No Time to Countdown: Backing Off in Frequency Domain

No Time to Countdown: Backing Off in Frequency Domain

No Time to Countdown: Backing Off in Frequency Domain Souvik Sen, Romit Roy Choudhury, Srihari Nelakuditi Current WiFi Contention Random Backoff = 15 AP1 Channel AP2 R1 Random Backoff = 25 R2 B1=15 AP1

DIF S AP2 B2=25 Time 2 Current WiFi Contention Random Backoff = 0 AP1 AP2 R1 B1=15 Channel Random Backoff = 10 R2 B1=0

AP1 DIF S AP2 B2=25 B2=10 Time 3 Current WiFi Contention Transmit AP1 AP2 R1 B1=15 Channel Carrier Busy R2 B1=0

AP1 Data and ACK DIF AP2 S B2=25 B2=10 Wait Time 4 Current WiFi Contention Random Backoff = 18 AP1 AP2 R1 B1=15 Channel Random Backoff = 10

R2 B1=0 B1=18 AP1 Data and ACK DIF AP2 S B2=25 B2=10 B2=10 Wait Time 5 Current WiFi Contention Channel Random Backoff = 8 AP1

AP2 R1 B1=15 R2 B1=0 B1=18 AP1 Random Backoff = 0 B1=8 Data and ACK DIF AP2 S B2=25 B2=10 B2=10 B2=0 Wait

Time 6 Current WiFi Contention Channel Carrier AP1 AP2 Busy R1 B1=15 R2 B1=0 B1=18 AP1 B1=8 Data and ACK

DIF AP2 S B2=25 B2=10 Transmit B2=10 Wait Wait B2=0 Data and ACK Time 7 Current WiFi Contention Channel Carrier High channel wastage due to backoff R1

R2 35% overhead at 54Mbps AP1 B1=15 DIF AP2 S B1=0 B1=18 Data and ACK B2=10 Wait B1=8 B2=0 Wai t Data and ACK Time 8 Current WiFi Channel Contention Backoff is not fundamentally a time domain operation Its implementation is in time domain

Time Domain Frequency Domain Can we implement backoff in frequency domain? Are there any benefits in doing so? 9 Frequency domain contention resolution 802.11 a/g/n PHY adopts OFDM Wideband channel divided into 48 narrowband subcarriers Copes better with fast, frequency selective fading Purely a PHY motivation Subcarriers: 1234 48 Frequenc y We propose Back2F MAC Opportunity: Pretend OFDM subcarriers as integers Emulate randomized backoff 1

Back2F: Main Idea Replace temporal with subcarrier transmission 6 18 0 47 AP 1 0 47 AP2 Backoff =6 R1 Backoff = 18 R2 1

Back2F: Main Idea Replace temporal with subcarrier transmission Others Backoff = 18 6 0 Listen Antenna AP 1 Backoff =6 18 0 47 Others Backoff =6 47 AP2

Listen Antenna Backoff = 18 R1 Both APs learn AP1 is R2 the winner 1 Back2F: Scheduled Transmission Active subcarriers imply backoff chosen by other APs Each AP knows its rank in the sequence Enables back to back TDMA like transmission Self Back off 0 1 2 Other

s Back off 4 7 AP 1 Rank in TDMA: 3 1 Is there a benefit with frequency domain backoff? - 1500 bytes at 54Mbps ~ 250 micro sec. - Avg. temporal backoff ~ 100 micro sec. Big overhead Frequency backoff = 1 OFDM symbol = 4 micro sec 4 microsecond overhead dramatic performance enhancement relative to Will APs Collide During Contention? Introduce a second round of contention Winners of first go to second 0

1 Subcarrier 2 3 First Round 4 5 1 Will APs Collide During Contention? Introduce a second round of contention Winners of first go to second 0 1 Subcarrier 2 0 1 Subcarrier 2 3

First Round 3 4 5 4 5 Second Round 1 Only a Few APs in Second Round? TDMA will not be effective Optimize for TDMA Instead of only winners, a few more APs to second round 0 1 Subcarrier 2 0

1 Subcarrier 2 3 First Round 3 4 5 4 5 Second Round 1 Optimize for TDMA Instead of only winners, a few more APs to second round 0 02 4 1 Rank

1 0 1 Subcarrier 2 02 4 3 First Round Rank 2 2 4 Rank 3 3 4 5 02 4 Enablin

g TDMA 5 Second Round 1 Improved Channel Utilization 02 4 Rank 1 0 1 Subcarrier 02 4 Rank 2 2 Rank 3

3 4 02 4 Enablin g TDMA 5 2 Improved Channel Utilization 02 4 02 4 Rank 1 0 1 Subcarrier 2

Data/ ACK Frequency Backoff Data/ ACK Rank 2 Data/ ACK Rank 3 3 02 4 4 Data/ ACK 802.11: Contention per packet Enablin g

TDMA 5 ....... ... Time TDMA Data/ ACK Data/ ACK ....... ... Back2F: Contention per TDMA Schedule Time 2 Multiple Collision Domains Does Back2F work with real-world scattered APs? 2 Multiple Collision Domains Does Back2F work with real-world scattered APs?

BO = 5 BO = 3 BO = 2 BO = 1 Blue waits for Purple, but Purple waits for Green But Blue and Green should transmit simultaneously Lost transmission opportunity However 802.11 does not suffer from this problem Blue will wait for DIFS, continue counting down and eventually transmit 2 Multiple Collision Domains Does Back2F work with real-world scattered APs? BO = 5 BO = 3 BO = 2 BO = 1 Blue waits for Purple, but Purple waits for Green But Blue and Green should transmit simultaneously Back2F Solution: Emulate 802.11

Lost transmission opportunity However 802.11 does not suffer from this problem Blue will wait for DIFS, continue counting down and eventually transmit 2 Multiple Collision Domains BO = 5 BO = 2 BO = 3 BO = 1 Frequency Backoff G -> 1 DIFS B -> 2 DIFS R -> 5 4 P -> 2 Time 3

Channel idle > DIFS Frequency Backoff, Wait for turn My turn Reduce BO by winners BO Frequency Backoff Transmit 2 Back2F: Performance Evaluation Three important questions: Can Back2F detect subcarriers reliably? What is Back2Fs collision probability? How much throughput gain over 802.11? 2 Back2F: Performance Evaluation Three important questions: Can Back2F detect subcarriers

reliably? Evaluated on USRP/Gnuradio What is Back2Fs collision probability? How much throughput gain over 802.11? Evaluated using traces at 65 locations 2 Back2F: Performance Evaluation Three important questions: Can Back2F detect subcarriers reliably? Evaluated on USRP/Gnuradio Practical Transmit Antenna Challenge: High Self Signal Listen Antenna 2 Self Signal Overflows into Adjacent Subcarrier 6

60 0 S SNR in dB 50 5 0 40 pt.FFT 6464pt. FFT Self Signal d 4 i 30 0 B 20 n 30 10 NR 2 0

0 1 0 10 10 60 20 20 30 30 4040 50 50 60 Subcarrier Number Subcarrier Number 2 Solution: Use a Higher Point FFT 60

6 0 50 128pt. pt. FFT 128 FFT Self Signal 5 0 40 dB S SNR in d 4 i 30 0 B 20 n 30

10 NR 2 00 10 10 60 1 0 20 20 30 4040 30 Subcarrier Number 50 50 60 Subcarrier Number 3 Solution: Use a Higher Point FFT 60

6 0 50 256 256pt. pt.FFT FFT Self Signal 5 40 0 dB S SNR in d 4 i 30 0 B n 20 3 0 10 NR

2 00 10 10 60 1 0 20 20 30 4040 30 Subcarrier Number 50 50 60 Subcarrier Number 3 Subcarrier Detection Performance Detection Accuracy

1 0. 8 8dB SNR 0. 6 10dB SNR 0. 4 14dB SNR 0 0. 21 12dB SNR 2 8 3 4 5

6 7 Distance in Subcarriers Reliable subcarrier detection at 14dB Robust subcarrier detection at 14dB 3 Back2F: Performance Evaluation Collect traces to answer: What is Back2Fs collision probability? How much throughput gain over 802.11 AP Clien t 20 AP locations 45 client locations 3 Back2F: Performance Evaluation Collect traces to answer: What is Back2Fs collision probability?

How much throughput gain over 802.11 1. RSSI Collision Probability 2. Per Subcarrier SNR 3. Optimal Bitrate 4. Traffic pattern Emulate 802.11, Back2F for various topologies 3 Back2F: Collision Probability Collision Probability(%) 0.7 Back2F: 2 rounds Back2F: single round 802.11 0.6 0.5 Benefit of second round 0.4

0.3 0.2 0.1 0 5 45 10 50 15 20 25 30 35 40 Number of APs Small collision probability in dense networks 3 Throughput Evaluation 1 HD streaming

Skype traffic Web browsing CDF 0.8 0.6 0.4 0.2 0 0.1 0.6 Higher 0.2 0.3 0.4 0.5 Throughput gain over 802.11 throughput gain for real time traffic 3 Throughput gain (%) over 802.11

Throughput Evaluation 0.6 6 1Mbps 38 5Mbps 54 Mbps w 6 Mbps 4 Mbps /o batch 0.5 0.4 0.3 0.2 0.1 0 5 45 10 50 15 20

25 30 35 40 ThroughputNumber Gainofofclients Upto 50% 3 Limitation and Discussion Robustness of subcarrier based backoff Back2F more sensitive to channel fluctuation Need for additional antenna Back2F is complementary to MIMO Gain over packet aggregation Aggregation may not be possible for real time traffic Back2F provides gain with aggregation at higher rates Interoperability with 802.11 May interoperate but will cause unfairness 3 Summary Randomization is an effective method for contention resolution 802.11 time domain backoff requires channel to remain idle

Observation: randomization possible in frequency domain Using OFDM subcarriers Back2F: practical system realizing frequency domain contention Prevents collisions, provides upto 50% improvement in throughput 3 Questions, comments ? Thank you Duke SyNRG Research Group http://synrg.ee.duke.edu

Recently Viewed Presentations

  • FGC Software Design Review - CERN

    FGC Software Design Review - CERN

    Powering of complete arcs & arcs together. 7/8 sectors up to 5.5 TeV (sector 56 to 6.5 TeV) Robust performance. Tunnel closure - 5th September - LHC enters operation phase. Becomes clear that it is the intersect of a number...
  • Présentation PowerPoint - Overblog

    Présentation PowerPoint - Overblog

    It delivers more channels It delivers better pictures There is more interaction For marketers, this means better targeting and immediate opportunities to sell, to collect data and to develop relationships More chanel means a huge selection There is a channel...
  • Fraud Awareness Profiling Fraudsters and Current Fraud Schemes

    Fraud Awareness Profiling Fraudsters and Current Fraud Schemes

    The majority of occupational frauds were committed by staff at the employee or managerial level. Employee. Manager. Owner/ Executive. 42%. 36%. 19%. IF YOU WANT TO SCORE THE BIG BUCKS, YOU NEED TO GET INTO A POSITION OF TRUST AND...
  • Interaksi Obat - Fk Uwks 2012 C

    Interaksi Obat - Fk Uwks 2012 C

    Interaksi obat dengan makanan. 11. Tetracylin + susu /makanan Fe, Ca, Mg membentuk ikatan Kompleks tidak bisa diabsorpsi. 12. Thyroid + kubis, wortel, kecambah, kool, sayur hijau, buah pir, lobak cina (Thiooxazolidone) menghambat aktivitas hormon thyroid.
  • NPV-Questions - H. Zafer Yuksel

    NPV-Questions - H. Zafer Yuksel

    The Yurdone Corporation wants to set up a private cemetery business. According to the CFO, Barry M. Deep, business is "looking up". As a result, the cemetery project will provide a net cash inflow of $106,000 for the firm during...
  • PRESENTATION NAME - Brain-Targeted Teaching

    PRESENTATION NAME - Brain-Targeted Teaching

    Why Tuck Everlasting? ... Students will use facial expressions to represent how they think Winnie reacted when she found out the Tucks could live forever. ... Students will use visualization to paint a picture of the forest where the spring...
  • Ashford District Governor Briefing Autumn 2015 Welcome Agenda

    Ashford District Governor Briefing Autumn 2015 Welcome Agenda

    2016/20 Draft Commissioning Plan. Demand for school places is mainly caused by inward migration connected to house-building and the birth rate (at 2 to 3 points above the County and National rates).
  • 課程名稱 - ocw.aca.ntu.edu.tw

    課程名稱 - ocw.aca.ntu.edu.tw

    政治學. 授課教師:國立臺灣大學 政治學系 王業立 教授 【本著作除另有註明外,採取創用 CC 「姓名標示-非商業性-相同方式分享 」 臺 灣 3.0 版 授權釋出】 第十八講:政體(四) 政治文化(一) 本課程指定教材為Michael G. Roskin, Robert L. Cord, James A. Medeiros,