analysis of an energy-efficient mac protocol based on polling for ieee 802.11 wlans

26
IEEE ICC 2015, 8–12 June, London, UK Raul Palacios a , Gedlu Mengistie Mekonnen a , Jesus Alonso- Zarate b , Dzmitry Kliazovich c and Fabrizio Granelli a a DISI University of Trento, Italy b CTTC, Barcelona, Spain c University of Luxembourg, Luxembourg Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

Upload: fabrizio-granelli

Post on 17-Aug-2015

107 views

Category:

Technology


1 download

TRANSCRIPT

Page 1: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

IEEE ICC 2015, 8–12 June, London, UK

Raul Palaciosa, Gedlu Mengistie Mekonnena, Jesus Alonso-Zarateb, Dzmitry Kliazovichc and Fabrizio Granellia

aDISI University of Trento, ItalybCTTC, Barcelona, Spain

cUniversity of Luxembourg, Luxembourg

Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

Page 2: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

2IEEE ICC 2015

London, UK

Outline

Scope

• Wi-Fi Footprint

• IEEE 802.11 PCF/DCF MAC Layer

• Energy Issues

Contribution

• Novel MAC Protocols

• Theoretical analysis

• Computer-based Simulation

Outcome

• Throughput

• Energy Efficiency

• Vs. Traffic/ Packet/ Rate/No. Stations

Page 3: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

3IEEE ICC 2015

London, UK

Wi-Fi Footprint

• The number of Wi-Fi public hotspots will increase by 350% in 2015.

• Wi-Fi home and hotspots will contribute by 31%-34% to the overall yearly cloud energy consumption, being the second main contributor after mobile networks.

Source: The Power of Wireless Cloud, Alcatel-Lucent’s Bell Labs, Centre for Energy-Efficient Telecommunications (CEET), University of Melbourne, Apr.

2013

Source: Wireless Broadband Access (WBA), Informa, Nov. 2011

1%

1%

2009 2010 2011 2012 2013 2014 2015

0.5 0.81.3

2.13.3

4.55.8

350%

Number of Wi-Fi Public Hotspots in the World (in million), 2009-2015

Metro & Core Networks

Data centers

Local (Wi-Fi Home/Wi-Fi Hotspots)

Mobile access networks (4G LTE)

2012 2015 Lo 2015 Hi

9173

32424

42957

57%55%

59%16%26%

1%

34%

31%10%

9%

Total Annual Wireless Cloud Energy Consumption (GWh), 2012-2015

Page 4: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

4IEEE ICC 2015

London, UK

off sleep idle

transmit

receive

doze

awakeTransition from the doze state to the awake state requires additional time

and energy consumption

Operating Modes of the Wi-Fi Interface

Page 5: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

5IEEE ICC 2015

London, UK

Power Characteristics of the Wi-Fi Interface

Transmit mode (1.65W)Receive mode (1.4W)Idle mode (1.15W)Sleep mode (45mW)

No

rma

lise

d p

ow

er c

onsu

mp

tion (

%) 100

90

80

70

60

50

40

30

20

10

0

Off mode (0W)

Source: Lucent IEEE 802.11 WaveLAN card

Page 6: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

6IEEE ICC 2015

London, UK

Radio’s On/Off Transitions [1]

[1] P. J. M. Havinga and G. J. M. Smit, “Energy-efficient TDMA medium access control protocol scheduling,” in Asian International Mobile Computing Conference (AMOC), Nov. 2000, pp. 1–10.

Switching between idle and sleep states takes about hundreds of μs.

It depends on the radio’s hardware design.

There is a peak of power consumption from sleep to idle state.

The on/off radio transitions cannot be neglected.

Page 7: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

7IEEE ICC 2015

London, UK

IEEE 802.11 MAC Layer for WLANs

activepower save

sleep

awake

DCF PCFmandatorydistributedcontentionbest effort

widely used

optionalcentralised

pollingquality of service

rarely used

PS modeno data to send

periodic wake-up listen to beacons

retrieve data

PCF

DCF

Page 8: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

8IEEE ICC 2015

London, UK

PCF Example

STA1

STA2

STA3

PIFS

Contention Free Period (CFP)

NAV Time +

APB C

E

AC

KDATA >STA1

DATA >STA2 PO

LL

DATA >AP

AC

K

Pt

Pt

PtSIFS SIFS SIFS SIFS SIFS SIFS SIFS

TPIFS TB TSIFS TDATA TSIFS TDATA

Pt

PIFS

Pi

Pr

Pi

Pr

Pi

Pr

Pi

Pr

POLL

DATA >APA

CK

POLL

NULL >APA

CK

Time +

Time +

Time +

Time +

TPOLL TACK TSIFSTACKTPOLLTDATA TSIFS TDATATACK TSIFSTPOLLTSIFS TDATA TSIFS TPIFSTACKTCE

AP: Access PointSTA: Wireless StationPIFS: PCF Interframe SpaceSIFS: Short Interframe

SpaceB: Beacon frameACK: AcknowledgmentCE: CFP End frameNAV: Network Allocation

Vector

Page 9: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

9IEEE ICC 2015

London, UK

PS

MInform the AP to buffer data

Enter the sleep state

Periodically awake

Listen to a beacon

Retrieve downlink data

802.11 PSM

802.11e APSD

802.11n PSMP

IEEE 802.11 PSM

Page 10: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

10IEEE ICC 2015

London, UK

IEEE 802.1ac TXOP PSM

TXOP

PSM

Not based on listen intervals

Sleep when others

transmit

Exploit virtual

sensing info Good for

high traffic and many stations

Not aware of radio’s

on/off transitions

Page 11: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

11IEEE ICC 2015

London, UK

Mult

i-Po

lling

UPCF [2]

EE-Multipoll [3]

PSR-PCF [4]

Limitations

PSMP-based

Scalability

Complexity

Performance

Related Works [2]-[4]

[2] Z.-T. Chou, C.-C. Hsu, and S.-N. Hsu, “UPCF: A New Point Coordi- nation Function with QoS and Power Management for Multimedia over Wireless LANs,” IEEE/ACM Trans. on Net., vol. 14, no. 4, pp. 807–820, 2006.

[3] J.-R. Hsieh, T.-H. Lee, and Y.-W. Kuo, “Energy-Efficient Multi-Polling Scheme for Wireless LANs,” IEEE Trans. on Wireless Communications, vol. 8, no. 3, pp. 1532–1541, 2009.

[4] K.-C. Ting, F.-C. Kuo, B.-J. Hwang, H. Wang, and C.-C. Tseng, “A Power-Saving and Robust Point Coordination Function for the Transmission of VoIP over 802.11,” in IEEE ISPA’10, 2010, pp. 283– 289.

Page 12: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

12IEEE ICC 2015

London, UK

Novel Centralized MAC Protocols

Bid

Poll

PCF-based

Two virtual phases

P1: Reserved P2: Round

robinBackwards compatible

Low overheadG

reen

Poll

BidPoll-based

Exploit beacon information

Sleep during the first phase

Cycling polling order scheme

TX time Vs. switching time

Page 13: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

13IEEE ICC 2015

London, UK

BidPoll Example [5]

[5] R. Palacios, F. Granelli, D. Gajic, and A. Foglar, “An Energy-Efficient MAC Protocol for Infrastructure WLAN Based on Modified PCF/DCF Access Schemes Using a Bidirectional Data Packet Exchange,” in IEEE CAMAD 2012, Sep. 2012, pp. 216–220.

AP: Access PointSTA: Wireless StationPIFS: PCF Interframe SpaceSIFS: Short Interframe

SpaceB: Beacon frameACK: AcknowledgmentCE: CFP End frameNAV: Network Allocation

Vector

Page 14: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

14IEEE ICC 2015

London, UK

GreenPoll Example [6]

[6] R. Palacios, F. Granelli, D. Gajic, C. Liß, and D. Kliazovich, “An Energy-efficient Point Coordination Function Using Bidirectional Transmissions of Fixed Duration for Infrastructure IEEE 802.11 WLANs,” in IEEE ICC 2013, 9–13 Jun. 2013, pp. 1036–1041.

AP: Access PointSTA: Wireless StationPIFS: PCF Interframe SpaceSIFS: Short Interframe

SpaceB: Beacon frameACK: AcknowledgmentCE: CFP End frameNAV: Network Allocation

Vector

Page 15: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

15IEEE ICC 2015

London, UK

Reference Scenario

Page 16: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

16IEEE ICC 2015

London, UK

GreenPoll Energy Efficiency Analysis: An Example

Definition of energy efficiency of protocol x

Sleep period of all sleeping STAs

GreenPoll energy consumption

Number of active STAs during the whole polling period

Page 17: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

17IEEE ICC 2015

London, UK

MAC/PHY System Parameters*

Parameter Value Parameter Value

SIFS, PIFS, DIFS, EIFS

10,19,28,88 μs

MAC Header + FCS 34 bytes

Slot time 9 μs MSDU Size50-2250

bytes

Preamble 16 μs Data Rate 6-54 Mbps

Signal 4 μs Control Rate6,12,24 Mbps

Signal Extension 6 μs Transmit Power Consumption 1.65 W

CWmin, CWmax 15, 1023 Receive Power Consumption 1.4 W

Sleep <-> Idle Time

500 μs Idle Power Consumption 1.15 W

Service 6 bits Sleep Power Consumption 0.045 W

Tail 16 bitsSleep->Idle Power

Consumption1.725 W

Size of B,RTS,POLL,CE

20 bytes No. of STAs 1-100

Size of CTS,NULL,ACK

14 bytes Simulation Time 15 s x10 times

*IEEE 802.11g MAC/PHY

Page 18: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

18IEEE ICC 2015

London, UK

0 10 20 30 40 50 60 70 80 90 1000

0.5

1

1.5

2

2.5DCF_AnalysisDCF_SimulationPCF_AnalysisPCF_SimulationBidPoll_AnalysisBidPoll_SimulationGreenPoll_AnalysisGreenPoll_Simulation

Total offered traffic load (Mbps)

Netw

ork

energ

y e

ffici

ency

(M

b/J)

Energy EfficiencyGreenPoll network energy distribution

Results Vs. Traffic Load

172%89% 112%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Sleep

Switch

Idle

Receive

Transmit

Total offered traffic load (Mbps)

Gre

enPoll e

nerg

y c

onsu

mpti

on d

istr

ibuti

on (

%)

Page 19: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

19IEEE ICC 2015

London, UK

250 500 750 1000 1250 1500 1750 2000 22500

0.5

1

1.5

2

2.5

3DCF_AnalysisDCF_SimulationPCF_AnalysisPCF_SimulationBidPoll_AnalysisBidPoll_SimulationGreenPoll_AnalysisGreenPoll_Simulation

MAC Service Data Unit (MSDU) length (Bytes)

Satu

rati

on n

etw

ork

energ

y e

ffici

ency

(M

b/J)

Energy efficiencyGreenPoll network energy distribution

Results Vs. Data Packet Length

330%

85%

250 500 750 1000 1250 1500 1750 2000 22500%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Sleep

Switch

Idle

Receive

Transmit

MAC Service Data Unit (MSDU) length (Bytes)

Gre

enPoll e

nerg

y c

onsu

mpti

on d

istr

ibuti

on (

%)

100% 153%

Page 20: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

20IEEE ICC 2015

London, UK

0 10 20 30 40 50 60 70 80 90 1000

2

4

6

8

10

12

DCF_Analysis

DCF_Simulation

PCF_Analysis

PCF_Simulation

BidPoll_Analysis

BidPoll_Simulation

GreenPoll_Analysis

GreenPoll_Simulation

Number of stations

Satu

rati

on n

etw

ork

energ

y e

ffici

ency

(M

b/J)

Energy efficiencyGreenPoll network energy distribution

Results Vs. No. of Stations

29%

1 2 3 4 5 10 15 20 25 50 75 1000%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Sleep

Switch

Idle

Receive

Transmit

Number of stations

Gre

enPoll e

nerg

y c

onsu

mpti

on d

istr

ibuti

on (

%)

19%

109%202%

Page 21: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

21IEEE ICC 2015

London, UK

Summary

• Wi-Fi (IEEE 802.11 PCF/DCF MAC): Not energy efficient• IEEE 802.11 PSM: Inefficient for heavy trafficProblem• BidPoll: split polling period into two phases: P1) on-demand

reserved RX/TX slots & P2) dynamically assigned RX/TX slots

• GreenPoll: Sleep in P1 when not served or after being served

Solution

• Performance evaluation of BidPoll and GreenPoll through theoretical analysis and computer-based simulationContribution

• Maximum energy efficiency gains: 330-146% Vs. MSDU & 94-172% Vs. Rate & 29-205% Vs. No. of stations

• Energy impact of GreenPoll on/off radio switch up to 20%Result

• Develop analysis and modeling for non-saturated case• Real-life experimentation on WARP v3 platformsFuture Work

Page 22: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

22IEEE ICC 2015

London, UK

Thanks for your kind attention!

Fabrizio Granelli [email protected]

Page 23: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

23IEEE ICC 2015

London, UK

Anlys

Simul

Complete performance assessment

802.11g DCF

802.11g PCF

BidPoll

GreenPoll

Contributions of The Paper

Page 24: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

24IEEE ICC 2015

London, UK

Theore

tica

l A

naly

sis

Error-free/Collision-free

channel

Constant packet length

Always data ready to be transmitted

No packet losses for queue overflow

No hidden terminalsPyth

on

S

imu

lato

r

Ideal channel

Poisson traffic generation

Balanced DL-UL data flows

Constant packet lengthUnbounded

transmit queue

No hidden terminals

General Assumptions

Page 25: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

25IEEE ICC 2015

London, UK

Idle to Sleep

Sleep to Idle

Pidle

Psleep

Pidle

Psleep

Transition

Transition

Idle to Sleep

Sleep to Idle

Pidle

Psleep

α=1 P∙ idle

Psleep

250μs

250μs

α=1.5 P∙ idle

α=2 P∙ idle

Transitions between Sleep and Idle

Page 26: Analysis of an Energy-Efficient MAC Protocol Based on Polling for IEEE 802.11 WLANs

26IEEE ICC 2015

London, UK

6 9 12 18 24 36 48 540

0.5

1

1.5

2

2.5

3DCF_AnalysisDCF_SimulationPCF_AnalysisPCF_SimulationBidPoll_AnalysisBidPoll_SimulationGreenPoll_AnalysisGreenPoll_Simulation

PHY data rate (Mbps)

Satu

rati

on n

etw

ork

energ

y e

ffici

ency

(M

b/J)

Energy efficiencyGreenPoll network energy distribution

Results Vs. PHY Data Rate

80%94%

89%

6 9 12 18 24 36 48 540%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Sleep

Switch

Idle

Receive

Transmit

PHY data rate (Mbps)

Gre

enPoll e

nerg

y c

onsu

mpti

on d

istr

ibuti

on (

%)

163%