1. 1 Mobile Computing and Wireless Networking Chengzhi Li University of Virginia chengzhi@cs.virginia.edu www.cs.virginia.edu/~cl4v

2. 2 What is Mobile Computing  Building distributed system with mobile computers and wireless networking – Mobile networking – MAC, Routing, Reliable data transport, … – Mobile information access – Disconnected operation, … – Adaptive applications – Proxies, transcoding, … – Energy aware systems – Goal-directed adaptation, … – Location sensitivity – GPS, …

3. 3 Evolution of Computing Single User OS Batch Timesharing Networking LANs + WSs Mobile Computing More Freedom from Collocation More Flexible Resource Usage

4. 4 Challenges  Battery constraints  limited wireless transmission range  limited life time  Broadcast nature of the wireless medium  Hidden & exposed terminal problems  Ease of snooping on wireless transmissions (security hazard)  Mobility  route changes  packet losses  network partitions

5. 5 Problem Space Link Network Transport Physical Upper layers

6. 6 Cellular Wireless Network infrastructure Internet

7. 7 Mobile Ad Hoc Network l Provide differentiated QoS levels to different wireless applications l Achievable by QoS-sensitive MAC and network layer scheduling

8. 8 Mobile Ad Hoc Networks  Mobile distributed multiple-hop wireless network  Formed by wireless hosts which may be mobile  Without necessarily using a pre-existing infrastructure  Routes between nodes may potentially contain multiple hops

9. 9 Applications of Ad Hoc Network  NTDR (Near Term Digital Radio) is the only “real” (non-prototypical) Ad Hoc network in use today.  NTDR use clustering and link state routing and self- organized into a two tier ad hoc network

10. 10 Why Wireless Networks ?  Potential ease of deployment  Decreased dependence on infrastructure

11. 11 Many Applications  Personal area networking  cell phone, laptop, ear phone, wrist watch  Military environments  soldiers, tanks, planes  Civilian environments  taxi cab network  meeting rooms  sports stadiums  boats, small aircraft  Emergency operations  search-and-rescue  policing and fire fighting

12. 12 Physical Layer  Traditionally, not much interaction between physical layer and upper layers  Many physical layer mechanisms not beneficial without help from upper layers  Example: Adaptive modulation

13. 13 Power Control  Transmit power determines  “Range” of a transmission  Interference caused at other nodes BCDA

14. 14 Benefits of Power Control  Transmit a packet with least transmit power necessary to deliver to the receiver  Save energy: Important benefit to battery-powered hosts  Reduce interference  Can allow greater spatial reuse

15. 15 Power Control  Power control introduces asymmetry  D transmits to C at low power, but B uses high transmit power to transmit to A  B may not know about D-to-C transmission, but can interfere with it BCDA

16. 16 Power Control  Transmit power determines  “Range” of a transmission  Interference caused at other nodes BCDA

17. 17 Power Control  Proposals for medium access control and routing with power control exist  Do not solve the problem satisfactorily  Ideal solution will  Reduce energy consumption, and  Maximize spatial reuse

18. 18 Link Layer

19. 19 Hidden Terminals & RTS/CTS Handshake

20. 20 Hidden Terminal Problem  Node B can communicate with A and C both  A and C cannot hear each other  When A transmits to B, C cannot detect the transmission using the carrier sense mechanism  If C transmits to D, collision will occur at B BCA D

21. 21 RTS/CTS Handshake  Sender sends Ready-to-Send (RTS)  Receiver responds with Clear-to-Send (CTS)  RTS and CTS announce the duration of the transfer  Nodes overhearing RTS/CTS keep quiet for that duration  RTS/CTS used in IEEE 802.11 D C BA CTS (10) RTS (10) 10

22. 22 Exposed Terminals & RTS, CTS, and Dual Busy Tones

23. 23 Exposed Terminal Problem  Node C can communicate with B and D both  Node B can communicate with A and C  Node A cannot hear C  Node D can nor hear B  When C transmits to D, B detect the transmission using the carrier sense mechanism and postpone to transmit to A, even though such transmission will nor cause collision BC D A

24. 24 Network Layer

25. 25 Mobile Ad Hoc Networks  May need to traverse multiple links to reach a destination

26. 26 Mobile Ad Hoc Networks  Mobility causes route changes

27. 27 Transport Layer

28. 28 Transport Protocols

29. 29 TCP  TCP performance degrades in presence of route failures  TCP cannot distinguish between packet losses due to route change and due to congestion  Reduces congestion window in response Unnecessary degradation in throughput

30. 30 TCP  TCP performance degrades in presence of route failures  TCP cannot distinguish between packet losses due to route change and due to congestion  Reduces congestion window in response Unnecessary degradation in throughput

31. 31 Problems in Ad Hoc Networking

32. 32 Problem Space  Practical considerations  Consumer demand or lack thereof  Standardization  Government regulations  Technical issues

33. 33 Physical Layer  Traditionally, not much interaction between physical layer and upper layers  Many physical layer mechanisms not beneficial without help from upper layers  Example: Adaptive modulation

34. 34 Adaptive Modulation  Channel conditions are time-varying AB

35. 35  Choose modulation scheme as a function of channel conditions

36. 36 Adaptive Modulation  If physical layer chooses the modulation scheme transparent to MAC  MAC cannot know the time duration required for the transfer  Must involve MAC protocol in deciding the modulation scheme  Some 802.11-compliant implementations use a sender- based scheme for this purpose  Receiver-based schemes can perform better

37. 37 Sender-Based “Autorate Fallback” MAC Protocol D C BA 1Mbps 2Mbps  Sender decreases rate after N consecutive ACKS are not received  Sender increases rate after Y consecutive ACKS are received DATA 2Mbps

38. 38 Performance of Sender-Based “Autorate Fallback” Expected ARF CCK (11Mbps) CCK (5.5Mbps) QPSK (2Mbps) BPSK (1Mbps)

39. 39 1Mbps 2Mbps  Sender sends RTS containing its best rate estimate  Receiver chooses best rate for the conditions and sends it in the CTS  Sender transmits DATA packet at new rate  Information in data packet header implicitly updates nodes that heard old rate Receiver-Based Autorate MAC Protocol D C BA CTS (1) RTS (2) 2 1

40. 40 Physical Layer  Several other physical layer capabilities call for changes to upper layers of protocol stack  Example: Power control

41. 41 Directional / Smart Antennas  Various capabilities  Sectored antennas (fixed beam positions)  Beam steering  Tracking a transmitter  MAC and routing protocols for ad hoc networks using such antennas  How to take into account antenna capabilities? Network may be heterogeneous

42. 42 Physical Layer  Are ad hoc networks benefiting from the progress made at physical layer ?  Other interesting areas  Efficient coding schemes  Various diversity techniques

43. 43 Physical Layer: Simulation Models  Insufficient accuracy in commonly used physical layer models  Physical link state is not binary as often assumed  Reliable packet reception does not depend just on distance  Transmit power  Interference level  Fading  Need to use realistic models  Modulation scheme  Coding

44. 44 Link Layer

45. 45 Interesting Link Layer Issues  Medium access control  Retransmission mechanisms  Transmission scheduling  Which pending packet should a node attempt to transmit?  Adaptive parameter selection  Frame size  Retransmission limit

46. 46 QoS in Medium Access Control  Many proposals for achieving fairness  Fair scheduling schemes attempt to provide equitable sharing of channel  Unpredictable nature of transmission errors makes it difficult to make hard guarantees  Need to develop a probabilistic framework

47. 47 QoS in MAC  Easier in a centralized protocol (such as 802.11 point coordination function), than in a distributed protocol  Distributed MAC appears more suitable for ad hoc networks, however  Perhaps a hybrid protocol will be best  How to design such a protocol ?

48. 48 Transmission Scheduling  When multiple packets pending transmission, which packet to transmit next?  Choice should depend on  Receiver status (blocked by some other transmission?)  Congestion at receivers  Noise level at receivers  Tolerable delay for pending packets –Need interaction between upper layers and MAC

49. 49 MAC for Multiple Channels  How to split bandwidth into channels?  How to use the multiple channels ? Dedicated channel for control ?

50. 50 Network Layer

51. 51 Reactive versus Proactive Routing  Reactive protocols  Maintain routes between nodes that need to communicate  Proactive protocols  Maintain routes between all node-pairs  Lot of activity on routing protocol design

52. 52 Routing  Reactive and proactive protocols are quite well-understood  Designing reactive protocols: “Solved” problem  Designing proactive protocols: “Solved” problem  At least, when using common assumptions about the network  Interesting problems exist when other issues are considered (such as QoS or physical layer properties)

53. 53 Reactive versus Proactive  Choice of protocol depends on  Mobility characteristics of the nodes  Traffic characteristics  How to design adaptive protocols ?  Existing proposals use a straightforward combination of reactive and proactive  Proactive within “radius” K  Reactive outside K  Choose K somehow

54. 54 Reactive versus Proactive  Need a more flexible way to manage protocol behavior  Assign proactive/reactive tag to each route (A,B) ?  How to determine when proactive behavior is better than reactive ?

55. 55 Address Assignment  How to assign addresses to nodes in an ad hoc network ?  Static assignment  Easier to guarantee unique address  Dynamic assignment  How to guarantee unique addresses when partitions merge?  Do we need to guarantee unique addresses ?

56. 56 Transport Protocols

57. 57 TCP  Several solutions have been proposed to fix this  These techniques somehow inform TCP sender that the packet losses are due to route failure  TCP does not decrease congestion window in response

58. 58 TCP  New route may differ significantly from old route  Proposals for TCP-over-ad-hoc tend to use old timeout and congestion window after a route change  Does not seem like a good idea  How to choose appropriate timeout and congestion window after detecting a route change ?

59. 59 Other Issues

60. 60 Algorithms

61. 61 Distributed Algorithms  Rich body of work on distributed algorithms in traditional distributed environments  Shared memory  Message ordering  Clock synchronization  Leader election

62. 62 Distributed Algorithms  Existing algorithms can usually be used on ad hoc networks without affecting correctness  Performance on ad hoc networks may not be good  Existing algorithm treat link repairs/failures as random events  With mobility, link failure/repairs are correlated with host movement

63. 63 Distributed Algorithms  How to design distributed algorithm exploiting the correlation between mobility and link failure/repair ?

64. 64 Distributed Algorithms  Traditionally, complexity is measured as a function of problem “size”  Number of nodes  Number of failures  How to analyze algorithm complexity as a function of mobility ?  What measure of mobility is amenable to such an analysis ?  Need to capture the correlation without making the measure too complex

65. 65 Security Issues

66. 66 What’s New ?  Wireless medium easy to snoop on  With ad hoc networking, hard to guarantee connectivity  Easier for intruders to insert themselves into network

67. 67 Authentication  How to authenticate a node ?  May not have access to a certification authority

68. 68 Resource Depletion Attack  Intruders may send data with the objective of congesting a network or depleting batteries A CB D T intruder U Bogus traffic

69. 69 Routing Attacks  Intruders may mis-route the data  not delivering it to the destination at all, or  delaying it significantly  How to detect such attacks ?  How to tolerate such attacks ?

70. 70 Traffic Analysis  Despite encryption, an eavesdropper can identify traffic patterns  Traffic patterns can divulge information about the operation mode  Traffic analysis can be prevented by presenting “constant” traffic pattern –Insert dummy traffic  How to make this cheaper ?

71. 71 Other Issues

72. 72 Incentives for Ad Hoc Routing  Why should I forward packets for some other nodes ?  Need some incentive mechanism  Policies to determine reward for performing each operation

73. 73 Applications  New applications for ad hoc networks ?

74. 74 Hybrid Environments  Use infrastructure when convenient  Use ad hoc connectivity when necessary or superior E A BS1BS2 X Z infrastructure Ad hoc connectivity

75. 75 Summary

76. 76 Summary  Plenty of interesting research problems  Research community disproportionately obsessed with routing protocols

77. 77 Summary  Interesting problems elsewhere at the two ends of the protocol stack  How to design algorithms and applications ?  How to exploit physical layer techniques ? Increase interaction between physical layer and upper layers Link Network Transport Physical Upper layers

78. 78 Summary  Hybrid environments require revisiting protocol design decisions