Upload
shawn-ross-fowler
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
A Review of Cross-Layer Scheduling and Resource Allocation for Wireless
Mesh Networks
Jason Ernst and Mieso Denko
IEEE TIC-STH 2009 SESMETSeptember 26-27 2009
Department of Computing & Information Science University of Guelph, ON, Canada 1
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 2
Introduction
• Wireless Mesh Networks
• OSI VS Cross-Layer Design
• Cross Layer Design Architectures
• Cross-Layer Design Techniques
• Our scheme
• Conclusions & Future Work
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 3
Wireless Mesh Networks
• 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)
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)
Department of Computing & Information Science University of Guelph, ON, Canada 5
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, self-healing network• Good for rural applications and sparsely populated areas
Department of Computing & Information Science University of Guelph, ON, Canada 6
OSI VS Cross-Layer Design
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 7
Cross-Layer Design
• Cross-Layer Design an Emerging Technology?• It has been around for about 5 years now
• Last major survey covered only the beginning of cross-layer 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 8
Cross-Layer Design Architectures
Application
Presentation
Session
Transport
Network
Link / MAC
Physical
OSI 7 Layer Stack
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 9
Cross-Layer Design Architectures
Application
Presentation
Session
Transport
Network
Link / MAC
Physical
OSI 7 Layer Stack
Direct Communication:• Layers which do not normally interact exchange information• Difficult to maintain• Poor extensibility
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 10
Cross-Layer Design Architectures
Application
Presentation
Session
Transport
Network
Link / MAC
Physical
OSI 7 Layer Stack
Status
Status:• Link quality• Queue sizes• Application requirements• Distance between nodes
• Easily enable cross-layer interactions by querying thestatus stack
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 11
Cross-Layer Design Techniques
• Power Control
• Rate Control
• Route Control
• Network Coding
• Mixed-Bias
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 12
Power Control
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 13
Power Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
No Interference between MCs
Department of Computing & Information Science University of Guelph, ON, Canada 14
Power Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
Interference between MCs
Department of Computing & Information Science University of Guelph, ON, Canada 15
Rate Control
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 16
Rate Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
Obstruction between devices
Department of Computing & Information Science University of Guelph, ON, Canada 17
Rate Control
Pervasive Computing and Wireless Networking Research Group (PerWin)
No Obstruction between devices
Department of Computing & Information Science University of Guelph, ON, Canada 18
Route Control
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 19
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 20
Network Coding
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 21
Summary of Techniques
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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 22
Cross-Layer Mixed Biasing
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 23
Our Scheme
• Our Mixed-Bias technique
• Additional characteristics• Distance between GW and MR• Queue Size• Link Quality
• Combined Technique• Biases against multiple characteristics at once
21
1
ccR
332211 RRRR (1) (2)
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 24
Our Scheme
• 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
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 25
Initial Results
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
Simulation Parameters
NS3 Simulation Tool
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 26
Initial Results
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
10 15 20 25 30Number of MRs
Avera
ge E
nd
-To
-En
d D
ela
y
M-B Distance
Combined M-B
M-B Queue Size
802.11
5 Flows – Effect of Varying MRs on End-to-End Delay
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 27
Initial Results
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
2 2.5 3 3.5 4 4.5 5
Number of Flows
Pa
ck
et
De
live
ry R
ati
o
M-B Distance
Combined M-B
M-B Queue Size
802.11
5 GWs - Effect of Varying Flows on Packet Delivery Ratio
Pervasive Computing and Wireless Networking Research Group (PerWin)
Department of Computing & Information Science University of Guelph, ON, Canada 28
Conclusions and Future Work
• 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)
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
Questions?
Pervasive Computing and Wireless Networking Research Group (PerWin)