Performance Analysis of IEEE802.11 Distributed Coordination Function (DCF) Author : Giuseppe Bianchi Presented by: 李政修 December 23 , 2003 Outline Overview of IEEE 802.11 DCF Mathematical model Notations Bi-dimensional Markov Chain One step transition probabilities Stationary distribution Performance evaluation of DCF Conclusion and future work March 26, 2003 Math884 project/wqh 2 Overview of IEEE 802.11 MAC and PHY layers specifications for wireless LANs MAC Protocols Fundamental: Distributed Coordination Function (DCF) CSMA/CA based Binary Exponential Backoff rules Optional: Point Coordination Function (PCF) March 26, 2003 Math884 project/wqh 3 Overview of IEEE 802.11 DCF Two access techniques Basic mechanism: 2 way handshaking RTS/CTS mechanism: 4 way handshaking RTS DATA CTS Dest Source Source DATA DESt ACK ACK March 26, 2003 Math884 project/wqh 4 Example of RTS/CTS Access Scheme CSMA/CA based CSMA: listen at least DIFS before talk CA: defer transmission for random back-off time after DIFS BO=3 (set) BO=0 A DIFS RTS SIFS BO=5 (set) DATA CTS B BO=8 (set) BO=5 (suspend) C BUSY DIFS March 26, 2003 DIFS RTS BO=15 (set) DIFS collision ACK BO=5(resume) NAV (RTS) NAV(CTS) DIFS RTS Math884 project/wqh BO=10 (set) DIFS 5 Overview of IEEE 802.11 DCF Backoff procedure—BEB algorithm CW Backoff counter: CWmax •Initial: uni~[0,CW-1] •Non zero: decremented for ccs each idle slot •Zero: transmit March 26, 2003 CWmin C c c Math884 project/wqh s cc c c s t 6 Analytical Model to Evaluate DCF Saturation throughput: The limit reached by the system throughput as the offered load increases, and represents the maximum load that the system can carry in stable conditions Assumptions Constant & independent collision probability for each transmitted packet Ideal channel condition (no hidden terminals and capture) Fixed number of stations operated under overload (saturation condition) March 26, 2003 Math884 project/wqh 7 Bi-dimensional Markov Chain model Behavior of a single station Notations Time scale: Discrete and integer, t, beginning of a slot time, when backoff time counter decrements or regenerated [t, t+1], interval between 2 consecutive slot time, can be variable length Makovian State: B(t) ={s(t), b(t)} b(t): backoff time counter at time t s(t): backoff stage at time t i CWi = 2 CWmin m m: maximum backoff stage, CWmax = 2 CWmin p: prob.of each transmitted packet being collided March 26, 2003 Math884 project/wqh 8 B(t0)=(0,3) SIFS DIFS Station A BO=3 Station B B(t9)=(1,7) SIFS RTS t0 t1 t2 t3 DIFS BO=4 PACKET CTS t8 BO=5 RTS BO=7 t4 …... t8 ACK Busy channel t9 collision DIFS Others BO=7 ACK March 26, 2003 DIFS BO=4 Busy channel NAV(RTS) BO=4 RTS BO=2 RTS Busy channel NAV(CTS) Math884 project/wqh 9 One step transition probabilities (1) 1) P{i,k|i,k+1}=1, k : [0,Wi-2], i : [0,m] At beginning of t Backoff counter not reach zero, no transmission Channel sensed idle for 1 mini-slot till t+1 At beginning of t+1 Backoff counter decremented by 1 i,k March 26, 2003 1 i , k+1 Math884 project/wqh 10 One step transition probabilities (2) P{0,k|i,0}=(1-p)/W0, k : [0,W0-1], i : [0,m] At beginning of t At beginning of t+1 Backoff counter reaches zero, successful transmitted [t,t+1] Contention window reset to CWmin (backoff stage = 0) Backoff counter chosen randomly in [0,W0-1] P{i+1,k|i,0}= p/Wi+1, k : [0,Wi+1-1], i : [1,m-1] At beginning of t Backoff counter reaches zero, transmit in [t,t+1], collision Contention window < CWmax At beginning of t+1 contention window doubled Backoff counter chosen randomly in [0,Wi+1-1] March 26, 2003 Math884 project/wqh 11 State transits upon backoff counter reach zero (Contention Window <CWmax) Tx Success 0,0 0,1 . . . 0 , W0-2 0, W0-1 (1-p)/W0 i,0 i+1, 0 … p/Wi+1 i+1 , 1 ... i+1,Wi+1-2 i+1,Wi+1-1 collision March 26, 2003 12 One step transition probabilities (3) P{m,k|i,0}= p/Wm, k : [0,Wm-1], i = m At beginning of t Backoff counter reaches zero, transmit in [t,t+1], collision Contention Window = CWmax At beginning of t+1 Contention Window remains at CWmax Backoff time counter chosen randomly in [0,Wm-1] March 26, 2003 Math884 project/wqh 13 State transits upon backoff counter reach zero (Contention Window = CWmax) 0,0 0,1 . . . m,0 … 0 , W0-2 0, W0-1 (1-p)/W0 m,1 … m , Wm-2 m , Wm-1 p/Wm March 26, 2003 Math884 project/wqh 14 Bi-dimensional Markov Chain model One step transition diagram 15 Results obtained from the model (1) Stationary distribution tao, Probability of a station transmit in a randomly chosen slot time p, Probability of a transmitted packet Using numerical techniques to solve tao, p encounters a collision March 26, 2003 Math884 project/wqh 16 Results obtained from the model (2) System throughput: as a function of tao (similar to that derived in class) March 26, 2003 Math884 project/wqh 17 Length of Ts and Tcollision March 26, 2003 Math884 project/wqh 18 Performance evaluation of 802.11DCF (1) Parameters Basic Access RTS/CTS Network size Sensitive Insensitive Prob.tao Sensitive Insensitive CWmin Dependent Independent CWmax Marginal effect Negligible effect Packet size March 26, 2003 More effective for longer packets Math884 project/wqh 19 Performance evaluation of 802.11DCF (2) Network size vs. throughput Basic and RTS/CTS access schemes March 26, 2003 Math884 project/wqh 20 Performance evaluation of 802.11DCF (3) Probability tao vs. Throughput Basic Access March 26, 2003 RTS/CTS Math884 project/wqh 21 Performance evaluation of 802.11DCF (4) CWmin vs. Throughput Basic Access March 26, 2003 RTS/CTS Math884 project/wqh 22 Performance evaluation of 802.11DCF (5) CWmax vs. throughput March 26, 2003 Math884 project/wqh (CWmin = 32) 23 Performance evaluation of 802.11DCF (6) Packet length vs. throughput March 26, 2003 Math884 project/wqh 24 Conclusion and future work Major contributions of the introduced paper Proposed analytical model Accurate: verified by comparison with simulations Simple Account for all exponential backoff details Evaluate basic and RTS/CTS access schemes Performance evaluation on saturation throughput What to improve considering the upper limit of retransmission times March 26, 2003 Math884 project/wqh 25 Conclusion and future work Extend one hop to multihop For fixed topology Find a mathematical solution March 26, 2003 Math884 project/wqh 26