Practical Localized Network coding in Wireless Mesh Networks

1
Practical Localized Network coding in Wireless
Mesh Networks

• Soji Omiwade, Rong Zheng, Cunqing Hua
• Department of Computer Science
• University of Houston
• June 13, 2008

2
Outline

• Motivation
• COPE network coding
• BFLY network coding
• Forwarding of coded packets
• Localized route optimizations
• Implementation and simulation results
• Coding decision analysis
• Conclusion

3
Motivation

• Wireless mesh networks
• Last mile network (access network)?
• Features
• Immobile nodes
• Available power supply
• High throughput requirement

Image courtesy of en.wikipedia.org/wiki/wireless_m
esh_network
4
Network Coding--Brief

• In certain cases, necessary to further increase
  throughput Ahlswede 00
• Example of such Butterfly
• System performance increase
• Throughput increase Katti 06
• Error checking Cai 02
• Security Jain 04
• Practical network coding
• COPE
• BFLY

5
COPE

• First practical application of coding in wi-mesh
  nets
• XOR packets together
• Use of certain node configurations to apply
  network coding

6
A
B
X-topology Flows 1?4, 2?3 Classical routing
7
A
A
B
B


A
B
Coding Gain
COPE coding
8
Other COPE structures
9
BFLY Contribution

• Drawbacks of COPE coding
• Coded packets are decoded in one hop no
  forwarding
• Coding oblivious routing
• Coding oblivious scheduling
• Our Contribution
• Allow forwarding of coded packet
• Make use of BFLY local structure
• Localized route optimizations
• Coding gain analysis tools

10
A
B
Butterfly topology Flows 1?4, 2?3 Classical
routing
11
A
B
COPE coding
12


COPE coding
13
A
B
COPE coding
Coding Gain
14
A
B
BFLY coding
15


BFLY coding
Coding Gain
16
BFLY Scheme

• Where to code?
• BFLY structure discovery
• Localized route optimization to increase coding
  opportunities
• When to code?
• Threshold based coding decision supported by
  analytical models
• How to decode?
• Packet pool for overheard packets
• Forwarding of coded packets
• XOR of coded packets with original packets

17
Butterfly Discovery (simu..sneak peek)

• Periodically determine butterflies
• Use of CTRL msgs (BFLY-Hellos)?
• Such msgs contain neighbour info
• Will need to contain 2-hop neighbour info
• Complexity
• Heavily congested networks
• At most 2 mins to determine all butterflies
• Hello frequency 1 sec

18
Difficulties

• Routing may not go through BFLY structures
• Making the coding gain unpredictable
• Solution
• Construct a coding aware routing scheme Sengupta
  07
• Would route route flows through butterflies
• Required centralized computation!
• Solution II
• Localized route optimizations

19
A
B
Non BFLY routes
20
A
B
Non BFLY routes
21
Beacon messages Route Optimizations
22
A
B
Beacon messages Route Optimizations
23
Simulation Setup

• Simulation environment ns2.30
• Propagation model Two ray ground
• 802.11g, rts/cts disabled, mac data rate 1Mbps
• Transmission range 250m Int. range 550m
• Transport Protocol
• Mainly UDP
• MAC Protocol 802.11
• Transmission range 250m
• Interference range 550m
• Performance metrics
• End to end delay
• Coding gain Throughput Ratio

24
BFLY-Topo C.G

• 1?4 2?3
• UDP .2 Mbps
• MAC 1 Mbps
• 50 runs
• n2n dist 200m

Coding Gain
25
Bfly-hybrid I
Bfly-hybrid
A
B


B
A
Coding Gain
25
26
Bfly-hybridII
A
B


Coding Gain
Coded-Packet Route Variety
27
6x6 Grid topology node-2-node dist 200m
Random number of flows Katti 06
flow_seed
Local route optimizations beneficial
28
6x6 Grid-random topology (11 random nodes)
Random flows 2,

• 1?4 2?3
• UDP .2 Mbps
• MAC 1 Mbps
• 50 runs

Coding Gain
29
Combating lossy links via modeling

• Lossy links make network coding detrimental
• Solution Analytically compute the gain
  achievable
• Model packet sending in a wireless network
• Obtain maximum coded and non-coded throughput
• For any link loss ratios given
• Key Assumption Fair sending amongst nodes

30
Lossy Network

• The coding gain is not a linear function
• 0-0 hardly affects the coding gain

31
Conclusion

• Designed a practical coding scheme---BFLY
• It complements COPE
• By allowing forwarding of coded packets
• Coding gains are topology dependent
• Implemented localized route optimization schemes
• Developed coding gain analysis tools

32
Future Work

• Evaluate our coding gain analysis tools with
• LP tools
• Simulations
• Coding with any scheme is beneficial only when
  link losses are properly taken into account??
• Develop distributed coding aware routing
  algorithms

33
References

• 1 Ahlswede, N. Cai, N. Li, S.-Y. R. Yeung, 0.
  W. Network Information Flow
• 2 S. Katti, H. Rahul, W. Hu, D. Katabi, M.
  Medard, and J. Crowcroft, XORs in The Air
  Practical Wireless Network Coding, In ACM
  SIGCOMM 2006, Pisa, Italy, September 11 -15
• 3 Ning Cai and Raymond W. Yeung, Network
  Coding and Error Correction, ITW2002 Bangalore
• 4 K. Jain, Security based on network topology
  against the wiretapping attack, IEEE Wireless
  Communications, pp. 68-71, Feb 2004.
• 5 S. Sengupta, S. Rayanchu and S. Banerjee. An
  Analysis of Wireless Network Coding for Unicast
  Sessions The Case for Coding-Aware Routing.
  Proceedings of IEEE INFOCOM07. 2007.

34
(No Transcript)
35
2000x2000, 1000x1000, 500x500
36
Packet Headers
37
(No Transcript)