Distributed Channel Management in Uncoordinated Wireless Environments Arunesh Mishra, Vivek Shrivastava, Dheeraj Agarwal, Suman Banerjee, Samrat Ganguly.

Distributed Channel Management in Uncoordinated Wireless EnvironmentsArunesh Mishra, Vivek Shrivastava, Dheeraj Agarwal, Suman Banerjee, Samrat GangulyUniversity of Wisconsin & NEC Labs

Presented by: Anuradha KadamFebruary 27, 2007

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Introduction

Wireless 802.11 hotspots: uncoordinated Unsatisfactory and unpredictable network

performance Primary focus: fairness problem Channel assignment: channel-hopping

Key Components

Channel Hopping Switching Overhead Impact on TCP Partially Overlapped channels Client-driven Assignment

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Background Channel Assignment Techniques

Non-overlapping channels Static approach – unfairness Least Congested Channel Search (LCCS) -

distributed CFAssign using Randomized Compaction (RaC) -

centralized

www.cs.wisc.edu/~arunesh/chop06.ppt

Background

Using Partially Overlapped channels “Partially Overlapped Channels not considered

harmful” As physical separation increased, amount of

interference decreased and this led to increase in throughput

At lower separation levels, throughput can be increased by increasing channel separation.

Increase spatial re-use by careful selection

Channel Hopping

• Channel Hopping Sequence• Periodicity• Throughputs of interfering APs get averaged out to equal values

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

MAXchop Algorithm

MAXchop Algorithm Initialize

Bootup or periodically (a week) Initialize channel assignment with pseudo-random hopping

sequence Hop

End of hopping period (Nsts) Computes new hopping sequence Based on information about hopping sequences of interfering

APs. Compute MinMax

Returns a color from C such that it distributes the interference equally among all neighbors of x.

For simplicity, assume color is chosen randomly

MAXchop Algorithm Partially Overlapped channels

ρ(u,i,x,j) if AP u on channel i interferes with AP x on channel j Return binary value or an accurate estimate of interference Px(u) I(i,j) received power if tx and rx on channels i and j Received power should be above a certain threshold to

cause interference – binary value ρ(u,i,x,j) = Px(u) I(i,j) – accurate estimation

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Practical Considerations

Implementing Channel SwitchingClient-AP coordination

Beacon message

Channel Switch Overhead 20 ms for Prism 2.5, 6 ms for Atheros Triggered during low periods of activity Slot duration large Gains v/s overhead

Practical Considerations

Interfering APs estimationClient drivenAP driven

Asynchrony in hoppingdifferent hopping periodsasynchronous time slotsover long periods performance is same

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Simulations

Packet-level simulations Hotspot topologies derived from Wigle Compare against LCCS and RaC AP locations for dense urban area Partitioned into 12 non-interfering

topologies

Simulations

Simulation Methodology

NS-2 simulator Slot durations loosely synchronized Switch latency of 20ms Two metrics:

Aggregate network throughput Fairness in per-AP throughput

Jain’s fairness index

5 clients on average

Simulation-Results (1)

Sample Topology 27 APs with uneven

density 8 suffer considerable

interference Remaining had

similar throughputs

Simulation-Results (1)

Simulation-Results (2)

Simulation-Results (2)

12 urban topologies Evaluate only partially-overlapping

channels. Channel hopping improves fairness over

LCCS by an average of 42%. Ch. Hopping gives performance

improvement of 30%.

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Implementation

Five APs One client/AP Typical hotspot area Different methods of

channel assignment NOV-LCCS, NOV-

MAXchop, POV-MAXchop, POV-static

TCP/UDP throughputs

Results - TCP

Throughput gains: 15.13% by POV-MAXchop over NOV-chop & 15.05% by POV-static over NOV-LCCS

Results - UDP

Throughput gains: POV-MAXchop improves by 10%

Improvement in fairness

Outline

Introduction Background MAXchop Algorithm Practical Considerations Simulations Implementation Conclusion

Conclusion

Channel hopping:simple and efficient methodGood fairness propertiesUtilize partially overlapped channels

Provide throughput gains in dense networks.

Questions??