1. Efficient Routing through Late Binding in Wireless Meshes
September 2005
doc.: IEEE /0909r0
September 2005
Efficient Routing through Late Binding in Wireless Meshes
Date:
Authors:
Notice: This document has been prepared to assist IEEE It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.
Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE
Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures < ieee802.org/guides/bylaws/sb-bylaws.pdf>, including the statement "IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed within the IEEE Working Group. If you have questions, contact the IEEE Patent Committee Administrator at
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

2. September 2005
doc.: IEEE /0909r0
September 2005
Abstract
Traditional routing in wired and wireless networks, including mesh networks, typically choose a deterministic path on which packets are routed. Path changes occur over a relatively longer time scale with detection of changes in network path properties. However, many characteristics of the wireless environment change at much smaller timescales than what routing protocols can react to. In this work, we contend that traditional routing approaches, when applied to wireless mesh networks miss significant opportunities of the wireless broadcast or multicast advantage in route selection.
In our proposed work, we define a new mechanism for routing in wireless mesh networks, where the exact path choice is instantiated only at packet transmission time, taking into account exact channel characteristics. Our analysis shows that such an approach can lead to significant performance improvements in many multi-hop wireless networking scenarios, and particularly in wireless mesh networks.
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

3. Overview and Assumptions
September 2005
Overview and Assumptions
A multi-hop and multi-radio wireless mesh network
Mesh Nodes and Stations
Problems explicitly addressed in this talk:
Routing strategy and its implications
Routing metric and its implications
Problems not explicitly addressed (but have interesting solutions for):
Channel assignment
Transmission power control
Load balancing
Suman Banerjee, UW-Madison

4. Objective High throughput routing
September 2005
Objective
High throughput routing
Maximize amount of traffic from STAs to Mesh Access Points (MAPs)
Tx power control not necessary for energy efficiency
Because mesh nodes are usually powered
Tx power control necessary to reduce contention
And thus achieve high throughput
Choice of Tx power depends on link characteristics
Suman Banerjee, UW-Madison

5. Talk Outline Routing with late binding Cost metrics
September 2005
Talk Outline
Routing with late binding
Cost metrics
Other optimizations
Summary and future work
Suman Banerjee, UW-Madison

6. Traditional routing Well-defined path derived from routing protocol
September 2005
doc.: IEEE /0909r0
September 2005
Traditional routing
Well-defined path derived from routing protocol S 1 2 3 D
Other links
ignored at
packet forwarding
time 4 5 6
Cost (ETX) = 4 (if no failures on path)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

7. Traditional routing Packet loss
September 2005
doc.: IEEE /0909r0
September 2005
Traditional routing
Well-defined path derived from routing protocol
Packet loss S 1 2 3 D 4 5 6
Cost (ETX) = 10 (given a sequence of failures)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

8. Problem with traditional routing
September 2005
Problem with traditional routing
A single deterministic apriori path
No “wireless multicast/broadcast advantage”
Suman Banerjee, UW-Madison

9. Routing with late binding
September 2005
Routing with late binding
Exploit diversity in wireless environments
Bind exact paths after packet transmission in each hop
No pre-determined single path
Exploits “wireless multicast/broadcast advantage”
Gain greater cost efficiency
Works with all reasonable metrics (not just hop count)
Suman Banerjee, UW-Madison

10. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Transmit each packet towards intended next hop
(intended next hop decided as before)
Packet loss S 1 2 3 D 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

11. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Transmit each packet towards intended next hop
(intended next hop decided as before) S 1 2 3 D 4 5 6
Packet loss
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

12. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Packet loss S 1 2 3 D 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

13. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Cost (Num. Tx) = 4
(Compare to Cost = 10 for a deterministic path) S 1 2 3 D 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

14. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Cost (Num. Tx) = 4
(Compare to Cost = 10 for a deterministic path) S 1 2 3 D 4 5 6
Number of path choices = 2 * 2 * 2 = 8
We just need one good path!
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

15. Applicability of late binding
September 2005
Applicability of late binding
Useful only if some links are lossy
Supported by empirical data in multiple studies
No longer have to disregard relatively higher loss links
Exploit them to our advantage
Delivery ratio
Link number
[See Mobicom’03 Couto et.al.]
Suman Banerjee, UW-Madison

16. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Need additional mechanism to deal with redundancy
What if both (or three paths) deliver the same packet? S 1 2 3 D 4 5 6
Interference
Need simple protection mechanisms
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

17. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Protection against redundancy S 1 2 3 D
RTS 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

18. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Protection against redundancy
CTS S 1 2 3 D 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

19. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Protection against redundancy S 1 2 3 D
Data 4 5 6
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

20. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
RTF: Request to Forward
(more than an ACK)
RTF S 1 2 3 D 4 5 6
4 on receiving 1’s RTF suppresses its own RTF
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

21. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
Each potential forwarder that did not hear another RTF, sends an RTF S 1 2 3 D
RTF 4 5 6
(If 4 did not hear 1’s RTF)
Staggered in time probabilistically, based on cost
(Cost(1,D) < Cost(4,D) and so 1 sends first)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

22. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding
CTF: Confirm to Forward S 1 2 3 D
CTF(1) 4 5 6
S confirms forwarding role for only one of them
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

23. Routing with late binding
September 2005
doc.: IEEE /0909r0
September 2005
Routing with late binding S 1 2 3 D
RTS 4 5 6
Repeats at next hop and so on …
RTF send by nodes that improve cost to destination
(4 will not send RTF to 1)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

24. Additional issues addressed
September 2005
Additional issues addressed
Handle re-orders and implications for TCP
Handle multicast traffic
Details have been worked out, conducting further evaluations
Suman Banerjee, UW-Madison

25. Talk Outline Routing with late binding Cost metrics
September 2005
Talk Outline
Routing with late binding
Cost metrics
Other optimizations
Summary and future work
Suman Banerjee, UW-Madison

26. Towards a routing metric
September 2005
Towards a routing metric
Consider the single-channel case
If packet loss rate = p, then expected number of transmissions is 1 / (1-p) [under independence assumption]
If packet size = s and Tx data rate = B, then channel occupancy duration is s / B
Need to consider number of interfered nodes as well
Reduction in network capacity
Implications on transmit power to be used
Suman Banerjee, UW-Madison

27. A proposed metric Loss rate = p Intefered node count = n
September 2005
doc.: IEEE /0909r0
September 2005
A proposed metric
Loss rate = p
Intefered node count = n
Tx data rate = B
Packet size = s
Cost = (n * s) / [B (1-p)] S 1 2 3 D 4 5 6
Loss rate = 0.5, Tx data rate = 1 Mbps
Interfered nodes = 4, Packet size = 1 Kb
Cost = 0.008
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

28. Metric for multi-channel case
September 2005
Metric for multi-channel case
Need to account for interference effects due to channel assignment
Possible to design good metrics (left out of scope here)
Suman Banerjee, UW-Madison

29. Talk Outline Routing with late binding Cost metrics
September 2005
Talk Outline
Routing with late binding
Cost metrics
Other optimizations
Summary and future work
Suman Banerjee, UW-Madison

30. Transmit power control
September 2005
doc.: IEEE /0909r0
September 2005
Transmit power control
Data rate: 1 Mbps
Packet size: 1 Kb
Link Loss Rate = 0% S D
Cost ~ 4 * 4 = 16 (Tx range = 1)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

31. Transmit power control
September 2005
doc.: IEEE /0909r0
September 2005
Transmit power control
Link Loss Rate = 0% S D
Cost ~ 8 * 4 = 32 (Tx range = 1.5)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

32. Transmit power control
September 2005
doc.: IEEE /0909r0
September 2005
Transmit power control
Link Loss Rate = 0% S D
Cost ~ 12 * 2 = 24 (Tx range = 2)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

33. Transmit power control
September 2005
Transmit power control
When loss rate = 0%, optimal tx range = 1
i.e., guarantee minimal connectivity of topology
Changes with loss rate in channels
Suman Banerjee, UW-Madison

34. Transmit power control
September 2005
doc.: IEEE /0909r0
September 2005
Transmit power control
Data rate: 1 Mbps
Packet size: 1 Kb
Link Loss Rate = 50% D S
Traditional
deterministic
routing
Cost ~ 8 * 2 = 16 (Tx range = 1)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

35. Transmit power control
September 2005
doc.: IEEE /0909r0
September 2005
Transmit power control
Link Loss Rate = 50% D S
Traditional
deterministic
routing
Cost ~ 16 * 1 = 16 (Tx range = 1.5)
Suman Banerjee, UW-Madison
Suman Banerjee, UW-Madison

36. Transmit power control
September 2005
Transmit power control
For 50% loss on channels and given example:
Traditional deterministic routing: Tx = 1 is same as Tx = 1.5
However, result is different for routing with late binding:
Tx 1.5 is better than Tx = 1
Also Tx = 1 (late binding) is better than Tx = 1.5 (traditional)
Bottomline: Tx power control had to re-considered for routing with late binding
Details offline
Suman Banerjee, UW-Madison

37. Load balancing May have multiple alternate mesh gateway nodes
September 2005
Load balancing
May have multiple alternate mesh gateway nodes
Typically STAs can choose any such gateway
Choice should be governed by traffic loads on paths
Defined an updated cost metric to reflect this
Suman Banerjee, UW-Madison

38. Summary of work Routing with late binding is beneficial
September 2005
Summary of work
Routing with late binding is beneficial
Re-visited two problems
Tx power control
Channel assignment (for multi-channel wireless meshes)
Results indicate that proposed mechanisms are efficient
Implementation on our wireless mesh testbed
Publicly available soon (please send if interested)
Proposed mechanisms have wider applicability (as our other results show)
Energy efficient reliable routing in wireless sensor networks
Efficient geographic routing
etc.
Questions and comments:
Suman Banerjee, UW-Madison

39. September 2005
Contact Information
Suman Banerjee
Wisconsin Wireless and NetworkinG Systems (WiNGS) Laboratory
Department of Computer Sciences
University of Wisconsin
Madison, WI 53706, USA
Suman Banerjee, UW-Madison