Isolating Physical PER for Smart Rate Selection in 802.11

Isolating Physical PER for Isolating Physical PER for Smart Rate Selection in Smart Rate Selection in 802.11802.11

Malik Ahmad Yar Khan and Darryl VeitchMalik Ahmad Yar Khan and Darryl Veitch

ARC Special Centre for Ultra-Broadband Information NetARC Special Centre for Ultra-Broadband Information Networks (CUBIN)works (CUBIN)

The University of Melbourne, AustraliaThe University of Melbourne, Australia

IEEE Infocom 2009IEEE Infocom 2009(acceptance rate = 19.65%) (acceptance rate = 19.65%)

OutlineOutline

IntroductionIntroduction The core ideaThe core idea RCA modificationRCA modification

– Fragmentation implementationFragmentation implementation– RTS/CTS implementationRTS/CTS implementation

PerformancePerformance ConclusionConclusion

IntroductionIntroduction

Rate selection algorithms in use in 8Rate selection algorithms in use in 802.11 suffer from a fundamental pro02.11 suffer from a fundamental problemblem– Implicity or explicity make decisions baImplicity or explicity make decisions ba

ses on total Packet Error Rate (PER)ses on total Packet Error Rate (PER) False rate decision lowers the systeFalse rate decision lowers the syste

m throughputm throughput

IntroductionIntroduction

Total PER = PPER + CPERTotal PER = PPER + CPER– PPER: Physical PERPPER: Physical PER– CPER: contention PERCPER: contention PER

It is often the case CPER It is often the case CPER dominates PPERdominates PPER– Poor latency, loss and throughputPoor latency, loss and throughput

PER under a ideal PER under a ideal environmentenvironment

Goal of the paperGoal of the paper

Providing a novel way to isolate Providing a novel way to isolate the physical errorthe physical error

Providing a modular way of Providing a modular way of isolating the contention error isolating the contention error which any rate adaptation which any rate adaptation algorithm can be enhancedalgorithm can be enhanced

FragmentationFragmentation

Only consider a packet is Only consider a packet is fragmented into 2 framesfragmented into 2 frames

The fragment 2 is sent out only The fragment 2 is sent out only after an SIFS after the final ACKafter an SIFS after the final ACK– Provided there are no hidden terminalsProvided there are no hidden terminals

FragmentationFragmentation

If the first fragment is lost, for nay If the first fragment is lost, for nay reason, the second will not even reason, the second will not even be sentbe sent

Observation: the loss of a second Observation: the loss of a second fragment on the first transmission fragment on the first transmission attemptattempt– Physical errorPhysical error

RTS/CTSRTS/CTS

Once the RTS-CTS exchange is Once the RTS-CTS exchange is completed, the data packet will completed, the data packet will have the channel reserved, and have the channel reserved, and will only encounter error for will only encounter error for physical reasonsphysical reasons

Block reservation in Block reservation in 802.11e802.11e

An option in IEEE 802.11eAn option in IEEE 802.11e The second data packet of such The second data packet of such

blocks could be used to directly blocks could be used to directly measure PPERmeasure PPER

Main conceptMain concept

Not to use these methods all the Not to use these methods all the time in practicetime in practice

Only trigger their use often Only trigger their use often enough to gather good statistics enough to gather good statistics on PPERon PPER

Scatterplots of PER with nScatterplots of PER with no cross traffico cross traffic

Scatterplots of PER with CScatterplots of PER with CT and fragmentationT and fragmentation

Scatterplots of PER with CT and Scatterplots of PER with CT and fragmentation (RTS/CTS on)fragmentation (RTS/CTS on)

Rate dependence of IP Rate dependence of IP throughputthroughput

Fragmentation Fragmentation implementationimplementation Only descriptors of second fragmentOnly descriptors of second fragment

s are passed to the RCAss are passed to the RCAs A mechanism is needed to trigger thA mechanism is needed to trigger th

e fragmentation of selected packetse fragmentation of selected packets– Maintains a simple ON/OFF variableMaintains a simple ON/OFF variable

RTS/CTS RTS/CTS implementationimplementation RTS/CTS protection is switched on RTS/CTS protection is switched on

by a per-packet decisionby a per-packet decision Use Long Retry Count (LRC) to Use Long Retry Count (LRC) to

record the statistic of RTS/CST record the statistic of RTS/CST transmissiontransmission

PerformancePerformance

Two main metricsTwo main metrics– Weighted mean rate summariesWeighted mean rate summaries– IP throughputIP throughput

All experiments are run in IEEE 802.11a iAll experiments are run in IEEE 802.11a in testbedn testbed– Full rate set:{6,9,12,18,24,36,48,54}Full rate set:{6,9,12,18,24,36,48,54}– Packet size: 1500 bytesPacket size: 1500 bytes– 10 stations if no further explanation10 stations if no further explanation

The wireless testbedThe wireless testbed

St[1-13] are stations, Sniff[1-2] St[1-13] are stations, Sniff[1-2] are sniffersare sniffers

Two implementation Two implementation protocolsprotocols SampleRateSampleRate

– A MIT master thesisA MIT master thesis– Implemented on MadWifiImplemented on MadWifi

AMRRAMRR– A multi-rate algorithm also implementeA multi-rate algorithm also implemente

d on MadWifid on MadWifi

Single user with different RCAsSingle user with different RCAs

TPER for single user TPER for single user case for AMRRcase for AMRR

SampleRate with different paSampleRate with different packet sizes (1472B)cket sizes (1472B)

SampleRate with different paSampleRate with different packet sizes (716B)cket sizes (716B)

UDP upstream system UDP upstream system throughputthroughput

TCP upstream system throughputTCP upstream system throughput

ConclusionConclusion

The paper investigates two ways The paper investigates two ways to isolate Physical PERto isolate Physical PER– RTS/CTSRTS/CTS– MAC level packet fragmentationMAC level packet fragmentation

Higher channel rates were used Higher channel rates were used when PPER was in fact low, even when PPER was in fact low, even when contention is presentwhen contention is present

Thank you!!Thank you!!