MIT 6 829 - Improving Loss Resilience in Wireless Networks

Unformatted text preview:

Improving Loss Resilience with Multi-Radio Diversity in Wireless NetworksThe problemBit-errors tend to clusterCurrent solutions are inefficient for recovering lossesSlide 5Multi-Radio Diversity (MRD) – UplinkMulti-Radio Diversity (MRD) – DownlinkAre losses independent among receivers?Slide 9Challenges in developing MRDBit-by-bit frame combiningBlock-based frame combiningFailure decreases with NB and burst sizeFlawed retransmission schemesRequest-for-acknowledgment (RFA) for efficient feedbackMRD-aware rate adaptationExperimental setupMRD improves throughputMRD maintains high bit-rateDelay AnalysisRelated workSummaryImproving Loss Resilience with Multi-Radio Diversity in Wireless NetworksAllen Miu, Hari BalakrishnanMIT Computer Science and Artificial Intelligence LaboratoryC. Emre KoksalComputer and Communication Sciences, EPFL2The problemNew wireless applications demand high performanceBut: wireless channels are loss-proneInterferenceNoise AttenuationMulti-pathMobility, etc… Inconsistent and poor performance3Bit-errors tend to clusterBit error positionCount# of corrupt frame samples = 185584Current solutions are inefficient for recovering lossesCoding (e.g., FEC)Hard on highly variable channelsRetransmissionWasteful: outage durations can be long (tens to hundreds of milliseconds)Bit-rate adaptationHard on variable channelsSlows down other clients5May use only one pathUses only one communication pathToday’s wireless LAN (e.g., 802.11)AP1Internet6May use only one pathAllow multiple APs to simultaneously receive transmissions from a single transmitterToday’s wireless LAN (e.g., 802.11)AP1AP2Multi-Radio Diversity (MRD) – UplinkInternetMRDC10%20%Loss independence simultaneous loss = 2%7Multi-Radio Diversity (MRD) –DownlinkInternetAP1AP2MRDCAllow multiple client radios to simultaneously receive transmissions from a single transmitter8Are losses independent among receivers?Broadcast 802.11 experiment at fixed bit-rate: 6 simultaneous receivers and 1 transmitterCompute loss rates for the 15 receiver-pair (R1, R2) combinationsFrame loss rate FLR(R1), FLR(R2) vs. simultaneous frame loss rate FLR(R1 ∩ R2)9Individual FLR > Simultaneous FLRFLR(R1 ∩ R2)FLRy = xR1R2R1*R210Challenges in developing MRDHow to correct simultaneous frame errors?Frame combiningHow to handle retransmissions in MRD?Request-for-acknowledgment protocolHow to adapt bit rates in MRD?MRD-aware rate adaptation112. Select bit combination at unmatched bit locations, check CRCPatterns CRC OkBit-by-bit frame combining1100 0000R11101 1010R20001 10101. Locate bits withunmatched value1100 0000 --1100 0010 X1100 1000 X1100 1010 OCorrected frameTX: 1100 1010 Problem: Exponential # of CRC checks in # of unmatched bits.Combine failure11012Block-based frame combiningObservation: bit errors occur in burstsDivide frame into NB blocks (e.g., NB = 6)Attempt recombination with all possible block patterns until CRC passes# of checks upper bounded by 2NBFailure rate increases with NB131.0Failure decreases with NB and burst sizeBurst error length parameterProbability of failureNB = 2NB = 4NB = 6NB = 16…0.80.60.40.20010 20 30 40 50Frame size = 1500B14Flawed retransmission schemesConventional link-layer ACKs do not workFinal status known only to MRDCTwo levels of ACKs are redundantCannot disable link-layer ACKs15Request-for-acknowledgment (RFA) for efficient feedbacklinkIPlinkMRDClinkMRD MRDDATAACK DATA RFAMRD-ACKDATAIP16MRD-aware rate adaptationStandard rate adaptation does not workReacts only to link-layer losses from 1 receiverUses sub-optimal bit-ratesMRD-aware rate adaptationReacts to losses at the MRD-layerImplication: First use multiple paths, then adapt bit rates.17Experimental setup~20 mR1R2L• 802.11a/b/g implementation in Linux (MADWiFi)• L transmits 100,000 1,500B UDP packets w/ 7 retries • 802.11a @ auto bit rate (6, 9, 12, 18, 24, 36, 48, 54)• L is in motion at walking speed, > 1 minute per trial• Variants: R1, R2, MRD (5 trials each)18MRD improves throughput2.3x Improvement R1 R2 MRDThroughput (Mbps)8.25 Mbps18.7 MbpsEach color shows a different trial19MRD maintains high bit-rateFrame recovery data(% of total losses at R1)via R2 42.3%frame combining 7.3%Total 49.6%Selected bit rate (Mbps)Fraction of transmitted frames609 12 18 24 36 48 5620Delay AnalysisOne way delay (s)Fraction of delivered packets10-310-4010-210-11User space implementation caused high delay21Related workPhysical layer spatial diversity techniquesAntenna diversity802.16 MIMO/802.11nMacro-diversity in CDMA networksRetransmission with memory [Sindhu ’77]Opportunistic forwarding [Biswas ‘05]Bit rate selection (AARF, RBAR, MiSer, OAR, Sample Rate)22SummaryDesign of Multi-Radio Diversity WLANBlock-based frame combiningRequest-for-acknowledgment protocolMRD-aware rate adaptationAnalysis of block-based frame combiningExperimental evaluationMRD reduces losses by 50% and improves throughput by up to


View Full Document

MIT 6 829 - Improving Loss Resilience in Wireless Networks

Download Improving Loss Resilience in Wireless Networks
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Improving Loss Resilience in Wireless Networks and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Improving Loss Resilience in Wireless Networks 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?