Pervasive Computing and Wireless Networking Research Group (PerWin)
A Review of Cross-Layer Scheduling
and Resource Allocation for Wireless
Mesh Networks
Jason Ernst and Mieso Denko
IEEE TIC-STH 2009 SESMET
September 26-27 2009
Department of Computing & Information Science
University of Guelph, ON, Canada
1
Introduction
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Wireless Mesh Networks
• OSI VS Cross-Layer Design
• Cross Layer Design Architectures
• Cross-Layer Design Techniques
• Our scheme
• Conclusions & Future Work
Department of Computing & Information Science
University of Guelph, ON, Canada
2
Wireless Mesh Networks
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Multi-hop wireless ad hoc network
• Mesh Routers MR
• Gateways GW
• Mesh Clients MC
• Majority of traffic between MC and GW
• Not MC to MC
• MR often assumed static, more resources
• CPU, Memory, Power (battery life)
Department of Computing & Information Science
University of Guelph, ON, Canada
3
Wireless Mesh Networks
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science
University of Guelph, ON, Canada
4
Wireless Mesh Networks
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Applications
• Commercial internet access
• Also applications in Military communication
• Search and Rescue
• Sensor applications where Mesh provides backbone
• Advantages
• Cheap and easy to deploy compared with wired
• Autonomous: self-configuration, self-optimization, selfhealing network
• Good for rural applications and sparsely populated areas
Department of Computing & Information Science
University of Guelph, ON, Canada
5
OSI VS Cross-Layer Design
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Cross-Layer
• Provides feedback from multiple layers
• More intelligent decisions made at routing, MAC layers
• Must be designed carefully to allow for extensions
• OSI
• Good design from software engineering point of view
• Provides good separation and abstraction compared with
a “flat” model
Department of Computing & Information Science
University of Guelph, ON, Canada
6
Cross-Layer Design
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Cross-Layer Design an Emerging Technology?
• It has been around for about 5 years now
• Last major survey covered only the beginning of crosslayer design
• Many developments in the last 5 years
• Somewhat controversial technique
•Many different ideas are being applied to cross-layering
• Cognitive radio techniques
• Adaptive control
• Network coding
Department of Computing & Information Science
University of Guelph, ON, Canada
7
Cross-Layer Design Architectures
Pervasive Computing and Wireless Networking Research Group (PerWin)
Application
Presentation
Session
Transport
Network
Link / MAC
Physical
OSI 7 Layer Stack
Department of Computing & Information Science
University of Guelph, ON, Canada
8
Cross-Layer Design Architectures
Pervasive Computing and Wireless Networking Research Group (PerWin)
Application
Presentation
Session
Direct Communication:
• Layers which do not normally interact exchange information
• Difficult to maintain
• Poor extensibility
Transport
Network
Link / MAC
Physical
OSI 7 Layer Stack
Department of Computing & Information Science
University of Guelph, ON, Canada
9
Cross-Layer Design Architectures
Pervasive Computing and Wireless Networking Research Group (PerWin)
Application
Status:
Presentation
• Link quality
• Queue sizes
• Application requirements
• Distance between nodes
Session
Transport
Status
Network
Link / MAC
Physical
• Easily enable cross-layer
interactions by querying the
status stack
OSI 7 Layer Stack
Department of Computing & Information Science
University of Guelph, ON, Canada 10
Cross-Layer Design Techniques
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Power Control
• Rate Control
• Route Control
• Network Coding
• Mixed-Bias
Department of Computing & Information Science
University of Guelph, ON, Canada 11
Power Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Power levels of competing nodes are adapted to ensure less
contention and interference
• Often combined with Rate Control, Route Control
• Makes use of Physical, MAC and Network layers
Department of Computing & Information Science
University of Guelph, ON, Canada 12
Power Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
No Interference between MCs
Department of Computing & Information Science
University of Guelph, ON, Canada 13
Power Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
Interference between MCs
Department of Computing & Information Science
University of Guelph, ON, Canada 14
Rate Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Allows MRs to control the transfer rates of associated MCs
• Rates are raised for a given link when the quality is higher
• Thresholds to ensure other MCs are not affected
• Solutions make use a wide range of layers
• Some take parameters from application layer
(multimedia applications)
• Generally Physical, MAC, Network layers are used
Department of Computing & Information Science
University of Guelph, ON, Canada 15
Rate Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
Obstruction between devices
Department of Computing & Information Science
University of Guelph, ON, Canada 16
Rate Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
No Obstruction between devices
Department of Computing & Information Science
University of Guelph, ON, Canada 17
Route Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Avoid congested links, links with poor quality
• Use SINR, queue sizes to determine which links to avoid
• Existing solutions make use of Network, Transport and Link
(MAC) layers
Department of Computing & Information Science
University of Guelph, ON, Canada 18
Route Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
Obstruction or Congestion on one link
Department of Computing & Information Science
University of Guelph, ON, Canada 19
Network Coding
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Allow multiple unicast transmissions simultaneously
• Assign a unique code for each link
• The correct information is decoded and separated from
other simultaneous transmissions
• SINR measure taken from physical layer to determine
which links may cause conflict
• Often combined with other previous techniques
• Usually uses the MAC / Physical layer
Department of Computing & Information Science
University of Guelph, ON, Canada 20
Summary of Techniques
Pervasive Computing and Wireless Networking Research Group (PerWin)
Reference
Technique
Layers
[12] J. Tang et. al
Power Control
MAC, Physical
[15] J. Thomas
Power Control
Network, MAC, Physical
[28] X. Wang et. al
Rate Control
Transport, MAC
[13] J. Tang et. al
Rate Control
Transport, Network, Link
[16] K. Karakayali et. al
Power / Rate
MAC, Physical
[1] C.E. Huang et. al
Power / Rate
Application, MAC, Physical
[7] H-Y. Wei
Route Control
Physical, Link, Network
[22] M.J. Neely et. al
Route Control
Transport, Network
[21] M.S. Kuran et. al
Route Control
Network, Link
[17] K. Li et. Al
Network Coding
MAC, Physical
Department of Computing & Information Science
University of Guelph, ON, Canada 21
Cross-Layer Mixed Biasing
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Mixed bias technique [Singh et al]
• Studied using different levels of bias
• A comparison against proportionally fair and max-min
algorithms
• Strong bias, weak bias
• Mixed bias combines a strong and a weak bias together
• Only bias against one characteristics
• Distance between GW and MR
Department of Computing & Information Science
University of Guelph, ON, Canada 22
Our Scheme
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Our Mixed-Bias technique
R

c
1
1
 2
c
(1)
R   1R1   2 R2  3 R3
(2)
• Additional characteristics
• Distance between GW and MR
• Queue Size
• Link Quality
• Combined Technique
• Biases against multiple characteristics at once
Department of Computing & Information Science
University of Guelph, ON, Canada 23
Our Scheme
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Scheduling / Resource assigned according to a cost function
at the gateways
• Multiple gateways are supported
• Each GW is responsible for scheduling / allocation for
MRs associated with it
• At each schedule a measure of the parameters is taken and
applied to the cost function
Department of Computing & Information Science
University of Guelph, ON, Canada 24
Initial Results
Pervasive Computing and Wireless Networking Research Group (PerWin)
Simulation Parameters
Parameter
Value
MRs
10 to 30
GWs
1 to 5
MCs
250
Flows
2 to 5
Environment Dimensions
1000 x 1000 m
Node Range
150 m
NS3 Simulation Tool
Department of Computing & Information Science
University of Guelph, ON, Canada 25
Initial Results
Pervasive Computing and Wireless Networking Research Group (PerWin)
0.1
Average End-To-End Delay
0.09
0.08
0.07
0.06
0.05
0.04
0.03
M-B Distance
0.02
Combined M-B
0.01
M-B Queue Size
802.11
0
10
15
20
25
30
Number of MRs
5 Flows – Effect of Varying MRs on End-to-End Delay
Department of Computing & Information Science
University of Guelph, ON, Canada 26
Initial Results
Pervasive Computing and Wireless Networking Research Group (PerWin)
0.9
M-B Distance
Packet Delivery Ratio
0.8
Combined M-B
0.7
M-B Queue Size
0.6
802.11
0.5
0.4
0.3
0.2
0.1
0
2
2.5
3
3.5
4
4.5
5
Number of Flows
5 GWs - Effect of Varying Flows on Packet Delivery Ratio
Department of Computing & Information Science
University of Guelph, ON, Canada 27
Conclusions and Future Work
Pervasive Computing and Wireless Networking Research Group (PerWin)
• Conclusions
• Cross-Layering should be viewed as intimidating
• Many existing approaches can apply cross-layer design
• The results show that our cross-layered mixed bias
approach is promising
• Future work
• Experiment with tuning the weightings and bias factors in
the mixed bias approaches
• Implement the scheme in real equipment to compare
• Many existing schemes make assumptions that limit the
application (single GW, no mobility of MCs or MRs)
Department of Computing & Information Science
University of Guelph, ON, Canada 28
Questions?
Pervasive Computing and Wireless Networking Research Group (PerWin)
Contact: Jason Ernst – [email protected]
Thank you for listening!
Department of Computing & Information Science
University of Guelph, ON, Canada 29