Analysis of IEEE 802.11e and Application of Game Models for Support of Quality-of-Service in Coexisting Wireless Networks Stefan Mangold ComNets Aachen.

Analysis of IEEE 802.11e and Analysis of IEEE 802.11e and

Application of Game Models for Application of Game Models for

Support of Quality-of-Service in Support of Quality-of-Service in

Coexisting Wireless NetworksCoexisting Wireless Networks

Stefan MangoldComNets Aachen University

30-June-2003

Stefan Mangold - ComNets Aachen University 2

OutlineOutlineOutlineOutline

IEEE 802.11 wireless LANBrief introduction: Distributed Coordination Function (DCF)

IEEE 802.11e QoS extensionOverview: Enhanced DCF (EDCF)

Achievable throughput with the EDCF

Model for achievable throughput per priority

Result evaluation with WARP2

Overlapping radio networks in unlicensed bandsGame model of competition

Result evaluation with YouShi

Analysis of competition scenario: stability, expected outcomes

Cooperation in repeated games

Conclusions


Motivation of this ThesisMotivation of this ThesisMotivation of this ThesisMotivation of this Thesis

IEEE 802.11 is the dominant radio system for wireless Local Area Networks (LANs):

Today’s Wireless LANs cannot support Quality of Service (QoS)

However, the demand is growing for new applications with QoS requirements

Wireless LANs operate in unlicensed frequency bands, where they have to share radio resources

Problems/Questions:How to support QoS in wireless LANs?

If wireless LANs can support QoS, what level of QoS can be maintained in unlicensed frequency bands?

New methods to support QoS in wireless LANs are developed and evaluated in this thesis.


IEEE 802.11 Wireless LANIEEE 802.11 Wireless LANIEEE 802.11 Wireless LANIEEE 802.11 Wireless LAN

Radio standard for data transport system that covers ISO/OSI layer 1 and 2:

Multiple Physical (PHY) layers: .11/.11a/.11b/.11g

One common Medium Access Control (MAC) layer

Here: IEEE 802.11a PHYOFDM multi-carrier transmission

Up to 54Mbit/s (@PHY)

5 GHz unlicensed bandShared resources

Main Service:MSDU Delivery

Reference model

MLMEmedium accesscontrol sublayer

PLMEPLCP sublayer

SME

PMD sublayer

IEEE 802.11

user

plan

e

man

agem

ent

plane

cont

rol p

lane

logical link controlsublayer

MAC-SAPMLME-

SAP

physical layer

transport layer

network layer

data link controllayer

OSI reference model

1

3

4

2

MLMEmedium accesscontrol sublayer

PLMEPLCP sublayer

SME

PMD sublayer

IEEE 802.11

user

plan

e

man

agem

ent

plane

cont

rol p

lane

logical link controlsublayer

MAC-SAPMLME-

SAP

physical layer

transport layer

network layer

data link controllayer

OSI reference model

1

3

4

2


Distributed Coordination Function (DCF)Listen before talk: CSMA/CA

Binary exponential backoffContention window increases with each retransmission

Received MPDUs (data frames) are acknowledged

Variable frame body sizes (up to 2312 byte)One queue per stationCollisions occur if many stations operate in parallel

Medium AccessMedium AccessMedium AccessMedium Access

CTS

RTS

time

ACKSIFS

DIFS

PIFS

SIFS

Contention Window(counted in slots,

9us per slot, 15 slots in 802.11a)

SIFS

defer access count down as long as medium is idle,backoff when medium gets busy

with 802.11a: slot: 9us SIFS: 16us PIFS: 25us DIFS: 34us

SIFS

DATA

busychannel

CTS

RTS

time

ACKSIFS

DIFS

PIFS

SIFS

Contention Window(counted in slots,

9us per slot, 15 slots in 802.11a)

SIFS

defer access count down as long as medium is idle,backoff when medium gets busy

with 802.11a: slot: 9us SIFS: 16us PIFS: 25us DIFS: 34us

SIFS

DATA

busychannel


IEEE 802.11 Wireless LAN BasicsIEEE 802.11 Wireless LAN BasicsIEEE 802.11 Wireless LAN BasicsIEEE 802.11 Wireless LAN Basics

MAC protocol is distributed (simple and successful)One queue per station (station = backoff entity)

MSDU can be fragmented into multiple MPDUs

RTS/CTS helps to improve efficiency

QoS involves achieving a minimum MSDU Delivery throughput and MSDU Delivery delays not exceeding a maximum limit

Delay variation and loss rate are often considered

IEEE 802.11 Task Group E (TGe) defines QoS mechanisms to be integrated into the legacy 802.11 MAC

This supplement standard is referred to as IEEE 802.11e (here: draft 4.0)

QoS Support in legacy 802.11? no!


Contention-based medium access: EDCFDifferent EDCF parameters per Access Category (AC)

DIFSAIFS[AC]

CWminCWmin[AC]

*) not in current draft standard

802.11e Medium Access: HCF802.11e Medium Access: HCF802.11e Medium Access: HCF802.11e Medium Access: HCF

CTS

RTS

time

ACKSIFS

AIFS[AC=med.]

AIFS[AC=high](=PIFS)

PIFS

AIFS[AC=low]

SIFS

CW[AC=high]

with 802.11a: aSlotTime: 9us SIFS: 16us PIFS: 25us DIFS: 34us AIFSN: 1…10[slots] AIFS: >=PIFS

SIFS

highpriority AC

lowpriority AC

mediumpriority AC

backoff

backoff

busychannel

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

aSlotTime

DCF: Random backoff counter isselected from interval 0...CW.Minimum interframe space is DIFS.Earliest channel access is DIFS.

EDCF: Random backoff counter isselected from interval 1...CW+1.Minimum interframe space is PIFS.Earliest channel access is DIFS.

earliest channel accessfor high priority AC

CW[AC=low]

CTS

RTS

time

ACKSIFS

AIFS[AC=med.]

AIFS[AC=high](=PIFS)

PIFS

AIFS[AC=low]

SIFS

CW[AC=high]

with 802.11a: aSlotTime: 9us SIFS: 16us PIFS: 25us DIFS: 34us AIFSN: 1…10[slots] AIFS: >=PIFS

SIFS

highpriority AC

lowpriority AC

mediumpriority AC

backoff

backoff

busychannel

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

aSlotTime

DCF: Random backoff counter isselected from interval 0...CW.Minimum interframe space is DIFS.Earliest channel access is DIFS.

EDCF: Random backoff counter isselected from interval 1...CW+1.Minimum interframe space is PIFS.Earliest channel access is DIFS.

earliest channel accessfor high priority AC

CW[AC=low]

CWmaxCWmax[AC]

(PF=2PF[AC]*)


Achievable ThroughputAchievable ThroughputAchievable ThroughputAchievable Throughput

Three different EDCF parameter setsAC (priority):higher medium(=legacy) lower

AIFSN[AC]:2 2 9

CWmin[AC]:7 15 31

CWmax[AC]:10231023 1023

PF[AC]:24/16 32/16 40/16

Question: achievable throughput per backoff entity in an isolated scenario? "saturation throughput"

Isolated scenario means: the same EDCF parameters are use by all backoff entities

Results depend on: frame body length, number of contending backoff entities, RTS/CTS, fragmentation

Approach: WARP2 stochastic simulation and analytical model (modifications of Bianchi’s legacy 802.11 model)


512 byte frame body: 512 byte frame body, RTS/CTS:


Legacy (Medium) PriorityLegacy (Medium) PriorityLegacy (Medium) PriorityLegacy (Medium) Priority

10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1

with address 4, w/o WEP encrypt.

number of backoff entities

satu

ratio

n th

rp. (

no

rm.)

BPSK1/2 (6 Mbit/s)16QAM1/2 (24 Mbit/s)64QAM3/4 (54 Mbit/s)

10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 80 100 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)





Low Priority (larger CWmin[AC])Low Priority (larger CWmin[AC])Low Priority (larger CWmin[AC])Low Priority (larger CWmin[AC])

10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


thrp. increases with increasingnumber of backoff entities

10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


thrp. increases with increasingnumber of backoff entities

10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1with address 4, w/o WEP encrypt.


satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1with address 4, w/o WEP encrypt.


satu

ratio

n th

rp. (

no

rm.)





High Priority (smaller CWmin[AC])High Priority (smaller CWmin[AC])High Priority (smaller CWmin[AC])High Priority (smaller CWmin[AC])

10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


deviation with highercollision probability

10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)



10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)



10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)



10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)


10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)



10 20 40 60 0

0.2

0.4

0.6

0.8

1



satu

ratio

n th

rp. (

no

rm.)




Modified Bianchi ModelModified Bianchi ModelModified Bianchi ModelModified Bianchi Model

(1-p)/ W0

p/ Wi

1

1

11

11 1

11

p/ W1

p/ Wi+1

p/ Wm

p/ Wmp/ Wm

p/ Wi

k: slot index i: stage index m: maximum backoff stage p: collision probability W0: CWmin+1 Wm: CWmax+1

(1-p)/ W0“fireand

new“

“coll.“

“coll.“

“coll.“

m depends on Persistent Factor (PF)in the EDCF (proposed), CWmax, andthe retry counter.

0,20,0 0,1 0,W0-2 0,W0-1

i-1,0

m,k=2m,0 m,k=1

i, k=2i,0 i, k=1 i,Wi-1i,Wi-2

-2mWm, -1mWm,

“coll.“

slot index k

stageindex i

W0 W0[AC] in EDCF

Wm Wm[AC] in EDCF m m[AC] in EDCF

(1-p)/ W0

p/ Wi

1

1

11

11 1

11

p/ W1

p/ Wi+1

p/ Wm

p/ Wmp/ Wm

p/ Wi

k: slot index i: stage index m: maximum backoff stage p: collision probability W0: CWmin+1 Wm: CWmax+1

(1-p)/ W0“fireand

new“

“coll.“

“coll.“

“coll.“

m depends on Persistent Factor (PF)in the EDCF (proposed), CWmax, andthe retry counter.

0,20,0 0,1 0,W0-2 0,W0-1

i-1,0

m,k=2m,0 m,k=1

i, k=2i,0 i, k=1 i,Wi-1i,Wi-2

-2mWm, -1mWm,

“coll.“

slot index k

stageindex i

W0 W0[AC] in EDCF

Wm Wm[AC] in EDCF m m[AC] in EDCF


Share of CapacityShare of CapacityShare of CapacityShare of Capacity

Saturation throughput shown so far is only valid for isolated scenariosNice to have, but useless for QoS support:

For QoS support, a backoff entity needs to know the expected throughput in mixed scenarios

Achievable throughput per backoff entity is referred to as "share of capacity"

Question: what is the share of capacity a backoff entity can achieve in a mixed scenario?

This is *THE* important question for EDCF QoS support

Bianchi model does not provide the answer

There is no solution available until today


Access Probability per SlotAccess Probability per SlotAccess Probability per SlotAccess Probability per Slot

25 34 43 52 61 70 79 88 97 106 115 124 133 142 151 160 1690

0.1

0.2

0.3

0.4

0.5

0.6

pro

b(a

cce

ss)

slot [s]

AIFSN[higher pr.]=2

AIFSN[medium pr.]=2

AIFSN[lower pr.]=9

8 backoff entities per AC[higher]

8 backoff entities per AC[medium]

8 backoff entities per AC[lower]

higher pr.medium pr.lower pr.

25 34 43 52 61 70 79 88 97 106 115 124 133 142 151 160 1690

0.1

0.2

0.3

0.4

0.5

0.6

pro

b(a

cce

ss)

slot [s]

AIFSN[higher pr.]=2

AIFSN[medium pr.]=2

AIFSN[lower pr.]=9





25 34 43 52 61 70 79 88 97 106 115 124 133 142 151 160 1690

0.1

0.2

0.3

0.4

0.5

0.6

pro

b(a

cce

ss)

slot [s]

AIFSN[higher pr.]=2

AIFSN[medium pr.]=2

AIFSN[lower pr.]=9





25 34 43 52 61 70 79 88 97 106 115 124 133 142 151 160 1690

0.1

0.2

0.3

0.4

0.5

0.6

pro

b(a

cce

ss)

slot [s]

AIFSN[higher pr.]=2

AIFSN[medium pr.]=2

AIFSN[lower pr.]=9






Approximation of Expected Idle TimesApproximation of Expected Idle TimesApproximation of Expected Idle TimesApproximation of Expected Idle Times

Expected size of contention windowN[AC] = number of backoff entities of AC

tau[AC] = probability that a backoff entity is transmitting

Access probability per slotExpressed by geometric distribution

N AC

N AC

-persistent CSMA with N Bianchi approximation with Ncontending backoff entities per AC contending backoff entities per AC

1 AC1E CW AC E CW AC

AC 1 1 AC

!

N ACslot AIFS AC

slot AC 1 1 AC 1 AC


CSMA Regeneration Cycle ProcessCSMA Regeneration Cycle ProcessCSMA Regeneration Cycle ProcessCSMA Regeneration Cycle Process C: inter-AC collision H: high priority access M: medium priority access L: low priority access

CWmax = max(CWmax[AC])

C

slot

1

H M L

slot-1

2

CWmax+1

P1,C

P1,LP1,MP1,H

P2,C

P2,LP2,MP2,H

1 111

P1,2

P2,3

Pslot-2, slot-1

Pslot-1, slot

Pslot, slot+1

PCWmax, CWmax+1

Pslot-1,C

Pslot-2,LPslot-1,MPslot-1,H

Pslot,C

Pslot,LPslot,MPslot,H

PCWmax+1,C

PCWmax+1,LPCWmax+1,M

PCWmax+1,H

C: inter-AC collision H: high priority access M: medium priority access L: low priority access

CWmax = max(CWmax[AC])

C

slot

1

H M L

slot-1

2

CWmax+1

P1,C

P1,LP1,MP1,H

P2,C

P2,LP2,MP2,H

1 111

P1,2

P2,3

Pslot-2, slot-1

Pslot-1, slot

Pslot, slot+1

PCWmax, CWmax+1

Pslot-1,C

Pslot-2,LPslot-1,MPslot-1,H

Pslot,C

Pslot,LPslot,MPslot,H

PCWmax+1,C

PCWmax+1,LPCWmax+1,M

PCWmax+1,H

State transition diagram for the Markov chain

States C, H, M, L represent busy system

States 1, 2, 3..., CWmax+1 represent idle system

Time is progressing in steps of a slot

State of the chain changes with state transition probabilities as indicated in the figure


Markov Chain (1)Markov Chain (1)Markov Chain (1)Markov Chain (1)

Resulting state transition probabilitiesaccess:

collision:

idle:

slot,H slot slot slot

slot,M slot slot slot

slot,L slot slot slot

P High 1 Medium 1 Low ,

P Medium 1 High 1 Low ,

P Low 1 High 1 Medium .

slot,C slot slot slot

slot slot slot

slot slot slot

slot slot slot

P High Medium 1 Low

High Low 1 Medium

Medium Low 1 High

High Medium Low .

slot,slot 1slot,H slot,M slot,L slot,C

0, slot CWmaxP

1 P P P P , else


Markov Chain (2)Markov Chain (2)Markov Chain (2)Markov Chain (2)

Resulting stationary distributionshigh:

other:

slot 1CWmax 1

H 1,H slot,H i,i 1 1slot 2 i 1

(this defines the relative priority of the AC "High")

p P P P p

AC High:

slot 1CWmax 1

M 1,M slot,M i,i 1 1 1slot 2 i 1

slot 1CWmax 1

L 1,L slot,L i,i 1 1 1slot 2 i 1

slot 1CWmax 1

C 1,C slot,C i,i 1 1slot 2 i 1

p P P P p : Medium p,

p P P P p : Low p,

p P P P p.


ResultResultResultResult

The priority vector

Share of capacity

Modified Bianchi model provides the saturation throughput

H M L

AC

1 , , High , Medium , Low

AC

H

M

L

Thrp HighsatThrp Thrp Thrp Medium .share sat sat

Thrp Lowsat


Scenario & Results (1)Scenario & Results (1)Scenario & Results (1)Scenario & Results (1)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

512 bytes frame body, no RTS/CTS

4+4+4 backoff entities

sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:

higher priority legacy priority lower priority

variable pr. (high low) sim.variable pr. (high low) apprx.legacy pr. sim.legacy pr. apprx.lower pr. sim.lower pr. apprx.

2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt. sim.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7



sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:



2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt. sim.

receiving station

variablepriority

legacypriority

lowpriority

receiving station

variablepriority

legacypriority

lowpriority

Four backoff entities per AC (4/4/4)Variable, legacy and low priority

Results of WARP2 simulation indicate accurate approximation



10/2/4 backoff entities per ACBackoff entities with variable priority are more dominant, as expected

Results of WARP2 simulation indicate accurate approximation

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7



sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:



2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt.

sim.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7



sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:



2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt.

sim.

receiving station

variablepriority

legacypriority

lowpriority

receiving station

variablepriority

legacypriority

lowpriority



0

0.1

0.2

0.3

0.4

0.5

0.6

0.7



sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:



2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt.

sim.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7



sha

re (

=th

rp.)

pe

r A

C (

no

rm.)

sim.

analyt.

AIFS:

CWmin:

16 PF:



2 2 2 2 2 2 2 2 2 2 2 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9

7 7 7 7 7 8 9 10 12 13 14 15 15 15 15 15 15 15 15 17 19 21 23 25 27 29 31 31

24 26 28 30 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 40

analyt.

sim.

receiving station

variablepriority

legacypriority

lowpriority

receiving station

variablepriority

legacypriority

lowpriority

2/10/4 backoff entities per ACBackoff entities with variable priority are more dominant, as expected

WARP2 simulation results deviate for different persistent factors


EDCF SummaryEDCF SummaryEDCF SummaryEDCF Summary

EDCF MAC protocol is distributed (as DCF, simple)Multiple queues per station (queue = backoff entity)The presented model can be used for prediction of expected share of capacity per backoff entityThe model can be extended towards delay and loss predictionEDCF supports QoS, but cannot guarantee as resulting share depends on activity of other backoff entities


QoS Support in 802.11e EDCF? yes, but no guarantee!


HCF Controlled Medium AccessHCF Controlled Medium AccessHCF Controlled Medium AccessHCF Controlled Medium Access

EDCF cannot guarantee QoS, because of distributed MAC

For guarantee, controlled medium access allows access right after PIFS, without backoff

Similar to polling in legacy 802.11 (PCF)

CTS

RTSAIFS[AC] DATA (MSDU)

ACK

EDCF-TXOP gained by contention-basedchannel access during contention period

duration < EDCF-TXOPlimit

QoS CF-Poll

optimal CAP allocationtime for HC 1

delayed start of TXOP

CAPallocation

PIFS

delayed CAP allocationtime for HC 1

time

tolerated by HC 1

under control of HC 1

busychannel

CTS

RTSAIFS[AC] DATA (MSDU)

ACK

EDCF-TXOP gained by contention-basedchannel access during contention period

duration < EDCF-TXOPlimit

QoS CF-Poll

optimal CAP allocationtime for HC 1

delayed start of TXOP

CAPallocation

PIFS

delayed CAP allocationtime for HC 1

time

tolerated by HC 1

under control of HC 1

busychannel


HCF in Overlapping BSSHCF in Overlapping BSSHCF in Overlapping BSSHCF in Overlapping BSS

Controlled medium access requires an isolated BSS

No other backoff entity must access the medium with highest priority (after PIFS), otherwise collisions occur!

This is a very strict requirement, and difficult to achieve in an unlicensed frequency band

Dynamic frequency selection may help, as in HiperLAN/2

512 byte frame body: 2304 byte frame body:

1 2 10 0

0.2

0.4

0.6

0.8

1

number of HCs allocating CAPs

satu

ratio

n th

rp. (

no

rm.)

CW=0, AIFSN=1


1 2 10 0

0.2

0.4

0.6

0.8

1


satu

ratio

n th

rp. (

no

rm.)

CW=0, AIFSN=1


1 2 10 0

0.2

0.4

0.6

0.8

1


satu

ratio

n th

rp. (

no

rm.)

CW=0, AIFSN=1


1 2 10 0

0.2

0.4

0.6

0.8

1


satu

ratio

n th

rp. (

no

rm.)

CW=0, AIFSN=1



HCF Controlled Access SummaryHCF Controlled Access SummaryHCF Controlled Access SummaryHCF Controlled Access Summary

The controlled medium access is often referred to as HCF

This coordination function is not distributed, it is centralized (requires a Hybrid Coordinator)

It works only in isolated scenarios, which is not a very likely scenario in unlicensed bands

The coexistence problem of overlapping BSSs will be discussed in the following



QoS Support with 802.11e HCF? not in unlicensed bands!


Scenario: two BSSs Sharing one ChannelScenario: two BSSs Sharing one ChannelScenario: two BSSs Sharing one ChannelScenario: two BSSs Sharing one Channel

CCHC(player 2)

CCHC(player 1)

vectors indicate"has control over"

CCHC'sdetection ranges

802.11station

802.11station

HiperLAN/2station

HiperLAN/2station

CCHC(player 2)

CCHC(player 1)

vectors indicate"has control over"

CCHC'sdetection ranges

802.11station

802.11station

HiperLAN/2station

HiperLAN/2station

Basic service sets are modeled as players that attempt to optimize their outcomesSingle stage game: one superframe (~200ms)Multi stage game: repeated interaction


The Superframe as Single Stage GameThe Superframe as Single Stage GameThe Superframe as Single Stage GameThe Superframe as Single Stage Game

[0...1]

QoS [0...1]

[0...1]

Allocation process during a superframe:

QoS:

SFDUR(n)[ms]

the periodic beacon is successfullytransmitted by one of the CCHCs

TBTT TBTT

time

1...L1 TXOPs allocatedby CCHC1 (here, L1=3)

d11(n) [ms] d3

1(n) [ms]d21(n) [ms]

DL1(n) = D3

1(n) [ms]D11(n) [ms] D2

1(n) [ms]

t11(n) t3

1(n)t21(n)

nth CCHC superframe = thenth single-stage game

allocatedby CCHC1

allocated byCCHC2

SFDUR(n)[ms]

the periodic beacon is successfullytransmitted by one of the CCHCs

TBTT TBTT

time

1...L1 TXOPs allocatedby CCHC1 (here, L1=3)

d11(n) [ms] d3

1(n) [ms]d21(n) [ms]

DL1(n) = D3

1(n) [ms]D11(n) [ms] D2

1(n) [ms]

t11(n) t3

1(n)t21(n)

nth CCHC superframe = thenth single-stage game

allocatedby CCHC1

allocated byCCHC2


Abstract Representation of QoSAbstract Representation of QoSAbstract Representation of QoSAbstract Representation of QoS

Throughput: normalized share of capacity

Delay: normalized resource allocation interval

Jitter: normalized delay variation

,

iL (n)i i

ll 1

1(n) d (n)

SFDUR(n)

i

i imax l l 1...L (n) 1

1(n) max D (n)

SFDUR(n)

i

i i imax l l 1

l 1...L (n) 1

1(n) max D (n) D (n)

SFDUR(n)


Player "i" and opponent player "–i" have individual requirementsPlayers select demands to meet requirementsThrough allocation process, players observe outcomes per single stage game: observed QoS

This single stage game is repeated with every superframePlayers adapt behaviors in the multi stage game

The PlayerThe PlayerThe PlayerThe Player

action ai of player i:select demand based

on requirement,observation, and

estimated demand ofplayer -i

z -1allocation process

requirement

time

ireq

ireq

iobs

iobs

n

n

idem

idem

n

n

demand of player -i

time

idem

idem

n

n

demand

time time

observation(outcome)

action ai of player i:select demand based

on requirement,observation, and

estimated demand ofplayer -i

z -1allocation process

requirement

time

ireq

ireq

iobs

iobs

n

n

idem

idem

n

n

demand of player -i

time

idem

idem

n

n

demand

time time

observation(outcome)


Allocation Process (Formal Description)Allocation Process (Formal Description)Allocation Process (Formal Description)Allocation Process (Formal Description)

Required:

If this process can be formally described through an accurate approximation, we can analyze

Expected outcomes (existence of Nash equilibrium (NE))

Stability (convergence to NE)

Fairness (position of NE in bargaining domain)

It can be discussed…… what QoS support is feasible for the individual players (player = CCHC = BSS)

… what level of QoS can be achieved

… if mutual cooperation improves the outcome per player

i i idem dem obsi i idem dem obs

, , i, i {1,2}

.


Observed payoffs in a single stage game:

Stationary distributions:p0: idle channel (EDCF background traffic)

p1: player 1 allocates radio resource

p2: player 2 backing off while player 1 allocates resource

State transition probabilities:

Markov ChainMarkov ChainMarkov ChainMarkov Chain

P23

P41

P43

P10 P21P34 P03

p0p4 p3 p2p1

P30 P01 P12

P23

P41

P43

P10 P21P34 P03

p0p4 p3 p2p1

P30 P01 P12

21,2dem

01 dem2 1dem dem

P , 0

1 11,2dem dem

12 dem2 2dem dem

P min 1, 01


Result and EvaluationResult and EvaluationResult and EvaluationResult and Evaluation

Resulting observations for both players:

Comparison with simulation results (YouShi):

i ii dem demobs i i i i

dem dem dem dem

i i idem dem dem

allocationinterval unwanted increaseof allocationinterval (

iobs

demand delayed )

TXOPlimit


The Utility FunctionThe Utility FunctionThe Utility FunctionThe Utility Function

Players attempt to meet their requirementsTherefore, players attempt to maximize the observed payoff (outcome), by using a utility function

i i i i i i i i idem obs req obs reqU U ( , , ) U ( , ), U


Existence of Nash Equilibrium (NE)Existence of Nash Equilibrium (NE)Existence of Nash Equilibrium (NE)Existence of Nash Equilibrium (NE)

Proposition: in the Single Stage Game of two coexisting CCHCs exists a Nash equilibrium in the action space A.

Proof: show that the outcome (the payoff V) is continuous in A, and show that it is quasi-concave in Ai.

There exists at least one Nash equilibrium, which can be calculated as:

a=action, V=payoff, N=number of players (N=2)

idemi i i

idem

a a* 0 grad V (a) i , with grad


Pareto EfficiencyPareto EfficiencyPareto EfficiencyPareto Efficiency

Players that take rational actions will automatically adjust into a NE (because there is at least one NE)

If the NE is unique, the respective action profile can be predicted as expected point of operation

However, there may exist action profiles in the single stage game that lead to higher payoffs

If such profiles do not exist, the NE is referred to as Pareto efficient (Pareto optimal)

Pareto efficiency can be determined by numerical search

Can be shown in bargaining domain … (next page)


Bargaining DomainBargaining DomainBargaining DomainBargaining Domain

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

payoff of player i: V i(a

i,a

-i)

payo

ff o

f pl

ayer

-i:

V -

i (ai ,a

-i)

Nash equilibrium

Pareto boundary

fair share

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

payoff of player i: V i(a

i,a

-i)

payo

ff o

f pl

ayer

-i:

V -

i (ai ,a

-i)

Nash equilibrium

Pareto boundary

fair share


Persist: demand=requirementShown are YouShi simulation results and analytical apprx.Poor delay performance for pl.2

Strategy: PersistStrategy: PersistStrategy: PersistStrategy: Persist

0

0.6

1

1

required observeddemanded

analyt. apprx.(dashed line not visible)

simulated(solid line)

arrow indicates thatplayers generallyattempt to maximizethroughput

some variationsbecause of EDCF

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.04

0.1

1

time (s), SFDUR = 200ms

required observed maxdemanded

analyt. apprx.(dashed line)


arrow indicates thatplayers generallyattempt to minimizedelays

0

0.6

1

2



analyt. apprx.(dashed line)

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.023

0.1

2



simulated(solid line) analyt. apprx.

(dashed line)

pl1 pl2


Persist/Best Response/CooperationPersist/Best Response/CooperationPersist/Best Response/CooperationPersist/Best Response/Cooperation

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

best

coop

persist

defect


pl 1pl 2

both players demandrequirements throughoutall stages (BEH-P)

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.2

0.4

1


Util

itie

s U

1,2

pl 1pl 2

in total, player 1 observes a higher payoff than player 2 whenboth demand their requirements

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

best

coop

persist

defect


pl 1pl 2

after 2s, both players changetheir behavior from BEH-P toBEH-B independently, then attempting to improve their individual payoffs

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.2

0.4

1


Util

itie

s U

1,2

pl 1pl 2

in total, player 2 gains and player 1suffers from playing the best responses

neither player is able to improve its outcomeby unilaterally changing its behavior from whatis demanded after the process converged into NE

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

best

coop

persist

defect


pl 1pl 2

both players cooperate after 2s

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.2

0.4

1


Util

itie

s U

1,2

pl 1pl 2

in total, payoffs are higher in cooperationthan in NE, therefor the NE is not Paretoefficient in this example


Cooperation can be beneficial for both players, and is established in repeated interactions (multi stage game)Cooperation and punishment:

Payoff discounting in multi stage game:

How to establish CooperationHow to establish CooperationHow to establish CooperationHow to establish Cooperation

n=n0

for a number of stages,depending on discouning factorCOOPERATE:

BEH-CPUNISH(1):

BEH-DPUNISH(n’):

BEH-D

opponentdefects

any behaviorof opponent


otherwise

n=n0

for a number of stages,depending on discouning factorCOOPERATE:

BEH-CPUNISH(1):

BEH-DPUNISH(n’):

BEH-D

opponentdefects



otherwise

i n iMSG

n 0V V (n)


Condition for CooperationCondition for CooperationCondition for CooperationCondition for Cooperation

It is more efficient to cooperate instead of defect (instead of playing best response), if…

It depends on the discounting factor (importance/shadow of future) if mutual support is achievable:

The more important the future is, the more likely is the establishment of cooperationFor example, CCHCs will interact for many superframes

n n' 1k k ki i i i i i i

CC DC CD CCk n k n 1 k n n' 1

V V V V

i ii CC DC

i iCD DC

V V

V V


Dependence on Discounting FactorDependence on Discounting FactorDependence on Discounting FactorDependence on Discounting Factor

1 2 3 4 5 6 7 8 9 100.5

1

1.5

stages of punishment through player -i

VM

SG

i o

f pla

yer

i

VCOOPi ( i=1)

i = 1

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

i

VCOOPi ( i=1)

i = 1

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

i

VCOOPi ( i=0.8)

i = 0.8

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

i

VCOOPi ( i=0.8)

i = 0.8

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

i

VCOOPi ( i=0.75)

i = 0.75

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

i

VCOOPi ( i=0.75)

i = 0.75

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

iV

COOPi ( i=0.6)

i = 0.6

1 2 3 4 5 6 7 8 9 100.5

1

1.5


VM

SG

i o

f pla

yer

iV

COOPi ( i=0.6)

i = 0.6

Future counts

Future is less important


Wrap UpWrap UpWrap UpWrap Up

There is always a Nash equilibrium in the single stage game

If the outcome of the Nash equilibrium is not satisfying, a player may attempt to punish the opponent, for establishment of mutual support

Depending on the behaviors of the CCHCs (the interacting players), and their requirements, cooperation can be achieved

QoS can be supported if cooperation is established



QoS Support with 802.11e HCF? not in unlicensed bands!

QoS Support with shared radio resources? with mutual support: yes!


ConclusionsConclusionsConclusionsConclusions

IEEE 802.11e EDCF will provide basic means for QoS support

The controlled medium access of HCF (polling) cannot support QoS in unlicensed frequency bands

New analytical model for EDCF is developedallows to predict and control QoS

New approach for coexisting radio networksmay help radio networks operating in unlicensed bands to support QoS

Results will be used in …Contributions to IEEE 802.11e

IEEE 802.19 coexistence discussions

Spectrum etiquette development at Wi-Fi alliance

Development of Spectrum Agile Radios (DARPA)

Backup SlidesBackup Slides


ArchitectureArchitectureArchitectureArchitecture

WirelessStation Wireless

Station

WirelessStationIBSS

Wired Station(Access Point, AP)

WirelessStation

WirelessStation

WirelessStation

BSS


WirelessStation

WirelessStation

WirelessStation

BSS

802.x LANvia Portal

DS

WirelessStation Wireless

Station

WirelessStationIBSS


WirelessStation

WirelessStation

WirelessStation

BSS


WirelessStation

WirelessStation

WirelessStation

BSS

802.x LANvia Portal

DS

Infrastructure Basic Service Set (BSS)one station is the access point

Independent Basic Service Set (IBSS)ad-hoc


Medium Access - ExampleMedium Access - ExampleMedium Access - ExampleMedium Access - Example

Station 1 initiates frame exchange firstOther stations set the Network Allocation Vector (NAV)Distributed approach difficult for station to support QoS

CTS

RTS

time

randombackoff(7 slots)

randomback-off(9 slots)

station 3defers, but

keeps backoffcounter (=2)

ACK

DATA

new randombackoff

(10 slots)

stationdefersDATA

ACK

ACK

DATA

remainingbackoff(2 slots)

SIFS

SIFS

SIFS

SIFS

SIFS

DIFS

DIFS

DIFS

DIFS

NAVs

station 1

station 2

NAVreset

stations set NAV uponreceiving RTS

station 6 sets NAV upon receiving CTS,this station is hidden to station 1

NAVupdates

station 5

station 4

station 3

station 6

NAV (timer)

transmission

CTS

RTS

time

randombackoff(7 slots)

randomback-off(9 slots)

station 3defers, but

keeps backoffcounter (=2)

ACK

DATA

new randombackoff

(10 slots)

stationdefersDATA

ACK

ACK

DATA

remainingbackoff(2 slots)

SIFS

SIFS

SIFS

SIFS

SIFS

DIFS

DIFS

DIFS

DIFS

NAVs

station 1

station 2

NAVreset

stations set NAV uponreceiving RTS

station 6 sets NAV upon receiving CTS,this station is hidden to station 1

NAVupdates

station 5

station 4

station 3

station 6

NAV (timer)

transmission


Multiple Backoff Entities per StationMultiple Backoff Entities per StationMultiple Backoff Entities per StationMultiple Backoff Entities per Station

transmission

one priority

backoff:DIFS

151023

legac y 802.11 stationwith one backoff entity:

PF[AC] notpart of

802.11e

upon parallel access at the same slot, the higher priority ACbackoff entity transmits, the other backoff entity/entities act

as if a collision occured

transmission

higher priority lower priority

4 Access Categories AC0 - AC3 representing 4priorities, with 4 independent backoff entities

AC0

backoff:AIFSN[0]CWmin[0]CWmax[0]

AC1


AC2


AC3


IEEE 802.11e station with four backoff entities:

8 priorities 0 - 7 according to 802.1D aremapped to 4 Access Categories (ACs)

7 4 015 236

AIFSN = 1,2,3…AIFS = SIFS + aSlotTime x AIFSN

backoffentity

backoffentity

transmission

one priority

backoff:DIFS

151023

legac y 802.11 stationwith one backoff entity:

PF[AC] notpart of

802.11e

upon parallel access at the same slot, the higher priority ACbackoff entity transmits, the other backoff entity/entities act

as if a collision occured

transmission

higher priority lower priority

4 Access Categories AC0 - AC3 representing 4priorities, with 4 independent backoff entities

AC0


AC1


AC2


AC3


IEEE 802.11e station with four backoff entities:

8 priorities 0 - 7 according to 802.1D aremapped to 4 Access Categories (ACs)

7 4 015 236

AIFSN = 1,2,3…AIFS = SIFS + aSlotTime x AIFSN

backoffentity

backoffentity


Markov ChainMarkov ChainMarkov ChainMarkov Chain

slot,AC

slot,C

slot,slot 1

P s t 1 AC| s t slot P , AC H,M,L,

P s t 1 C| s t slot P ,

P s t 1 slot 1| s t slot P , slot 1 CWmax 1

ACt

Ct

slott

limP s t AC p , AC H,M,L,

limP s t C p ,

limP s t slot p , slot 1 CWmax 1

State transition probabilities

Stationary distributions


Allocation Process (Example)Allocation Process (Example)Allocation Process (Example)Allocation Process (Example)

Two single stage games (two superframes):

Two players interact with each otherA third player models the EDCF background trafficFor analysis, a formal description of this process is needed

0 40 80 120 160 200 240 280 320 360 400

beacons(at TBTTs)

player 1

player 2

player 3 (EDCF)

TX

OP

s

time [ms]

collisioncollision collision

0 40 80 120 160 200 240 280 320 360 400

beacons(at TBTTs)

player 1

player 2

player 3 (EDCF)

TX

OP

s

time [ms]

collisioncollision collision


Best Response: adapt demand to achieve highest outcome (myopic competition)

Action profile (demand) converges to NE

0

0.6

1

1

required observed demanded

apprx. sim.

observed thrp. decreases, because now theopponent player 2 plays its best response

as both players play theirbest responses, the demandsconverge into Nash equilibrium (NE)

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.04

0.1

1



sim. apprx.

demanding NE

0

0.6

1

2

required observed demanded

sim. apprx.

now demanding high thrp. when converging into NE

this player gains from playingthe best response

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.023

0.1

2



sim. apprx.

demanding NE

Strategy: Best ResponseStrategy: Best ResponseStrategy: Best ResponseStrategy: Best Response

pl1 pl2


Cooperation: reduced demand, shorter resource allocationsNow both players achieve higher outcomes (next page…)

Strategy: CooperationStrategy: CooperationStrategy: CooperationStrategy: Cooperation

0

0.6

1

1


apprx.(not visible)

sim.

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.04

0.1

1



apprx.

sim.

0

0.6

1

2


apprx. sim.

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4

0

0.023

0.1

2


required observed maxdemandedapprx.

sim.

pl1 pl2

Analysis of IEEE 802.11e and Application of Game Models for Support of Quality-of-Service in Coexisting Wireless Networks Stefan Mangold ComNets Aachen.

Documents

qos support

todays wireless lans

qos requirementswireless

e qos extensionoverview

qos mechanisms

level of qos

e medium access

wireless lanradio standard