Enhancing IEEE 802.11 MAC in congested environments Imad Aad, Qiang - PowerPoint PPT Presentation

Enhancing IEEE 802.11 MAC in congested environments Imad Aad, Qiang Ni, Chadi Barakat, Thierry Turletti ASWN, Boston-MA, USA August 9 th , 2004 Enh. 802.11 – p.1

Outline IEEE 802.11 Very brief description Mathematical model description Enhacement Related work Slow decrease (SD) Performance Evaluation Enh. 802.11 – p.2

MAC sub-layer Time DIFS Data Source (Tx) CW SIFS ACK Destination (Tx) DIFS Contention Window Other NAV Defer access = NAV+DIFS Backoff Enh. 802.11 – p.3

MAC sub-layer backoff = rand () × CW Collision → equal backoffs → too many nodes → Should increase CW !! at the i th retransmission: CW ( i ) = CW min × 2 i at a successful transmission: CW = CW min Enh. 802.11 – p.3

MAC Throughput Model [Bianchi] S = E [ payload − information − transmitted − in − a − slot − time ] E [ length − of − a − slot − time ] Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] S = E [ payload − information − transmitted − in − a − slot − time ] E [ length − of − a − slot − time ] P s P tr E [ P ] S = (1 − P tr ) σ + P tr P s T s + P tr (1 − P s ) T c Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] S = E [ payload − information − transmitted − in − a − slot − time ] E [ length − of − a − slot − time ] P s P tr E [ P ] S = (1 − P tr ) σ + P tr P s T s + P tr (1 − P s ) T c Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] S = E [ payload − information − transmitted − in − a − slot − time ] E [ length − of − a − slot − time ] P s P tr E [ P ] S = (1 − P tr ) σ + P tr P s T s + P tr (1 − P s ) T c P s = nτ (1 − τ ) n − 1 = nτ (1 − τ ) n − 1 1 − (1 − τ ) n P tr P tr = 1 − (1 − τ ) n Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] To find τ , 2 nonlinear equations to solve, 1: p = 1 − (1 − τ ) n − 1 Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] To find τ , 2 nonlinear equations to solve, 2: (1−p)/W 0 1 1 1 1 0,0 0,1 0,2 0,W −2 0,W −1 0 0 i−1,0 p/W i 1 1 1 1 i,0 i,1 i,2 i,W −2 i,W −1 i i p/W i+1 p/W m 1 1 1 1 m,0 m,1 m,2 m,W −2 m,W −1 m m p/W m Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] To find τ , 2 nonlinear equations to solve: p = 1 − (1 − τ ) n − 1 2(1 − 2 p ) τ = (1 − 2 p )( W +1)+ pW (1 − (2 p ) m ) Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] To find τ , 2 nonlinear equations to solve: p = 1 − (1 − τ ) n − 1 2(1 − 2 p ) τ = (1 − 2 p )( W +1)+ pW (1 − (2 p ) m ) → Matlab → very close to simulations Enh. 802.11 – p.4

MAC Throughput Model [Bianchi] 850 802.11, simul 802.11, model SD, δ = 0.5, model SD, δ = 0.5, simul SD, δ = 0.25, model SD, δ = 0.25, simul 800 Total throughput (KBytes/s) 750 700 650 600 5 10 15 20 25 30 35 40 45 50 Number of contending flows, n Enh. 802.11 – p.4

Outline Enh. 802.11 – p.5

CW slow decrease After each collision, CSMA/CA increases CW Upon a successful transmission, reset CW BUT! congestion did not “reset”! Enh. 802.11 – p.6

CW slow decrease To reset or not to reset, that is the question! Enh. 802.11 – p.7

Related work In 1994, Bharghavan et al. proposed MACAW: MILD: Multiplicative Increase ( CW = CW × 1 . 5 ) Linear Decrease ( CW = CW − 1 ) Enh. 802.11 – p.8

Related work In 1994, Bharghavan et al. proposed MACAW: Enh. 802.11 – p.9

Related work In 1994, Bharghavan et al. proposed MACAW: Enh. 802.11 – p.10

Our approach We propose a slow CW decrease mechanism (SD), e.g. CW = 0 . 9 × CW Enh. 802.11 – p.11

Simulation scenario Simulation time (sec): 42 0 50 100 150 200 Enh. 802.11 – p.12

Simulation scenario Simulation time (sec): 44 0 50 100 150 200 Enh. 802.11 – p.12

Simulation scenario Simulation time (sec): 50 0 50 100 150 200 Enh. 802.11 – p.12

Simulation scenario Simulation time (sec): 100 0 50 100 150 200 Enh. 802.11 – p.12

Simulation scenario Simulation time (sec): 140 0 50 100 150 200 Enh. 802.11 – p.12

Simulation scenario Simulation time (sec): 150 0 50 100 150 200 Enh. 802.11 – p.12

Throughput vs. n 250 SD, basic, qlen = 2 802.11, basic, qlen = 2 No decrease, basic, qlen = 2 200 Throughput (KBytes/s) 150 100 50 0 0 50 100 150 200 250 300 Time (s) Enh. 802.11 – p.13

Settling time vs. δ 1.2 Eq. (8), pkt-size = 1050 Simul, λ =1, pkt-size = 1050 1 0.8 Settling time, T l , (s) 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 Multiplicative factor, δ Enh. 802.11 – p.14

Delays vs. n 0.9 "delays_09_comm_noRTS_qlen2.dat" "delays_noenh_comm_noRTS_qlen2.dat" 0.8 0.7 0.6 Packet delay (s) 0.5 0.4 0.3 0.2 0.1 0 0 50 100 150 200 250 300 Time (s) Enh. 802.11 – p.15

Throughput gain vs. δ 1.4 Basic, λ = 1, pkt-size = 1050 RTS/CTS, λ = 1, pkt-size = 1050 Basic, λ = 0.1, pkt-size = 105 1.35 1.3 1.25 Throughput gain, G 1.2 1.15 1.1 1.05 1 0.95 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Multiplicative factor, δ Enh. 802.11 – p.16

802.11 throughput model (1−p)/W 0 1 1 1 1 0,0 0,1 0,2 0,W −2 0,W −1 0 0 i−1,0 p/W i 1 1 1 1 i,0 i,1 i,2 i,W −2 i,W −1 i i p/W i+1 p/W m 1 1 1 1 m,0 m,1 m,2 m,W −2 m,W −1 m m p/W m Enh. 802.11 – p.17

SD throughput model (1−p)/W 0 1 1 1 1 0,0 0,1 0,2 0,W −2 0,W −1 0 0 i−1,0 p/W i 1 1 1 1 i,0 i,1 i,2 i,W −2 i,W −1 i i p/W i+1 p/W m 1 1 1 1 m,0 m,1 m,2 m,W −2 m,W −1 m m p/W m Enh. 802.11 – p.18

Throughput vs n 850 802.11, simul 802.11, model SD, δ = 0.5, model SD, δ = 0.5, simul SD, δ = 0.25, model SD, δ = 0.25, simul 800 Total throughput (KBytes/s) 750 700 650 600 5 10 15 20 25 30 35 40 45 50 Number of contending flows, n Enh. 802.11 – p.19

Throughput Gain vs. CW min 1.3 simul, n=5 model, n=5 simul, n=20 model, n=20 1.25 simul, n=50 model, n=50 1.2 Throughput gain of SD 1.15 1.1 1.05 1 0.95 0 20 40 60 80 100 120 140 CWmin Enh. 802.11 – p.20

802.11 Fairness, varying CW min 1 0.9 0.8 Average Jain fairness index 0.7 0.6 0.5 0.4 0.3 0.2 10 flows, CWmin = 32 10 flows, CWmin = 63 10 flows, CWmin = 127 0.1 0 500 1000 1500 2000 Window size Enh. 802.11 – p.21

802.11 Fairness, varying n 1 0.9 0.8 0.7 Average Jain fairness index 0.6 0.5 0.4 0.3 0.2 10 flows 0.1 25 flows 50 flows 80 flows 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Window size Enh. 802.11 – p.22

SD Fairness, varying n 1 0.9 0.8 0.7 Average Jain fairness index 0.6 0.5 0.4 0.3 0.2 SD, 10 flows SD, 15 flows SD, 20 flows 0.1 SD, 40 flows SD, 50 flows SD, 80 flows 0 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Window size Enh. 802.11 – p.23

Fairness comparison 1 0.9 0.8 0.7 Average Jain fairness index 0.6 0.5 0.4 0.3 0.2 802.11, 10 flows 0.1 SD, 10 flows 802.11, 80 flows SD, 80 flows 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Window size Enh. 802.11 – p.24

Coexisting SD and 802.11 30 10 flows, 802.11, simul 10 flows, SD λ = 0.5, simul 20 flows, 802.11, simul 20 flows, SD λ = 0.5 , simul 10 flows, 802.11, model 25 10 flows, SD λ = 0.5, model 20 flows, 802.11, model 20 flows, SD λ = 0.5, model Throughput/node (KBytes/s) 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 Proportion of 802.11 nodes Enh. 802.11 – p.25

Energy consumption 11 10 9 8 Energy/Bit (x10e-6 Joules) 7 6 5 4 3 2 802.11, Tx 1 SD, Tx 802.11, Rx SD, Rx 0 0 5 10 15 20 25 30 Number of contending flows Enh. 802.11 – p.26

On the application layer, FTP 300 250 200 FTP duration (s) 150 100 50 802.11 SD 0 0 5 10 15 20 25 30 Number of contending flows Enh. 802.11 – p.27

Noisy channel 1.4 1.3 1.2 1.1 Throughput gain of SD 1 0.9 0.8 0.7 0.6 1 flow 4 flows 15 flows 0.5 25 flows 40 flows 50 flows 0.4 0 0.02 0.04 0.06 0.08 0.1 Packet Error Rate (PER) Enh. 802.11 – p.28

Conclusion Deep analysis of simple Slow Decrease (SD) functions SD outperforms 802.11 in: throughput delay fairness (if congested) battery consumption etc. 802.11 outperforms SD if channel is severely noisy Enh. 802.11 – p.29

The End Thank you! ... questions ? imad.aad@epfl.ch Enh. 802.11 – p.30