Exploiting Diversity in Wireless Networks

1
Exploiting Diversity in Wireless Networks

• Nitin H. Vaidya
• University of Illinois at Urbana-Champaign
• www.crhc.uiuc.edu/wireless
• Presentation at Mesh Networking Summit
• Snoqualmie, WA, June 23-24, 2004

2
Capacity of Wireless Networks

• Limited by
• Interference
• Available spectrum
• Need to find ways to get most out of available
• spectrum

3
Diversity / Multiplicity / Heterogeneity

• Diversity provides flexibility in using available
  resources
• Can help improve performance

4
Diversity / Multiplicity / Heterogeneity

• Research Agenda
• Abstractions that capture diversity
• Protocols that exploit diversity

5
Diversity / Heterogeneity

• Many dimensions
• Physical layer
• Architecture
• Upper layer

6
Channel Diversity
7
Channel Diversity

• Multiple channels can help improve performance
• Obvious approaches
• Exploit diversity to choose channel with best
  gain
• Use multiple channels simultaneously to improve
  capacity
• Developing practical protocols for the obvious
  approaches is still a challenge

8
Alternative Approach

• Exploit protocol characteristics to benefit from
  the diversity
• Examples
• Pipelining
• Backup routes

9
IEEE 802.11

• Channel contention resolved using backoff(and
  optional RTS/CTS)

Backoff
Data / ACK
RTS/CTS
10
Simple Observation

• Backoff keeps channel idle ? unproductive
• Most protocols have such idle contention periods

Unproductive
Backoff
Data / ACK
RTS/CTS
11
Pipelining Using Multiple Channels

• Control Channel Backoff and RTS/CTS
• Data Channel Data and ACK

Backoff
RTS/CTS
Backoff
RTS/CTS
RTS/CTS
Backoff
Stage 1
Data / ACK
Data / ACK
Stage 2
12
Pipelining works well only if pipeline stages are
balanced !
Control Channel
Data / ACK
Data / ACK
Data Channel
13
Solution Partial Pipelining

• Only partially resolve channel contention in the
  pipelined stage

14
Partial Pipelining

• Stage 1 Narrow-Band Busy Tone Channel
• Stage 2 Data channel

Control Channel
Data / ACK
Data / ACK
Data Channel
15
Partial Pipelining

• No packets transmitted on busy tone channel
• Bandwidth can be small

16
Partial Pipelining

• By migrating backoff to a narrow-band channel,
  cost of backoff is reduced

Data Channel Bandwidth
Area cost of backoff
Busy Tone Channel Bandwidth
Backoff Duration
17
Moral of the Story

• Looking beyond physical layerdiversity
  exploitation schemes helps
• Protocol characteristics can be exploited

18
Another Example
19
Multiple Interfaces

• Consider devices equipped with both 802.11a and b

802.11a 802.11b
Higher max rate Lower max rate
Lower range Higher range
20
Channel Diversity

• 802.11b network
• denser than the 802.11a network
• but provides lower rate
• Example approach
• Use 802.11a as primary network
• Use 802.11b network to provide backup routes when
  802.11a routes fail
• The 802.11b network could be used for other
  things too

21
Protocol Interactions

• For TCP, route failure more painful than a
  degradation in available capacity
• The backup routes can avoid a route failure
• Benefits of added capacity can be magnified by
  exploiting protocol behavior

22
Research Agenda

• Develop practical protocols that can exploit
  diversity
• Pay attention to protocol characteristics

23
Antenna Heterogeneity
24
Antenna Heterogeneity

• Fixed beam antennas prevalent on mobile devices
• Omnidirectional antennas (often with diversity)
• Other antennas likely to become more prevalent
• Switched, steered, adaptive, smart
• Can form narrow beamforms, which may be changed
  over time
• Re-configurable antennas
• Beamforms can be changed over time by
  reconfiguring the antenna, but not necessarily
  narrow beams

25
Antenna Heterogeneity

• Beamforms All antennas are not made equal
• Timescale Can beamforms be changed at packet
  timescales?

26
Protocol Design

• Protocols designed for fixed beam antennas
  inadequate with movable beam antennas
• State of the art
• MAC Protocols for specific antenna
  capabilities

27
Research Challenge

• How to design antenna-adaptive protocols ?
• Need to develop suitable antenna abstractions
  that span a range of antenna designs
• Forces us to think about essential
  characteristics of antennas
• Example Variability of beamforms a more
  fundamental property than directionality

28
Diversity / Heterogeneity

• Many dimensions
• Physical layer
• Architecture
• Upper layer

29
Pure Ad Hoc Networks

• No infrastructure
• All communication over (one or more) wireless hops

B
C
D
E
A
Z
Ad hoc connectivity
Y
X
30
Hybrid Networks

• Infrastructure Ad hoc connectivity

infrastructure
AP1
AP2
B
C
D
E
A
Z
Ad hoc connectivity
Y
X
31
Hybrid Networks
R

• Infrastructure may include wireless relays

infrastructure
AP1
AP2
R
P
C
B
R
D
A
Z
Ad hoc connectivity
Y
X
32
Hybrid Networks

• Heterogeneity
• Some hosts connected to a backbone, most are not
• Access points/relays may have more processing
  capacity, energy

33
Heterogeneity Beneficial

• Infrastructure provides a frame of reference
• Provide location-aware services
• Reduce route discovery overhead

AP0
AP1
AP2
AP3
R3
R2
R1
D
B
A
34
Heterogeneity Beneficial

• Reduce diameter of the network
• Lower delay
• Potentially greater per-flow throughput

35
Infrastructure Facilitates New Trade-Offs
(hypothetical curves)
overhead
connectivity
User density distribution affects the trade-off
Ad hoc-ness
36
Research Issues

• How to trade complexity with performance ?
• Parameterize ad hoc-ness ?
• Should the spectrum be divided between
  infrastructure and ad hoc components?
• What functionality for relays / access points?

37
Misbehavior
38
Misbehavior

• Misbehavior occurs with limited resources
• Violating protocol specifications benefits
  misbehaving hosts
• Example Small backoffs in 802.11 ? higher
  throughput

39
Research Agenda

• Protocols that maximize performance while
  discouraging/penalizing misbehavior
• Challenge
• Wireless channel prone to temporal and spatial
  variations
• Different players see different channel state
• Impossible to detect misbehavior 100 reliably

40
Conclusions
41
Conclusions

• Diversity/Heterogeneity natural to wireless
  networks
• Need better abstractions to capture the diversity
• Need protocols that can exploit available
  diversity
• Need to be able to survive misbehavior

42
Other Research

• Distributed algorithms for multi-hop wireless
  networks
• Clock synchronization
• Message ordering
• Leader election
• Mutual exclusion

43
Thanks! www.crhc.uiuc.edu/wireless

• Advertisement
• National Summit for Community Wireless Networks
• Urbana-Champaign, Illinois
• August 20-22, 2004
• http//www.cuwireless.net