Linklevel Measurements from an 802.11b Mesh Network

1
Link-level Measurements from an 802.11b Mesh
Network
2005 Autumn CS712

• Daniel Aguayo, John Bicket, Sanjit Biswas, Robert
  Morris
• M.I.T Computer Science and Artificial
  Intelligence Laboratory
• Glenn Judd
• Carnegie Mellon Univeristy
• SIGCOMM04

2005. 9. 29 Presented by Sangho
Lee shlee_at_cosmos.kaist.ac.kr
2
Table of Contents

• 1. Introduction
• 2. Experimental Methodology
• 3. Distribution of Delivery Probabilities
• 4. Spatial Distribution of Loss Rates
• 5. Time Variation of Loss Rate
• 6. Effect of Signal-to-Noise Ratio
• 7. Effect of Transmit Bit-Rate
• 8. Interference from 802.11 Sources
• 9. Effect of Multi-path
• 10. Conclusions

3
1. Introduction (1/2)
1.1 Roofnet map
1.2 Omni-directional antennas
Black Points PCs with an 802.11b card connected
and omni-directional antenna
4
1. Introduction (2/2)

• 1.3 Problems
• Internet Service via a few wired gateway
• Implementation strategy
• neighbor abstraction
• partitions all the pairs of nodes into pairsthat
  can communicate directly, and pairs that
  cannot.(wireless versus. Wired)
• Existing radio, MAC, routing technology
• Bad performance
• MAC, routing protocols poor fit to networks
  actual behavior
• 1.4 Goal
• provide insight into which differences are
  important
• draw conclusion relevant to the design of future
  MAC and routing protocols

Measurement study of the Roofnet wireless
network!!
5
2. Experimental Methodology

• 802.11b card (Intersil Prism 2.5 chip-set)
• pseudo-IBSS (ad hoc) mode
• Each node in turn sends 1500-byte 802.11
  broadcast packets, the rest of the nodes
  passively listen.
• No link-level ACKs, retransmissions
• 90 sec. at each of the 802.11b bit-rates (1,2,
  5.5, 11Mbit/s)

Link 37
Receiver 37
Sender
Link 1
Link ..
Receiver 1
Receiver ...
Link 3
Link2
Link ..
Receiver 3
Receiver 2
Receiver ..
802.11 Broadcast Packets
Simplified version of the 802.11b IBSS (ad hoc)
mode IBSS Independent Basic Service Set
6
3. Distribution of Delivery Probabilities

• most Roofnet node pairs that can communicate have
  intermediate loss rates.

7
4. Spatial Distribution of Loss Rates
4.1 distance versus delivery probability at 1Mbps

• No discernible relationship between distance and
  delivery probability

8
5. Time Variation of Loss Rate
Highest variation
Bit rate 1Mbps, Total Average Loss rate 50
Delivery probability
lowest variation
Delivery probability
Time (sec.)
The lines indicate averages over successive 200
ms interval

• Different Route selection and Error correction
  strategy are required
• The relatively smooth bottom graphs are the most
  common types
• The links are not really alternating between up
  and down

9
6. Effect of Signal-to-Noise Ratio (1/2)

• Many links have marginal signal-to-noise ratio?
• Prism 2.5 specification1 suggest the range of
  S/N values for which the packet error rate would
  be between 10 and 90 is only 3dB wide.

10
6. Effect of Signal-to-Noise Ratio (2/2)
Roofnet (2Mbps)
emulator and two Prism 802.11b cards

• S/N does not predict delivery probability for
  intermediate-quality links

11
7. Effect of Transmit Bit-Rate (1/2)

• 7.1 802.11b transmit bit-rates differ in
  robustness

The pairs sorted by the throughput at 11Mbit/s
12
7. Effect of Transmit Bit-Rate (2/2)

• 7.2 802.11b bit-rate selection algorithms
• A. should wait until a high bit-rate is
  performing very badly before it reduces the
  bit-rate
• B. 11Mbit/s often provides higher throughput
  than 5.5Mbit/s even when the loss rate at
  11Mbit/s is higher than 50
• C. performance at a low bit-rate is not a
  good predictor of performance at higher rates
• Bit-rate selection must be based on explicit
  measurements of throughput at the different rates

13
8. Interference from 802.11 Sources (1/2)

• 8.1 Interference from other 802.11activity?

Used channel for experiments
All packets are from non-Roofnet sources before
experiments
Foreign Packets
14
8. Interference from 802.11 Sources (2/2)
X-axis of foreign packets received per second
by the receiver in the pair Y-axis of 1500
byte packets lost per second at 1Mbit/s

• No correlation between foreign traffic observed
  and packets lost

15
9. Effect of Multi-path (1/4)

• 9.1 Reflection
• a delayed and attenuated copy of the signal

original
B
B
reflection
A
building

• Prism 2.5 (RAKE receiver) suppress reflected
  copies with delays of up to 250 nsec.1
• Outdoor urban radio propagation findthat delay
  spreads often exceed one µsec 13,5
• Theoretical models demonstrate that such delay
  spreads significantly increase packet loss
  rates4

16
9.2 channel emulator to investigate
multi-path effects
9. Effect of Multi-path (2/4)
delay
attenuation

• Two channel used for emulation experiments
• 200 broadcast packets at 1,2,5.5,11 Mbit/s
• Reflected ray Increments of 0.02 µsec., 0.2dB

17
9.3 emulation experiments
9. Effect of Multi-path (3/3)

• Black bar more than 90 loss
• Gray bar between 10 and 90 loss
• Multiple of the modulations symbol boundary
• Delayed path delivers valid symbols that the
  receiver cannot distinguish from the direct
  symbols
• The process of imposing sound or visual
  information onto the radio waves created in a
  transmitter.

Delivery probability
Delay of second ray (nsec.)
18
9.2 distance between all pairs of Roofnet nodes
9. Effect of Multi-path (4/4 )

• Reflected delay one µsec. gt 300meters
• Median distance between Roofnet Nodes 500 meters

19
10. Conclusions

• No clear distinction between working and
  non-working links
• Link distance, S/N ration do have an effect on
  loss rate, but the correlation is weak
• Important cause of intermediate loss rates is
  multi-path fading due to reflections

20
references

• 1 ISL3873 Wireless LAN Integrated Medium
  Access Controller with Baseband Processor.
  Intersil Corporation, 2000. Application Note
  FN4868.
• 4 M. V. Clark, K. K. Leung, B. McNair, and Z.
  Kostic. Outdoor IEEE 802.11 cellular networks
  Radio link performance. In Proc. of IEEE ICC
  2002, April 2002.
• 5 D. C. Cox. Delay Doppler characteristics of
  multipath propagation at 910 MHz in a suburban
  mobile radio environment. In IEEE Transactions on
  Antennas and Propagation, AP-20(5)625-635,
  September 1972.
• 13 E. S. Sousa, V. M. Jovanovic, and C.
  Daigneault. Delay spread measurements for the
  digital cellular channel in Toronto. In IEEE
  Trans. on Veh. Tech., vol. 43, no. 4, pp. 1-11,
  November 1994.
• Roofnet MIT Roofnet http//pdos.lcs.mit.edu/roof
  net