A General Model of Wireless Interference

1
A General Model of Wireless Interference

• Lili Qiu, Yin Zhang, Feng Wang, Mi Kyung Han
• University of Texas at Austin
• Ratul Mahajan
• Microsoft Research
• ACM MOBICOM 2007
• Sept. 12, 2007

2
Motivation

• Interference is critical to wireless network
  performance
• Its impact is not well understood
• Lack of a general model for multihop wireless
  networks
• Make wireless network optimization/control hard
• Often have to resort to indirect metrics to
  optimize

1 Mbps
1 Mbps
1 Mbps
Throughput 2 Mbps
A
B
D
C
1 Mbps
1 Mbps
1 Mbps
Throughput 1 Mbps
A
B
D
C
3
State of the Art

• Abstract models of radio propagation
• e.g., Interference range is twice the
  communication range
• Inaccurate for real networks
• Direct measure wireless interference
• Lack scalability and predictive power
• Interference models
• Single-hop networks (e.g., Bianchis model)
• Multihop networks only handle restricted traffic
  (e.g., Reis et al. and Garetto et al.)
• Only two senders or two flows
• Only broadcast traffic
• Only saturated demands

4
Our Contributions

• First general interference model for IEEE 802.11
  multihop wireless networks that supports
• Interference among an arbitrary number of senders
  
• Both broadcast and unicast traffic
• Both saturated and unsaturated demand

5
Background on IEEE 802.11

• 802.11DCF uses CSMA/CA
• Nodes stay silent when the channel is sensed busy
• Virtual carrier sense using Network Allocation
  Vector (NAV)
• NAV is updated based on overheard packets
• Physical carrier sense by measuring the total
  energy received at the sender
• Broadcast
• If medium is idle for DIFS, transmit immediately
• Otherwise, wait for DIFS and then a random
  backoff between 0, CWmin

DIFS
Data Transmission
Random Backoff
6
Background on IEEE 802.11 (Cont.)

• Unicast
• Use ACKs and retransmissions for reliability
• Binary backoff
• CW doubles after each failed transmission until
  CWmax
• Restore CW to CWmin after a successful
  transmission

DIFS
Data Transmission
Random Backoff
ACKTransmission
SIFS
7
Assumptions of Our Model

• A sender can transmit if
• The total energy it received ? CCA threshold
• A receiver can correctly receive a transmission
  if its
• RSS ? radio sensitivity
• SINR ? SINR threshold
• Our model can easily be extended to BER-model

8
Overview of Our Model
sender model
given network
RF profile measurement
pairwise RSS
throughput
traffic demand
receiver model
goodput

• How it works
• Measure pairwise RSS via broadcast probes
• One node broadcast at a time, others measure RSS
  ? O(n) probes
• Saturated broadcast sender/receiver models
• Markov-chain model
• Extend to unsaturated broadcast
• Extend to saturated/unsaturated unicast

9
Broadcast Sender Overview

• Estimate how much a sender can send
• Key estimate joint active probability
• Markov chain
• State i a set of active nodes Si
• Throughput of node m tm ?im?Si ?i

00
01
01
00
0..10
.
.
10
11
.
11
.
10
Broadcast Sender Overview (Cont.)

• State transition probability
• Staying idle P00(nSi)
• Idle to active P01(nSi)
• Active to idle P10(nSi)
• Staying active P11(nSi)
• Assume node independence
• Compute stationary probabilities ?i by solving LP
• Highly efficient for sparse M

11
Broadcast Sender Transition Probabilities
Under the assumption that both transmission and
idle times are exponential
12
Broadcast Sender Clear Probability

• How to estimate ImSi?
• ImSiWmBm?s?Si\m Rsm
• Assume each term is lognormal variable
• Validated by our testbed measurements
• Approximate the sum using a lognormal variable by
  matching mean and variance

13
Broadcast Sender Clear Probability
14
Broadcast Sender Handle Similar Packet Sizes

• Synchronization occurs when packet sizes used by
  different nodes are similar
• When several nearby nodes transmit together, they
  will end transmission together
• Independence assumption fails
• Handle synchronization
• Construct synchronization graph Gsyn
• Two nodes are connected iff C(mn) ? 0.1 and
  C(nm) ? 0.1
• Find all synchronization groups
• Each connected component in Gsyn is a
  synchronization group
• If m and n in the same synchronization group
• m?Sj and n ?Sj? M(i,j) 0
• P10(mnSi) Tslot/T?(m) instead of
  (Tslot/T?(m))G

15
Broadcast Sender Handle Unsaturated Demands

• Estimate Q(m) probability m has data to send
  when its backoff counter is 0 and channel is
  clear at m
• Under saturated demands, Q(m) 1
• Under unsaturated demands, compute Q(m)
  iteratively to ensure that demands are not
  exceeded

Initialize Q(m) 1
Solve the Markov chain
Update Q(m)
16
Handling Unsatuated Demands
17
Broadcast Sender Enhance Scalability

• Basic model requires 2N states and 2Nx2N
  transition probabilities
• Prune states including too many synchronized
  transmissions
• Transmissions involving ? 3 nearby nodes
• Transmissions involving ? 2 groups each with 2
  nearby nodes
• Remove transitions with low probabilities ? make
  M(i,j) sparse
• Reset M(i,j) to 0 if it is below 0.001

18
Broadcast Receiver

• Goodput
• Key Estimate
• Challenge slot-level loss rate ! pkt loss rate
• Partial pkt corruption e.g., 10 slot loss rate
  could lead to 100 pkt loss if lossy slots are
  spread over all pkts)
• Estimate slot-level loss rate
• Estimate packet loss rate

19
Broadcast Receiver (Cont.)

• Estimate slot-level loss probabilities
• Loss due to low RSS
• Loss due to low SNR
• Estimate packet loss probabilities
• Loss due to low RSS
• Loss due to low SNR ( and ) refer
  to paper

20
Broadcast Receiver (slot-level loss due to low
SNR)
21
Broadcast Receiver(Packet-level loss due to low
SNR)
22
Broadcast Receiver(Packet-level loss due to low
SNR)
23
Unicast Model Overview

• Challenges
• Binary backoff
• Sending rate depends on loss rates
• DATA losses due to collisions with ACKs
• Model ACK sending rate, which in turn depends on
  DATA sending rate and loss rates
• ACK losses
• asymmetric RSS induced losses
• collisions with DATA
• Our solution
• Use an iterative process to capture the
  inter-dependency between DATA and ACKs

24
Unicast Model Iterative Framework
Initialize Lmn 0, Q(m) 1
Compute CW(m), OH(m) using Lmn
Derive transition matrix M
Compute ?i using M
Compute Qnew(m) based on ?i and previous Q(m)
Compute using ?i
25
Unicast Sender Extensions to Broadcast Sender
Model

• Model binary backoff
• Compute average CW under packet loss rate L
• When there are multiple receivers, CW(m) is
  weighted average over all receivers
• Compute ACK overhead
• Include SIFS/ACK overhead when DATA is received
  successfully
• OH(m) is the weighted average of OH(m,n) over all
  receivers n
• Compute Q(m)
• The only change is to account for retransmissions

26
Unicast Sender
27
Unicast Sender
28
Unicast Receiver Extensions to Broadcast
Receiver Model

• Challenges
• Asymmetric link quality
• Collisions due to ACKs
• Approaches
• Extend to include RSS induced losses for
  both data and ACK
• Extend to include SINR induced slot-level
  loss due to collision between ACK/data, data/ACK,
  ACK/ACK (in addition to data/data)
• Refer to the paper for details

29
Evaluation Methodology

• Qualnet simulation
• Controlled environment and direct assessment of
  individual components in our model
• Vary topologies, senders, demand types, freq.
  band
• Testbed experiments
• More realistic scenarios
• RF fluctuation, measurement errors, and variation
  across hardware
• UW traces (Reis et al.)
• 14-node testbed inside an office building
• 2-sender traces
• UT traces
• 22-node, 802.11 a/b/g NetGear WAG511, Madwifi,
  click
• Vary senders, demand types
• Metric
• root mean square error (RMSE)

30
Simulation Evaluation Saturated Broadcast
2 saturated broadcast
(a) throughput
(b) goodput
More accurate than UW 2-node model
31
Simulation Evaluation Saturated Broadcast
10 saturated broadcast
(a) throughput
(b) goodput
Accurate for 10 saturated broadcast
32
Simulation EvaluationUnsaturated Unicast
10 unsaturated unicast
(a) throughput
(b) goodput
Accurate for unsaturated unicast
33
Testbed Evaluation
UW traces 2 senders, 30 mW, broadcast, saturated
(b) goodput
(a) throughput
More accurate than UW-model for 2-sender
34
Testbed Evaluation (Cont.)
UT traces 5 senders (30 mW), broadcast, saturated
(a) throughput
(b) goodput
Accurate for saturated broadcast
35
Testbed Evaluation (Cont.)
UT traces 3 senders (1mW), broadcast, unsaturated
(a) throughput
(b) goodput
Accurate for unsaturated broadcast
36
Testbed Evaluation
UT traces 2 senders (30 mW), broadcast, saturated
(a) throughput
(b) goodput
37
Summary

• Main contributions
• A general interference model that handles
• An arbitrary number of senders
• Broadcast unicast traffic
• Heterogeneous traffic
• Validated by simulation and testbed evaluation
• More general than the existing models
• More accurate than the previous 2-sender model
• Future work
• Model e2e performance in multihop networks
• Apply the model to wireless network optimization