Rate adaptation in 802.11

Mateusz Wielgosz

Outline

Multiple rates in 802.11

Access scheme

Variants of 802.11 standard

Rates and modulation schemes

Performance anomaly of 802.11

Rate Adaptation algorithms

o ARF and AARF

o Sample Rate

o OAR (ERA)

o CARA / LDRA

Future directions

References

Multiple rates in 802.11

Initially only one base rate (1997).

Multirate enhancement since (1999)

Higher rates restricted by SNR.

Need for rate adaptation mechanisms.

Multitude of rate adaptation techniques, resulting from various metrics –

throughput, airtime fairness, power consumption, overall efficiency of network…

Algorithms include statistical approach, channel conditions measurement, loss

differentiation, machine learning, opportunistic attempts, ant colony

optimization...

Access scheme

RTS/CTS is an optional mechanism, introducing some

overhead, reducing chances of collisions.

Fragmentation mechanism.

Variants of 802.11 standard

802.11 (1997)

802.11a (1999)

802.11b (1999)

802.11g (2003)

802.11-2007 (a, b, d, e, F, g, h, i, j)

802.11n (2009)

802.11-2012 (k, r, y, n, w, p, z, v, u, s)

802.11ac (2012)

Rates and modulation schemes

Example for 802.11a, BER=10-5:

Rate (Mbp/s) SINR (dB) Modulation scheme Coding rate

6 6.02 BPSK 1/2

9 7.78 BPSK 3/4

12 9.03 QPSK 1/2

18 10.79 QPSK 3/4

24 17.04 16-QAM 1/2

36 18.80 16-QAM 3/4

48 24.05 64-QAM 2/3

54 24.56 64-QAM 3/4

Different modulation schemes and coding rates may produce non-linear

performance-SINR ratio. Assumption that next higher rate is “better".

Performance anomaly of 802.11

Degraded performance of hosts using higher rates in presence of hosts using

lower rates.

Basic CSMA/CA admission control gives all users equal access probability.

Users with lower rates occupy medium for longer periods of time than high-rate

users that transmit data packets in short time intervals.

Troublesome for rate adaptation algorithms without loss differentiation, it

resulted in various attempts at improving temporal fairness among users.

Rate adaptation algorithms

decrease rate decision increase rate decision RTS/CTS decision entiety Loss Diff. Type Year

ARF 2 missed frame 10 successful transmissions no sender STA - Loss 1997

AARF 2 missed frame x successful transmissions no sender STA - Loss 2004

AMRR +33% fail 10%- fail no sender STA - Loss 2004

ARC failure while cw >= OPTcw success while cw=< OPTcw no sender STA - Loss 2009

Onoe 50% frames lost in 1 second window success ratio (credit system) no sender STA - Loss 2004

RBAR channel conditions channel conditions yes receiver STA - SNR 2001

OAR channel conditions* channel conditions* yes* receiver STA* -* SNR* 2002

RSSLA RSS compared with adaptive Threshold RSS compared with adaptive Threshold no sender STA - RSS 2003

FAR channel conditions channel conditions yes hybrid - SNR 2005

SampleRate four succesive failures probing of promising rates no sender STA - Loss 2005

BARA channel conditions channel conditions no sender STA - SNR 2007

TA-ARA - when traffic can't be satisfied yes sender STA - SNR 2010

Minstrel probing of promising rates probing of promising rates no sender STA - Loss 2011

LD-ARF 2 missed frame due to channel fading 10 successful transmissions optional receiver STA + Loss 2005

CARA 2 fail 10 successful transmissions yes* sender STA + Loss 2006

RRAA loss ratio compared with current threshold loss ratio compared with current threshold yes * sender STA + Loss 2006

ERA frame missed due to channel fading x ACK (jak w AARF) no sender STA + Loss 2007

LDRA frame missed due to channel fading channel conditions no sender STA + SNR 2008

AARF-CD 2 fail 10 successful transmissions yes* sender STA + Loss 2008

SARA sukcesy i porażki w transmisji sukcesy i porażki w transmisji yes sender STA + Loss 2008

MutFed 2 missed frame channel conditions no hybrid + SNR 2010

MiSer optimal power consumption optimal power consumption yes sender STA - SNR 2003

Smart Sender probing and transmission history probing and transmission history no sender STA - Loss/SNR 2008

TFRC fairness criteria fairness criteria yes sender STA* - SNR 2009

ARF and AARF

Automatic Rate Fallback (1997), Adaptive Automatic Rate Fallback (2004).

ARF: Increase rate after 10 consecutive successful transmissions. If first

transmission at higher rate fails – return to previous rate. Decrease rate after 2

consecutive failed transmissions.

AARF: Increase rate after reaching T - threshold for rate increase. If first

transmission at higher rate fails – return to previous rate, double T. Decrease rate

after 2 consecutive failed transmissions.

Reduces unnecessary attempts in stable conditions.

Sample Rate

Keeps statistics for last 10-second period of transmission.

Periodically checks performance of other promising rates.

Decrease rate after 4 consecutive failures.

Every 10th frame, probe promising rates.

Transmission time of 1500 bytes in ideal channel = 1.873ms at 11Mbps, 2.976ms at 5.5Mbps, 6.834 at

2Mbps, 12.995ms at 1Mbps. If due to occasional failures, average transmission time at 11Mbps is 3.123ms,

then 5.5 Mbps will be probed as potentially better rate than 11Mbps.

OAR and ERA

Opportunistic Auto Rate (OAR), Effective Rate Adaptation (ERA).

OAR: “meta algorithm” (RBAR), aiming at airtime fairness. Using fragmentation

mechanism send base_rate/transmission_rate frames at once.

Example – 1 frame at 2Mbps, 5 frames at 11Mbps.

ERA uses fragmentation mechanism to determine cause of failed transmission.

Failed packet is divided into short and remaining fragment. If short fragment is

retransmitted at previous rate. If it’s successful, failure is attributed to collision. If

not, it’s retransmitted at basic rate. If this transmission is successful, previous

failure is attributed to channel fading. If not, it’s deemed to have been caused by

collision.

CARA and LDRA

Collision Aware Rate Adaptation (CARA), Loss Differentiation Rate Adaptation (LDRA)

CARA: Employs RTS/CTS mechanism, but only if ACK is missed. RTS exchange is

held at basic rate, and assumption is, that all RTS failures occur only due to

collision. So if it fails, collision is deduced and consecutive failure counter is not

incremented.

LDRA: Similar mechanism, but instead of RTS, it attempts at retransmission of lost

packet at basic rate to determine cause of failure.

Future directions

QoS supporting rate adaptation algorithm.

Devise algorithm that adapts rate according to both channel conditions and

traffic demands.

It would provide sufficient bit rate, delay, jitter, BER…

References

1. Lacage, M., Manshaei M.H., Turletti, T. “IEEE 802.11 Rate Adaptation: A Practical Approach”

2. Kamerman, A., Monteban, L. “WaveLAN-II: A High-Performance Wireless LAN for the Unlicensed Band”

3. Jongseok K., Seongkwan K. Sunghyun C., Daji Q. “CARA: Collision-Aware Rate Adaptation for IEEE 802.11

WLANs”

4. Biaz S. , Wu S. “ERA: An Efficient Rate Adaptation Algorithm with Fragmentation”

5. Wu S., Biaz S. “Loss Differentiated Rate Adaptation in Wireless Networks”

6. Sadeghi B., Kanodia V., Sabharwal A., Knightly E. “Opportunistic Media Access for Multirate Ad Hoc

Networks”

7. Heusse M., Rousseau F., Berger-Sabbatel G., Duda A. “Performance Anomaly of 802.11b”

8. John C. Bicket “Bit-rate Selection in Wireless Networks”

9. Labiod H., Afifi H., De Santis C. “Wi-Fi, Bluetooth, Zigbee and WiMax”

10. Ao X., Jiang, S., Tang L. “Traffic-aware Active Link Rate Adaptation via Power Control for Multi-hop Multi-

rate 802.11 Networks”