Interaction Between Hidden Node Collisions and Congestions in Multihop AdHoc Networks

1
Interaction Between Hidden Node Collisions and
Congestions in Multihop Ad-Hoc Networks

• Hicham KHALIFE Naceur MALOUCH
• Université Pierre et Marie Curie - LIP6
• FRANCE
• IEEE ICC06

2
IEEE 802.11 (reminder)
Virtual Carrier Sensing
RTS
1
DATA
S
R
CTS
ACK
The IEEE 802.11 DCF uses the 4 way
handshake RTS/CTS/DATA/ACK
3
Motivation and Objectives

• Two types of losses in wireless Ad Hoc Networks
  Congestion Collision losses
• Analyze and investigate the effect of each loss
  type and their interaction
• Approach
• 1- Understand
• 2- Model
• 3- Quantify
• 4- React

Cs transmission range
Contribution
E
D
B
C
Congestion
Collision at Node B
Ds interference range
4
Packet Loss in IEEE 802.11
RTS X 7 Loss
1
DATA X 4 Loss
S
R
CTS
ACK

• Packet discarded at the source Mac layer
• If
• RTS retries exceed the Long Retry Threshold
  (7 default value)
• Or
• DATA retries exceed the Short Retry Threshold
  (4 default value)
• Separated by Backoffs

5
Packet Loss in a Chain Topology
Simplified Model If node D transmission will
take more than the time needed for the 7 RTS
retries at A, A will discard the packet
Poisson l
Poisson
C
D
E
B
6
Packet Loss Probability
Poisson l

• Pr(Loss) Pr(Loss qD k) Pr(qD k)
• Pr(Loss qD lt k) Pr(qD lt k)

k
D
A
E
B
C
1
Pr(Loss qD 0)
Pr(Loss 0 lt qD lt k)
qD Number of packets in D queue
C Link Capacity r Long retry threshold
7
Packet Loss Probability
Poisson l
k
i
D
A
E
B
C

• If Node Ds queue is empty an arriving at A
  packet will be sent immediately

Pr(Loss qD 0) Pr(qD 0) x Pr(receiving at
least k packets during A transmission time
T(constant) )

• If D is transmitting, A packets sent will
  collide , lost packet at A if k-i packets are
  received during T.i

Pr(Loss 0 lt qD lt k) ? Pr(qD i) x
Pr(receiving k i packets during the
transmission of the existing i)
8
Packet Loss Probability
Poisson l
k
i
D
A
E
B
C

• The packet Loss Probability in a five nodes chain
  topology

• Related work
• Collision probability Ray et al. 2005
• WLAN Bianchi 2000, Kumar et al. 2005,
  ElRakabawy et al 2005

9
Validation Through NS Simulation
(packets/s)
10
Model Extension
Poisson l
B
C
D
E
l
Only 1 collision case A-D simultaneously
transmitting Obtained packet loss
probability pf(l)
11
Model Extension
l
l
k
D
A
E
B
C

• p f(l)?
• p f1(l)
• l f2(p, l?
• A first approximation
• p f1(l)
• l (1 p)l???Upper bound)
• Can be resolved with the fixed point theorem

12
Model Extension
p f(l)
p lt 1 l/l
(packets/s)
13
Discussion

• Loss probability formula
• Relation between loss probability, input rate
  and MAC output queue size
• Taking advantage of the k value
• Idea
• Eliminating losses due to hidden node collisions.
  
• How
• Setting queue sizes to the value smaller than k

14
Interaction Between Collision and Congestion
Collision
Congestion
(packets/s)
Node
15
Conclusion and Ongoing Work

• Loss probability formula in the five nodes chain
  topology
• Solution to improve TCP throughput

Reducing buffer size increases performance
B
C
D
E

• Future work
• Extending the formula to limited queue sizes
• General formula to generalized topologies

16
Thank You!Merci !