC

ts

NItKCtIS

sxtx

ortx

Ni

Endtimet

for

Totali

i

Totali

i

tI

txi

Totali

i

Time=N Time=N

33

Admission Control: w/o Mobility

Total Throughput Individual Throughput

Data Rates are fixedWithout RTS/CTS0.2With RTS/CTS gap will be smaller

34

Admission Control: w/o MobilityProbability of being selected Number of Stations selected at time t

Fairness Constraint Data Rate vs Throughput

35

Admission Control: with Mobility

Total Throughput Individual Throughput

Data Rates are changed in every iteration

36

Admission Control: with Mobility

Probability of being selected Number of Stations Selected at time t

Fairness Constraint Data Rate vs Throughput

37

Indoor Throughput

38

Indoor ThroughputAccess Point Coverage Determination Signal Power (RSSI Map)

•Access Point Coverage gives the number of•Mobiles attached per AP

Signal Power gives the data rate of each mobile

Client model (power level 1-30mW)Omni-directional antennas

APs (power level 1-100mW)Model the interference between the APs and the mobiles

39

40

PerformanceTotal Throughput: 5 AP Individual Throughput: 5 AP

Data Rate vs Throughput: 5 APThroughput: 50 STAs

41

Future work

42

Intelligent Network

• Problems– Coverage

– Throughput

– Security

– Interference

– Power Efficiency

• Applications– WLAN

– Mesh Networks

– UWB

43

Adaptive Antenna: Infrastructure BSS: Only AP has AA, RTS/CTS/ACK omni directional

• Rate Adaptation Mechanism• Decrement with timeout• Increment with received ACK

• Power: 1mW • 10mW in direction • (45o, 90o, 180o, 360o)• 0.01mW out of direction

44

Adaptive Antenna

Ad hocAll STAs haveAdaptive Antenna

InfrastructureOnly AP have AdaptiveAntenna

45

Multi Hop Networking : Motivation

• The smaller the range the higher the throughput

)()(

)()(

)(

164

)(

4

1

)()1(||

)(||

22

22

nnr

cnMaxT

Wnr

LnnT

nrMax

boundaryatnwhennr

Min

transmitsalsokthennrXXif

jtransmitsithennrXXif

nodesnwithDareaunitofdiskaDefine

Tx

Tx

area

jk

ji

46

Multi Hop Networking: Algorithm

AP

STA3

STA2

STA1

PCFACTIVE

Overlap

ACTIVE

DCFACTIVE

STA4

Operation in PCF

47

Positioning

• Outdoor– GPS– Cellular Networks

• Indoor– WLAN– UWB– Motivation

• Location Aware Applications

• Wireless Security

48

Hybrid Method for Positioning

• Hybrid Method– Achieve the

accuracy of fingerprinting with less data collection effort,

– Error bound,

NiCGAYAgAgAh

CYdatasetsgiven

MjAgYAgYAL

NewhereeAgY

wallLossBAdPAg

reportedisYYY

AatBofstrengthsignalreceivedY

MjBscoordinateknown

Acoordinateunknown

jijjjj

ii

YYjj

YYjo

jjjj

jj

M

ji

j

ojoj

,...,1:)),((1)),((),(),(

),(

,...,1:)),(()),(()(

)2,0(),(

),(log10),(

),...,(

:

,...,1:

^

)(1)(1

100

1

^

49

Conclusion

• Markov Model– Independent Markov Model

– Joint Markov Model

– Different Data Rates

– (Un) Saturated Traffic

– Application: Admission Control

– Application: Indoor Throughput

• Next Generation WLANs– Adaptive Antenna

– Multi-hop Networking

– Positioning

50

Appendix

• 802.11a*• Slot 9• SIFS 16• PIFS 25• DIFS 34• EIFS 96

• 802.11*• Slot 50• SIFS 28• DIFS 128• EIFS 384

sec