Robust and Energy Efficient MAC/PHY Strategies of Wi-Fi Sunghyun - - PowerPoint PPT Presentation

▶
robust and energy efficient mac phy strategies of wi fi
SMART_READER_LITE
LIVE PREVIEW

Robust and Energy Efficient MAC/PHY Strategies of Wi-Fi Sunghyun - - PowerPoint PPT Presentation

Jan 24, 2023 198 likes •701 views

Robust and Energy Efficient MAC/PHY Strategies of Wi-Fi Sunghyun Choi, Ph.D., FIEEE Multimedia & Wireless Networking Lab. Seoul National University, Korea http://www.mwnl.snu.ac.kr Introduction Wi-Fi has become an indispensable part of

slide-1
SLIDE 1

Robust and Energy Efficient MAC/PHY Strategies of Wi-Fi

Sunghyun Choi, Ph.D., FIEEE

Multimedia & Wireless Networking Lab. Seoul National University, Korea http://www.mwnl.snu.ac.kr

slide-2
SLIDE 2

2

Introduction

  • Wi-Fi has become an indispensable part of out daily lives!
  • 8 billion global Wi-Fi shipments expected in 2015
slide-3
SLIDE 3

3

Introduction

  • New features of the emerging Wi-Fi
  • PHY rate
  • Multiple antenna system: up to 8 antennas
  • Wider bandwidth: up to 160 MHz
  • Higher order modulation: up to 256 QAM
  • MAC efficiency
  • Frame aggregation: Aggregate MPDU (A-MPDU)

 Require robustness and energy efficiency

slide-4
SLIDE 4

4

Contents

  • Robust Wi-Fi in mobile environments
  • MoFA: Mobility-Aware Frame Aggregation in Wi-Fi
  • ChASER: Channel-Aware Symbol Error Reduction
  • Energy efficient Wi-Fi
  • Power consumption of Wi-Fi
  • WiZizz: Energy Efficient Bandwidth Management
slide-5
SLIDE 5

5

Robust Wi-Fi in Mobile Environment

slide-6
SLIDE 6

6

Introduction

  • Paradigm shift of Wi-Fi
  • Now, people hold their Wi-Fi devices and move
  • Performance degradation due to mobility (user and/or environment)
  • Faster PHY rate (higher modulation, multiple streams, and wide bandwidth)
  • Longer frame duration (Aggregate MPDU, A-MPDU)
slide-7
SLIDE 7

7

Introduction

  • Aggregate MAC protocol data unit (A-MPDU)
  • Core technology of IEEE 802.11n/ac
  • Packing several MPDUs into a single A-MPDU
  • Amortizing protocol overhead over multiple frames
  • Positive/negative acknowledgement for individual MPDUs (subframes) using BlockAck

 Aggregating more subframes results in much higher throughput!

slide-8
SLIDE 8

8

Introduction

  • Channel estimation and compensation in Wi-Fi
  • Obtaining channel state information (CSI) using training symbols in PLCP preamble
  • Conducted only at the beginning of a frame reception
  • OFDM pilot symbols designed only to track the difference of the local oscillators
  • No way to catch up with CSI variations during a frame reception

IEEE 802.11n Mixed‐mode frame format of A‐MPDU

slide-9
SLIDE 9

9

Channel Estimation and Compensation in Wi-Fi

  • Limitation of channel estimation and compensation

Testbed experiment 1. Error Vector Magnitude (EVM) and IQ constellation

  • Microsoft Sora SDR platform (Rx) and Qualcomm Atheros AR9380 (Tx)
  • As mobility increases, EVM increases!

Rx symbol dispersion at the latter part of AMPDU is much larger than that at the front part of A‐MPDU `

slide-10
SLIDE 10

10

Channel Estimation and Compensation in Wi-Fi

  • Limitation of channel estimation and compensation

Testbed experiment 2. Throughput measurement

  • Programmable 802.11n commercial device
  • Qualcomm Atheros AR9380 / Intel IWL5300
  • Using hostAP to build an AP on linux machine
  • Controlling device drivers (ath9k/iwlwifi)
slide-11
SLIDE 11

11

  • MoFA: Mobility-Aware Frame Aggregation in Wi-Fi
  • A-MPDU length (aggregation bound) adaptation with ease of implementation
  • Simple modification of device driver (using commercial programmable 802.11n NIC)
  • ChASER: Channel-Aware Symbol Error Reduction
  • Chasing channel variation without overhead
  • Receiving process modification (using SDR platform)

Two Proposed Approaches

slide-12
SLIDE 12

12

MoFA: Mobility-Aware Frame Aggregation in Wi-Fi

Source: Seongho Byeon, Kangjin Yoon, Okhwan Lee, Woonsun Cho, Seungseok Oh, and Sunghyun Choi, "MoFA: Mobility‐aware Frame Aggregation in Wi‐Fi," in Proc. ACM CoNEXT 2014, Sydney, Australia, Dec. 2‐5, 2014.

slide-13
SLIDE 13

13

MoFA: Mobility-Aware Frame Aggregation in Wi-Fi

11111111101101110010000000

  • ?

1111111110110111

  • 1111110111110111
  • ,

1111110111110111111

slide-14
SLIDE 14

14

MoFA: Mobility-Aware Frame Aggregation in Wi-Fi

  • Implementation issues

1) Standard‐compliant algorithm (with ease of implementation) 2) Prototype in commercial 802.11n devices (AR9380) with ath9k driver 3) Need to modify transmitter‐side only

A‐RTS: Adaptive use of RTS/CTS in order to

  • vercome hidden

interference

slide-15
SLIDE 15

15

MoFA: Mobility-Aware Frame Aggregation in Wi-Fi

  • Performance of MoFA in time-varying mobile environments
  • One-to-one scenario: Stays and moves half-and-half with a regular pattern
  • Divided into two regions (dashed line in the left figure)

 Performance of MoFA reaches up to the most outer curve which is obtained by the

  • ptimal fixed time bound in each region
slide-16
SLIDE 16

16

MoFA: Mobility-Aware Frame Aggregation in Wi-Fi

  • Performance of MoFA in time-varying mobile environments
  • Multiple node scenario: Three mobile nodes and two stationary nodes
  • 127%, 109%, and 35% higher network throughput than no aggregation, 802.11n

default setting, and optimal bound for 1 m/s

  • STA4 (stationary and close to AP) gets the biggest benefit

10 20 30 40

Mobile STA1 Mobile STA2 Mobile STA3 Static STA4 Static STA5

Throughput (Mb/s) No aggregation 802.11n default setting

  • Opt. time bound for 1 m/s

MoFA

slide-17
SLIDE 17

17

ChASER: Channel-Aware Symbol Error Reduction

Source: Okhwan Lee, Weiping Sun, Jihoon Kim, Hyuk Lee, Bo Ryu, Jungwoo Lee, and Sunghyun Choi, "ChASER: Channel‐Aware Symbol Error Reduction for High‐Performance WiFi Systems in Dynamic Channel Environment,“ in Proc. IEEE INFOCOM 2015, Apr. 26 ‐ May 1, 2015.

slide-18
SLIDE 18

18

ChASER: Channel-Aware Symbol Error Reduction

  • Channel estimation using unknown data symbols
  • Exploit unknown data symbols using

/

  • Exponential weighted moving average filter

1

/

  • CRC-assisted error correction
  • Evaluation methodologies
  • Microsoft Sora SDR platform
  • SDK version 1.6
slide-19
SLIDE 19

19

ChASER: Channel-Aware Symbol Error Reduction

  • Implementation issues (Microsoft Sora SDR platform)
  • High complexity and difficult to implement
  • Feasibility verification
  • We cannot control the commercial 802.11 device’s Rx process (hardware-level)
  • Real-time rx processing?
  • Processing latency due to multiple thread  update CSI every 4 OFDM symbols
  • Sora does not provide real-time AGC at RF front-end  Offline gain control
  • Testbed experiments
  • Baseline 802.11n vs. ChASER
  • Fixed MCS 3, good channel condition

Microsoft SORA

slide-20
SLIDE 20

20

ChASER: Channel-Aware Symbol Error Reduction

  • ChASER chases the wireless channel variation with high fidelity
  • Eliminate caudal losses by tracking channel variation
  • Standard compliant, but high performance gain (up to 56%)
slide-21
SLIDE 21

21

Energy Efficient Wi-Fi

slide-22
SLIDE 22

22

Introduction

  • IEEE 802.11ac standard offers data rate as high as 6933 Mb/s
  • Higher order modulation: up to 256QAM
  • The number of spatial streams and antennas: up to 8
  • Channel bonding: up to 160 MHz
  • WiFi is a primary energy consumer in battery-powered mobile devices
  • IEEE 802.11n 3x3 MIMO receiver consumes more energy than IEEE 802.11a receiver [1]
  • 2x in active mode
  • 1.5x in idle/listening (IL) mode.
  • More energy consumption in IEEE 802.11ac

Consume more energy

[1] C.‐Y. Li, C. Peng, S. Lu, and X. Wang, “Energy‐based rate adaptation for 802.11n,” in Proc. Mobicom, Aug, 2012.

slide-23
SLIDE 23

23

Background

  • Time and energy spent in IDLE/CCA mode [2]
  • Real-world Wi-Fi traces
  • IDLE/CCA is the dominant source of energy consumption in Wi-Fi

[2] X. Zhang and K. Shin, “E‐mili: energy‐minimizing idle listening in wireless networks,” IEEE Trans. Mob. Computing., vol. 11, no. 9, 2012.

slide-24
SLIDE 24

24

Background

  • IDLE/CCA mode power consumption in IEEE 802.11 [3, 4]

idle mix LNA fil amp ADC

P P P P P P     

ADC ANT

P Bandwidth N  

BP Filter LP Filter Mixer Mixer BP Filter Power AMP DAC LNA BP Filter Mixer BP Filter ADC LP Filter Mixer TX RF Circuitry RX RF Circuitry

CPU for Baseband Signal Processing

[3] S. Cui, A. J. Goldsmith, and A. Bahai, “Energy‐constrained Modulation Optimization,” IEEE Trans. Wireless Communications, 4(5), 2005. [4] J. Thomson and B. Baas, “An Integrated 802.11a Baseband and MAC Processor,” in Proc. IEEE Int’l Solid‐State Circuits Conf. (ISSCC) Digest

  • f Technical Papers, 2002.
slide-25
SLIDE 25

25

Measurement Environment

  • Network Interface Card (NIC)
  • Qualcomm Atheros 9880 (QCA 9880)
  • IEEE 802.11ac
  • 3 x 3, 80 MHz, 256QAM
  • Device driver
  • ath10k
  • 3.18.0 Linux kernel
  • Measurement tools
  • NI USB-6218 Data Acquisition (DAQ)
  • PEX1-MINI-E Adaptor
  • Current sense resistors (40 mΩ)
  • External power source (Power Monitor)
  • LabVIEW
slide-26
SLIDE 26

26

Measurement Result

  • Power consumption of QCA 9880 in Idle mode
  • The power consumption highly depends on BW and NANT
  • Power consumption of 160 MHz is obtained from our model

584 607 740 868 723 769 961 1265 855 926 1183 1661

slide-27
SLIDE 27

27

Measurement Result

  • Power consumption of QCA 9880 in RX mode
  • The power consumption highly depends on BW, MCS, and NANT

587 ~ 654 807 ~ 910 1000 ~ 1150 1137 ~ 1478 688 ~ 726 980 ~ 1170 1287 ~ 1590 2405 ~ 3087 834 ~ 922 1500 ~ 1780 2122 ~ 2690 3692 ~ 4714

slide-28
SLIDE 28

28

WiZizz: Energy Efficient Bandwidth Management

Source: Okhwan Lee, Jihoon Kim, and Sunghyun Choi, "WiZizz: Energy Efficient Bandwidth Management in IEEE 802.11ac Wireless Networks,“ in Proc. IEEE SECON 2015, Seattle, USA, June 22 ‐ 25, 2015.

slide-29
SLIDE 29

29

WiZizz: Energy Efficient Bandwidth Management

  • WiFi in Zizz (WiZizz)
  • Save the power consumption of STAs
  • More suited to portable devices than SMPS (e.g., Smartphone)
  • Key idea
  • Listen with the narrowest bandwidth (e.g., 20 MHz)
  • Transmit/receive data frames with a larger bandwidth (e.g., 160 MHz)

160 MHz DATA 160 MHz 20 MHz DATA 20 MHz Baseline WiZizz Bandwidth Time Time Save energy Bandwidth

slide-30
SLIDE 30

30

WiZizz: Energy Efficient Bandwidth Management

  • PHY-level Filtering
  • WiZizz STAs can ignore 40/80/160 MHz PPDUs addressed to others
  • Bandwidth information in the preamble
  • Reduce RX mode power consumption

DATA-STA1 ACK- STA1

AP STAs

ACK- STA2

IC RX TX

DATA- STA2

STA1 STA2

Power Power Time Time

Save energy

slide-31
SLIDE 31

31

WiZizz: Energy Efficient Bandwidth Management

  • Dynamic WiZizz
  • Switching delay (Dup, Ddown) < SIFS
  • Use RTS/CTS sequence to announce bandwidth switching and to set Network

Allocation Vector (NAV)

  • 20 MHz duplicate frame
  • Upward switching condition
  • If switching overhead is relatively small compared to the frame duration
  • Downward switching condition
  • End of data packet reception

80 20

( ) 3 ( )

RTS CTS data ack data ack

T T T r T SIFS T r T SIFS       

slide-32
SLIDE 32

32

WiZizz: Energy Efficient Bandwidth Management

  • Dynamic WiZizz example

Time Bandwidth DATA ACK 20 MHz Switch to 80 MHz Switch to 20 MHz RTS RTS RTS RTS BW of STA CTS

Network Allocation Vector (NAV)

Upward switching condition Downward switching condition

slide-33
SLIDE 33

33

WiZizz: Energy Efficient Bandwidth Management

  • Delay in switching bandwidth
  • QCA 9880
  • Average over 15 runs

BW NANT Upward (Dup) Downward (Ddown) 20↔80 3 73.6 48.53 20↔80 1 73.04 45.07 20↔40 3 40.53 22.67

slide-34
SLIDE 34

34

WiZizz: Energy Efficient Bandwidth Management

  • Pseudo-Dynamic WiZizz
  • Can be readily implemented with current hardware
  • Switching delay (Dup, Ddown) > SIFS
  • Use action frame to announce bandwidth switching and to set NAV
  • Upward switching condition
  • If switching overhead is relatively small compared to the frame duration
  • Downward switching condition
  • Receives the WiZizz action frame addressed to it
  • Receives a frame addressed to others, its duration is longer than the switching delay
  • Receives a frame with the more data bit in the frame control field set to 0

80 20

( ) ( )

action ack data up down data

T SIFS T T r D D T r      

slide-35
SLIDE 35

35

WiZizz: Energy Efficient Bandwidth Management

  • Pseudo-Dynamic WiZizz example

Time Bandwidth DATA

More Data = 0

20 MHz Switch to 80 MHz Switch to 20 MHz

NAV

Action Action Action Action

BW of STA

ACK ACK

Upward switching condition Downward switching condition

ACK ACK ACK

slide-36
SLIDE 36

36

WiZizz: Energy Efficient Bandwidth Management

  • Prototype and testbed experiments (QCA 9880)
  • Pseudo-dynamic WiZizz
  • Single node with various source rates
  • Multiple node environment
  • Saturated downlink traffic

Single node (80 MHz) Multiple node (80 MHz)

500 1000 1500 2000 0.01 0.1 1 10 Power consumption (mW) Source rate (Mb/s)

Baseline WiZizz

500 1000 1500 2000 2500 5 10 15 20 Power consumption (mW) Number of STAs

Baseline WiZizz

55% energy saving 25% energy saving

slide-37
SLIDE 37

37

WiZizz: Energy Efficient Bandwidth Management

  • Simulation (ns-3)
  • Dynamic WiZizz
  • Using measurement-based power model
  • Performance of 160 MHz bandwidth can be obtained
  • Saturated downlink traffic

80 MHz 160 MHz

1000 2000 3000 4000 5 10 15 20 Power consumption (W) Number of STAs

Baseline WiZizz

1000 2000 3000 4000 5 10 15 20 Power consumption (W) Number of STAs

Baseline WiZizz

73% energy saving 57% energy saving

slide-38
SLIDE 38

38

Conclusion (1/2)

  • Robust Wi-Fi in mobile environments
  • MoFA: Mobility-Aware Frame Aggregation in Wi-Fi
  • Standard-compliant adaptive frame aggregation control at transmitter
  • Prototyping using programmable 802.11n commercial device
  • Open-source linux device driver: Ath9k / iwlwifi
  • ChASER: Channel-Aware Symbol Error Reduction
  • Eliminate caudal losses by tracking channel variation at receiver
  • Receiver architecture modification
  • Prototyping using Microsoft Sora SDR platform
slide-39
SLIDE 39

39

Conclusion (2/2)

  • Energy efficient Wi-Fi
  • Power consumption of Wi-Fi
  • IEEE 802.11n/ac consume more energy than IEEE 802.11a/b/g
  • Impact of wider channel bandwidth on the power consumption is significant
  • WiZizz: Energy Efficient Bandwidth Management
  • Practical, standard-congenial bandwidth management
  • Achieve significant performance gain over the baseline 802.11ac
slide-40
SLIDE 40

Do You Have any Questions?

slide-41
SLIDE 41

Thank you

for your attention

slide-42
SLIDE 42

42

References

Seongho Byeon, Kangjin Yoon, Okhwan Lee, Woonsun Cho, Seungseok Oh, and Sunghyun Choi, "MoFA: Mobility-aware Frame Aggregation in Wi-Fi," in Proc. ACM CoNEXT 2014, Sydney, Australia, Dec. 2-5, 2014. Okhwan Lee, Weiping Sun, Jihoon Kim, Hyuk Lee, Bo Ryu, Jungwoo Lee, and Sunghyun Choi, "ChASER: Channel-Aware Symbol Error Reduction for High-Performance WiFi Systems in Dynamic Channel Environment,“ in Proc. IEEE INFOCOM 2015, Hong Kong, Apr. 26 - May 1, 2015. Okhwan Lee, Jihoon Kim, and Sunghyun Choi, "WiZizz: Energy Efficient Bandwidth Management in IEEE 802.11ac Wireless Networks,“ in Proc. IEEE SECON 2015, Seattle, USA, June 22 - 25, 2015.

slide-43
SLIDE 43

43

Back-up slides

slide-44
SLIDE 44

44

Impact of Mobility (1/2)

  • Impact of modulation
  • MCS 4 / MCS 7 (using amplitude modulation)

highly susceptible to mobility

  • Impact of 11n features
  • Spatial multiplexing (MCS 15) and

channel bonding (BW 40) are highly affected by the mobility

  • STBC (2 X 1) does not alleviate

the performance degradation

slide-45
SLIDE 45

45

Impact of Mobility (2/2)

  • Rate adaptation: Minstrel
  • Window-based rate adaptation algorithm

 A-MPDU length adaptation will increase the accuracy of Minstrel rate selection

  • Achieving the maximum throughput at

2 ms aggregation time bound

  • Malfunction for large aggregation

time bound due to high SFER for currently selected PHY rate

  • Undesirably using too high MCS index
  • Unnecessarily frequent MCS changes
slide-46
SLIDE 46

46

Measurement Result

  • Power consumption of QCA 9880 in TX mode
  • The power consumption highly depends on NANT
  • ≪
  • 1355

1360 1414 1544 2160 2157 2257 2434 2877 2896 3043 3280

slide-47
SLIDE 47

47

Measurement Result

  • Power consumption of QCA 9880 in TX mode
  • W.R.T TX power
slide-48
SLIDE 48

48

Related Work

  • Spatial Multiplexing Power Save (SMPS)
  • Save the power consumption of STAs
  • Static
  • Use a single antenna
  • Dynamic
  • Use a single antenna in IDLE/CCA mode
  • STA enables its additional antennas when it receives the start of a frame sequence (e.g., RTS) addressed to it

4 ANTs DATA 4 ANTs 4 ANTs

1 ANT DATA 4 ANTs 1 ANT Baseline SMPS # of ANTs Time Time Save energy

slide-49
SLIDE 49

49

Power Consumption Modeling

  • RX/IDLE listening mode power consumption in 802.11ac [1]
  • Non-linear regression analysis in SPSS
  • Average error rate is 0.329%

1 2

: Bandwidth (MHz) : Data rate (Mb/s)

idle rx rx f

P i N BW i N P BW r   

 

 

1 2 3 rx rx ss rx f

P N f N BW N r P        

OURS SS DS TS

IWL5300 (11n) 2.5 354 0.2 3.34 4.2 4.6 493.1 4.117 241.4 AR9380 (11n) 2.31 19.8 0.3 0.6 4.6 7 414.7 1.654 34.62 QCA9880 (11ac) 2.22 54.36 0.472 1.08 6.619 11.93 472.1 1.978 79.88

1

2

3

f

P

1

i

2

i

( )

S S

f N [1] C.‐Y. Li, C. Peng, S. Lu, and X. Wang, “Energy‐based rate adaptation for 802.11n,” in ACM Mobicom, Aug. 2012.