On the Performance Behavior of IEEE
802.11 Distributed Coordination Function
M.K.Sidiropoulos, J.S.Vardakas and M.D.Logothetis
Wire Communications Laboratory,
Department of Electrical & Computer Engineering,
University of Patras,
265 00 Patras, Greece
E-mail: [email protected]
Outline
 Purpose of the paper.
 DCF-An Example.
 Mathematical Analysis ( Assumptions ).
 802.11 DCF: Markov Chain Model and Steady State Analysis
leading to a Saturation Throughput formula.
 Simulation results ( IEEE 802.11b network).
 Conclusion.
Purpose of the paper
 We propose a new Markov model for the DCF of IEEE 802.11
based on Bianchi’s, Wu’s and Ziouva’s models.
 and derive an analytical formula for the Saturation Throughput
for both Basic and RTS/CTS access schemes.
 Simulation Study:
• Validation of our new Markov model based on
throughput results by the NS-2.
• Average end-to-end packet delay for both access schemes.
DCF-An Example
DCF employs 2 mechanisms:
 Basic access scheme: A 2-way handshaking technique.
Note that:
 After a DIFS time interval each station defers for an additional random
backoff time.
 The backoff counter is frozen if a transmission is detected on the channel
BACKOFF SUSPENSION
DCF-An Example (cont.)
RTS/CTS:
 Request-To-Send / Clear-To- Send.
 It is a 4-way handshaking technique.
 Introduced to tackle the hidden terminal problem.
 Improve throughput performance in case of long packets.
Mathematical Analysis
Assumptions:
 Ideal channel conditions ( error-free channel).
 Finite number of stations, each of which has always a packet
available for transmission. (saturation conditions)
 Constant and independent collision probability p.
 Probability pb independent of the backoff procedure.
802.11 DCF: Markov Chain Model
(1-p)/W0
1-pb
0,0
0,1
1-pb
pb
0,2
1-pb
pb
0,W0-2
1-pb
pb
0,W0-1
pb
p/W1
(1-p)/W0
1-pb
i-1,0
1-pb
1-pb
i-1,1
pb
1-pb
i-1,2
pb
i-1,wi-1-2
pb
i-1,wi-1-1
W0-1
pb
p/Wi
(1-p)/W0
1-pb
i,0
1-pb
i,2
i,1
pb
1-pb
pb
1-pb
i,Wi-2
pb
i,Wi-1
pb
p/Wi+1
p/Wm
1/W0
m,0
1-pb
pb
m,1
1-pb
pb
m,2
1-pb
pb
m,Wm-2
1-pb
pb
m,Wm-1
Saturation Throughput model.
Normalized Throughput: ( fraction of the channel time used
for payload transmissions)
E[ payload transmitted in a slot time]
S
E[average length of a slot time]
Ps Ptr E[ P ]
S 
(1  Ptr )  Ptr PsTS  Ptr (1  Ps )Tc
 Ps :
a successful transmission in a slot.
 Ptr :
at least one transmission in a slot.




average packet payload.
duration of an empty slot time
average time of a successful transmission
average duration of a collision .
E[P]:
σ:
Ts:
Tc :
Steady State Analysis
Stationary Distribution of the chain ( Steady State ):
bi ,k  lim P{s (t )  i, b(t )  k}
t 
From the chain
we have:
bi ,0  p b0,0
i
, i  [0, m]
m 1
 Wi  k 

 1  p    b j ,0  bm,0  i  0

Wi (1  pb ) 
j 0

bi ,k  
 Wi  k  1  b
0

i

m
i
,0
 W 1 p
i
b

(1)
Steady State Analysis (cont.)
Normalization condition :
1
m
Wi 1

i 0 k 0
 Wi  2i W
Contention Window : 
m
W

2
W
 i
bi , k
(2)
i  m
i  m
(3)
(1), (2), (3)  b0,0  f ( p, pb , w, m, m )
'
 
m
b
i 0
i ,0
1  p m 1

b0,0 
1 p
Steady State Analysis (cont.)
Channel access probability :
  f ( p, pb )
p  1  (1   )
(4)


(5)

n 1
Probability of channel being busy : pb  1  (1   )


Collision probability :
n 1
(4), (5) 
 , p
numerical techniques
Ptr  1  (1   )
n
n (1   )n1
Ps 
1  (1   )n
Saturation Throughput model.
Normalized Throughput: ( fraction of the channel time used
for payload transmissions)
E[ payload transmitted in a slot time]
S
E[average length of a slot time]
Ps Ptr E[ P ]
S 
(1  Ptr )  Ptr PsTS  Ptr (1  Ps )Tc
 Ps :
a successful transmission in a slot.
 Ptr :
at least one transmission in a slot.




average packet payload.
duration of an empty slot time
average time of a successful transmission
average duration of a collision .
E[P]:
σ:
Ts:
Tc :
Simulation Study
 Performance metrics measured
by simulation:
• Saturation throughput.
• End-to-end average packet delay.
 Simulations in NS-2
 IEEE 802.11b single-hop
network.
 Network Topology:
• No hidden stations, all have LOS.
• CBR traffic over UDP links towards
the AP.
• No mobility.
Model Validation: Simulation vs. Analysis:
1Mbps.
Saturation Throughput
0.9
Basic and RTS/CTS
Basic 1Mbps
0.8
 Close match of analytical
model and simulation results.
0.7
0.6
0.5
Basic, simulation
Basic, new model
Basic, Wu's model
0.4
0
10
20
30
40
50
Number of Stations
 Our model is closer to
simulation than Wu’s.
Saturation Throughput
0.9
 The RTS/CTS gives higher
throughput than Basic due to
the short RTS frames.
( Only exception for n =5).
0.8
0.7
RTS/CTS, 1Mbps
0.6
RTS, simulation
RTS, new model
RTS, Wu's model
0.5
0
10
20
30
Number of Stations
40
50
Saturation Throughput
Model Validation: Simulation vs. Analysis:
5.5 and 11 Mbps.
 In both cases analysis and
simulation are in satisfactory
agreement.
Basic
0.6
RTS
0.4
5.5 Mbps
0.2
Basic, Simulation
Basic, New Model
RTS, Simulation
RTS, New Model
0.0
0
10
20
30
40
50
Number of Stations
Saturation Throughput
0.6
0.5
Basic
0.4
 Basic access scheme gives
higher throughput than
RTS/CTS when channel bit
rate ↑. ( RTS, CTS packets
are transmitted at 1Mbps).
RTS
0.3
11 Mbps
0.2
Basic, Simulation
Basic, New Model
RTS, Simulation
RTS, New Model
0.1
0.0
0
10
20
30
Number of Stations
40
50
 Throughput ↓ as bit rate↑
( DIFS,SIFS, Backoff delay
remain unchanged)
Average Delay Simulation
Saturation Delay (sec)
2.5
2.0
1.5
1 Mbps
Basic
1.0
RTS
5.5 Mbps
0.5
11 Mbps
0.0
0
10
20
30
40
50
Number of stations
 As network size ↑  delay ↑ for both access schemes.
 As channel bit rate ↑  delay ↓.
 RTS/CTS delay is lower than Basic delay only for 1Mbps.
Not efficient to use RTS/CTS for high data rates
Conclusion
 We have developed an analytical model to enhance Bianchi’s and Wu’s
analytical model for the saturation throughput of the DCF of the IEEE 802.11
protocol.
 Our model gives greater throughput results than Wu’s model for both access
schemes, Basic and RTS/CTS.
 Via numerous simulations with NS-2 we have shown that our model is close to
simulation, for all network sizes .
 As channel bit rate increases:
 throughput decreases
Average delay decreases.
 Basic vs. RTS/CTS:
In low rates RTS is better than Basic.
In higher rates Basic is preferable than RTS ( gives greater throughput
and lower delay).
