2014 YU-ANTL Lab Seminar Performance Analysis of the IEEE 802.11 Distributed Coordination Function Giuseppe Bianchi April 12, 2014 Yashashree Jadhav Advanced Networking Technology Lab. (YU-ANTL) Dept. of Information & Comm. Eng, Graduate School, Yeungnam University, KOREA (Tel : +82-53-810-3940; Fax : +82-53-810-4742 http://antl.yu.ac.kr/; E-mail : yashashree@ynu.ac.kr)
Outline (1) Background MAC DCF Basic Access Mechanism RTS/CTS Mechanism Main Idea Contribution Markov Model Probabilities Two Dimensional Markov chain Packet Transmission Probability Throughput
Outline (2) Basic Access Mechanism RTS/CTS Access Mechanism Model Validation & Simulation Model Validation Maximizing Saturation Throughput Throughput vs Number of Stations Throughput vs Initial Window Size Throughput vs Max. Back‐off Stage Throughput vs Packet Length Conclusion
MAC (1) IEEE802.11 is a set of standards for wireless local area network (WLAN) This paper’s interest is in MAC layer The MAC layer is a set of protocols which is responsible for maintaining order in the use of a shared medium The MAC layer defines two different access methods The Distribution Coordination Function (DCF) Random access scheme Based on CSMA/CA Protocol The Point Coordination Function (PCF) Based on TDMA Paper focus on DCF
MAC (2) WLAN MAC and PHY Layer
DCF (1) When a station wants to transmit a new packet Monitor the channel activity If senses idle for DIFS (Distributed Inter Frame Space), the station transmits CSMA/CA If sensed busy (immediately or during the DIFS),the station persists to monitor until it is measured idle for DIFS The station generates a random back‐off interval before transmitting to minimize the collision probability
DCF (2) It describes two techniques to employ for packet transmission Basic access mechanism (two‐way handshaking) Source transmits the packet If destination receives successfully transmits a positive ACK RTS/CTS mechanism (four‐way handshaking) Source sends RTS If destination receives RTS then sends CTS So the channel reservation is done Source then transmits the packet
DCF (3) IEEE 802.11 DCF At each packet transmission, the back‐off time is uniformly chosen in the range(0,w‐1) where w=contention window w depends on the number of transmissions failed for the packet At first, w=CWmin (minimum contention window) At each unsuccessful, w is doubled (binary back‐off) up to a maximum value CWmax=2mCWmin The back‐off time counter is Decremented as long as channel is sensed idle Frozen when a transmission is detected on the channel Reactivated when the channel is sensed idle for more than a DIFS The station transmits when the back‐off time reaches zero
Basic Access Mechanism station has to wait for DIFS before sending data receiver acknowledges at once (after waiting for SIFS) if the packet was received correctly (CRC) automatic retransmission of data packets in case of transmission errors
RTS/CTS Access Mechanism station can send RTS with reservation parameter after waiting for DIFS ( reservation determines amount of time the data packet needs the medium) acknowledgement via CTS after SIFS by receiver (if ready to receive) sender can now send data at once, acknowledgement via ACK other stations store medium reservations distributed via RTS and CTS
802.11 – Slot Time in Bianchi’s Model
Contribution Analytical evaluation of the saturation throughput Ideal channel conditions (no hidden terminals and capture) Fixed number of stations where each station having a packet available for transmission Behavior of single station is studied with a Markov model The packet transmission probability (τ) of a station in randomly chosen slot time is obtained which is independent of access mechanism The throughput of the both access mechanism is expressed as a function of τ In saturation, each station has immediately a packet available for transmission Each packet needs to wait for a random back‐off time before transmitting At each transmission attempt each packet collides with constant and independent probability (p)
Markov Model (1) s(t) : stochastic process of back‐off stage of a station at time t b(t): stochastic process of back‐off time counter for a station Defines W=CWmin m=maximum back‐off stage such that CWmax=2mW Wi= 2iW where i Є(0,m) is the back‐off stage It is possible to model the bi‐dimensional process {s(t),b(t)} with the discrete‐timeMarkov chain
Markov Model (2) Probabilities P{i, k |i, k+1}=1 k Є (0,Wi ‐2) and i Є (0, m) At the beginning of each slot time the back‐off time is decremented P{0, k |i, 0}=(1-p)/W0 k Є (0,W0 ‐1) and i Є (0, m) New packet following a successful transmission (probability=1‐p) and starts with back‐off stage 0.The back‐off is initially chosen between (0, W0‐1) P{i, k |i-1, 0}=p/Wi k Є (0,Wi ‐1) and i Є (1, m) Unsuccessful transmission (probability=p) occurs at back‐off stage i-1,The new back‐ off is uniformly chosen between (0, Wi) P{m, k |m, 0}=p/Wm k Є (0,Wm ‐1) Once the back‐off stage reaches the value m, it is not increased in subsequent packet transmission
Markov Model (3) Two Dimensional Markov chain
Markov Model (4) Packet Transmission Probability bi, k= lim t-> ∞ P{s (t)=i, b(t)=k} , k Є (0,Wi ‐1) and i Є(0,m) Stationary distribution of the chain Closed‐form solution is needed All the bi, k values can be expressed as functions of the values b0,0 and p τ = probability that a station transmits in a randomly chosen slot time transmission occurs when back‐off counter=0 regardless of the back‐off stage
Markov Model (5) Packet Transmission Probability When m=0 (no exponential back‐off) One station transmits, collision occurs when at least one of the other n‐1station transmits Using the two equations it can be derived that τ (p) Can be shown to be a monotone decreasing function that Starts from ,reduces up to
Throughput (1) S=Normalized system throughput [fraction of time the channel is used to successfully transmit payload bits] Ptr=probability that there is at least one transmission in the considered slot time=p=1‐(1‐ τ)n Ps=probability that a transmission in the channel is successful =
Throughput (2) E[P]=average packet payload size PtrPs=probability of successful transmission in a slot time 1-Ptr=probability of the empty slot time Ptr (1-Ps)=probability of collision Ts =average time the channel is busy due to successful transmission Tc =average time the channel is busy during a collision σ=duration of an empty slot time S depends mainly on Ts and Tc
Basic Access Mechanism H=packet header=PHYhdr + MAChdr δ=propagation delay E[P* ]=Average length of the longest packet payload involved in a collision
RTS/CTS Access Mechanism H=packet header=PHYhdr + MAChdr δ=propagation delay
Model Validation & Simulation (1) Used event‐driven custom simulation program in C++ It closely follows all the 802.11 protocol details for each in dependent transmitting station The analytic model is extremely accurate The analytic results (lines) practically coincide with the simulation results (symbols) in both basic and RTS/CTS access
Model Validation & Simulation (2)
Model Validation & Simulation (3) Maximizing Saturation Throughput Max throughput achievable by Basic is very close to by RTS/CTS Throughput of RTS/CTS is less sensitive on τ RTS/CTS throughput has a much lower dependence on the system engineering parameters
Model Validation & Simulation (4) Throughput vs Number of Stations The greater the network size, the lower is the throughput [Except W=32] For Basic Access it varies with the values of n For RTS/CTS it is almost independent of n
Model Validation & Simulation (5) Throughput vs Initial Window Size For both Basic Access and RTS/CTS , a high value of W depends on the n
Model Validation & Simulation (6) Throughput vs Max. Back‐off Stage For both Basic Access and RTS/CTS , with W=32 and n=10 –50 Choice of m doesn’t practically affect the system throughput as long as is m is greater than 4 or 5
Model Validation & Simulation (7) Throughput vs Packet Length RTS/CTS mechanism is effective when packet size increases
Conclusion Simple but extremely accurate analytical model to study 802.11 DCF Covers both Basic Access and RTS/CTS mechanism as well as the hybrid one Provides good simulation results with comparison The best analytical model so far for DCF Finite number of terminals No hidden terminal Fixed Data Rate