1. 1 Heterogeneity in Multi-Hop Wireless Networks Nitin H. Vaidya University of Illinois at Urbana-Champaign www.crhc.uiuc.edu/~nhv © 2003 Vaidya

2. 2 Summary  Heterogeneity is essential  Heterogeneity is beneficial Research Agenda  Develop protocols that exploit the heterogeneity  Develop mechanisms to better evaluate wireless systems Proof by example …

3. 3 Heterogeneity  Many dimensions of heterogeneity:  Architecture  Physical capability of hosts  Higher layers

4. 4 Architecture Multi-hop wireless networks  Pure ad hoc networks  Hybrid networks

5. 5 Pure Ad Hoc Networks  No “infrastructure”  All communication over (one or more) wireless hops E A BC D X Z Ad hoc connectivity Y

6. 6 Hybrid Networks  Infrastructure + Ad hoc connectivity E A BC D AP1AP2 X Z infrastructure Ad hoc connectivity Y

7. 7 Hybrid Networks  Infrastructure may include wireless relays E A C D AP1AP2 X Z infrastructure Ad hoc connectivity Y B R P

8. 8 Hybrid Networks  Heterogeneity  Some hosts connected to a backbone, most are not  Access points may have more processing capacity, energy E A BC D AP1AP2 X Z infrastructure Ad hoc connectivity Y

9. 9 Hybrid Networks  Heterogeneous wireless technologies E A BC D AP1AP2 X Z infrastructure Y Type 1 (3G) Type 2 (802.11)

10. 10 Hybrid Networks  Heterogeneity is essential  Pure ad hoc or pure infrastructure networks inadequate for many environments  Heterogeneity is beneficial …

11. 11 Benefit over Pure Ad Hoc Networks  Infrastructure provides a frame of reference  Can assign approximate locations to the mobiles –Provide location-aware services –Reduce route discovery overhead AP0AP1AP2AP3 A B D R2 R1 R3 A

12. 12 Benefit over Pure Ad Hoc Networks  Infrastructure can reduce diameter of the network  Lower delay  Potentially greater per-flow throughput E A C D AP1AP2 X Z infrastructure Ad hoc connectivity Y B R P

13. 13 Infrastructure Facilitates New Trade-Offs (hypothetical curves) User density distribution affects the trade-off Ad hoc-ness = K connectivity overhead Poor Man’s Ad Hoc Network

14. 14 Infrastructure Helps in Resource Allocation Address Assignment  Unique IP addresses need to be assigned to hosts in a network  DHCP used in traditional networks  Difficult to use DHCP in pure ad hoc networks  But Can also be deployed on the infrastructure in a hybrid network

15. 15 Infrastructure Helps in Resource Allocation Address Assignment  Impossible to detect address duplication in networks that can get partitioned Unbounded delays cause difficulty  Clusters of hosts may partition from the infrastructure, rejoin, over time  Need a mechanism to assign unique addresses despite partitions Impossible with unbounded message delays

16. 16 If a problem cannot be solved Change the problem

17. 17 Weak Duplicate Address Detection Packets from a given host to a given address should be routed to the same destination, despite duplication of the address  Achievable despite unbounded delay, but incurs overhead  Infrastructure to the rescue: Use weak DAD only for nodes partitioned from the infrastructure  Can this extend to other resource allocation problems?

18. 18 Benefit over Pure Infrastructure Networks  Ad hoc routing increases the “reach” of the infrastructure  Connectivity can be traded with overhead  Example: Limit “ad hoc-ness” to K hops

19. 19 Hybrid Networks: Research Issues How to implement infrastructure?  How to deploy relays/access points? Density, distribution  What functionality should be given to relays and access points? Should they cooperate? With each other? With mobiles?  Are relays an optimization or necessary components?  Should the spectrum be divided between the infrastructure and ad hoc components?

20. 20 Hybrid Networks: Research Issues How to design protocols?  How to trade “complexity” with “performance” ? Parameterize ad hoc-ness ?  How to design protocols that maximize “performance” for a given complexity? Power control: How should the heterogeneity affect power control? MAC: Should the infrastructure do more work? Routing: Reduce overhead using infrastructure Transport: How to approach theoretical capacity bounds? How to deal with potentially unbounded delays?  The answers to the above questions are inter-dependent Power control, MAC, routing, transport protocols affect each other’s behavior Cross-layer design needed

21. 21 Heterogeneity  Many dimensions of heterogeneity:  Architecture  Physical capability of hosts –Antennas –Topology control mechanisms –Processing capability –Energy availability  Higher layers

22. 22 Antenna Capabilities  “Fixed beam” antennas prevalent on mobile devices  Omnidirectional antennas  “Movable beam” antennas likely to become more prevalent over time  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  Different devices may incorporate different antennas

23. 23 Antenna Heterogeneity All antennas are not made equal  Beamforms: Only directional, or omni too?  Timescale: Can beams be “moved” at packet timescales?  Single beam or multiple beams?  Variations with time?

24. 24 Antenna Capabilities  Protocols designed for omnidirectional (fixed beam) antennas inadequate with movable beam antennas  State of the art: MAC Protocols designed for specific antenna capabilities  Need “antenna-adaptive” MAC and routing protocols that allow for antenna heterogeneity

25. 25 Antenna Heterogeneity  Heterogeneity is essential  Enforcing homogeneity will limit benefits from antenna improvements  Heterogeneity is beneficial  Devices can employ best antennas that they can “afford” –Device constraints: energy, processing, size, weight, $$ –Access points may use more capable antenna than mobiles  Antenna-adaptive protocols allow separation of the antenna as a “layer” in the protocol stack

26. 26 Antenna Heterogeneity: 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 beam patterns a more fundamental property than directionality

27. 27 Evaluation of Wireless Networks

28. 28 Capacity  Capacity analysis: Capacity results useful to determine the gap between actual performance and the best case scenario Significant progress in recent years Need further work to model heterogeneous environments

29. 29 Evaluation of Wireless Protocols  Benchmarks: Need benchmarks for comparison of different protocols State of the art: Toy benchmarks, almost no real data (for evaluating multi-hop wireless networks)  Simulations Commonly used simulation models are poor Need better physical layer models accessible to protocol community

30. 30 Evaluation of Wireless Protocols  Experimentation:  “Full scale” experiments not always practical  Need mechanisms to build and experimentally evaluate “scaled models” of the network –Physical dimensions –Mobility –Number of hosts –Traffic density  How to “scale down” the network, and still maintain essential behaviors?

31. 31 Conclusions

32. 32 Conclusions  Heterogeneity essential, and beneficial  Heterogeneity  Complexity ?  Not necessarily  Thinking about heterogeneity useful in arriving at better abstractions  Need protocols that can exploit heterogeneity  Need approaches for realistic comparative evaluation of protocols

33. 33 Thanks!