1. 1 State of Multi-Hop Wireless Networking Nitin Vaidya Electrical and Computer Engineering University of Illinois at Urbana-Champaign Sept. 15. 2008

2. 2 Deep Thought So the secret to good self-esteem is to lower your expectations to the point where they're already met ? Calvin (and Hobbes) : Bill Watterson

3. 3 Caveat This talk is based on opinions not necessarily substantiated by real data

4. 4 Multi-Hop Wireless  Many possibilities …

5. 5 Multi-Hop Wireless  Mobile ad hoc networks Interconnect cars, planes, tanks, soldiers, people

6. 6 Multi-Hop Wireless  Mesh networks (roof-top, pole-top) internet Mesh Client

7. 7 Multi-Hop Wireless  Sensor networks

8. 8 Multi-Hop Wireless  Opportunistic Delay/Disruption/Disconnection-tolerant networks

9. 9 Why Multi-Hop Wireless ?

10. 10 Why Multi-Hop Wireless ?  Lack of infrastructure

11. 11 Why Multi-Hop Wireless ?  Some clients difficult to reach directly via infrastructure, due to obstacles AP Relay

12. 12 Why Multi-Hop Wireless ?  Decreasing dependence on wired infrastructure  Add wireless “infrastructure” internet Mesh Client

13. 13 Why Multi-Hop Wireless ?  Low-power clients unable to communicate directly with infrastructure

14. 14 Why Multi-Hop Wireless ?  For improved capacity High interference High transmit power

15. 15 Why Multi-Hop Wireless ?  For improved capacity Low interference Low transmit power

16. 16 Why Multi-Hop Wireless ?  Poor connectivity

17. 17 A Selective History 1973-87 DARPA Packet Radio Networks (PRNET/SURAN) 1997 IEEE 802.11 1997 IETF MANET 1999 TinyOS 2000 MeshNetworks founded 2000+ CUWiN open-source mesh 2000 ACM MobiHoc 2001 “Embedded, Everywhere”  Sensor networks 2001 Interplanetary Internet, IETF draft, Vint Cerf 2003 ACM Sensys 2004 Motorola acquires MeshNetworks 2004 IEEE 802.11s study group for mesh networking 2004 ZigBee Time

18. 18 Research Activity versus Relevance

19. 19 Research Activity Much activity in  Mobile ad hoc networks No infrastructure Large diameter High mobility  Sensor networks Low power Large diameter Small diameter useful in practice, but not “interesting”

20. 20 Unscientific Measure of Interest: Google 9/11/08  Ad hoc networks: 2,290,000  Mesh networks: 764,000  Sensor networks: 1,670,000  Vehicular networks: 1,710,000  Delay tolerant networks: 196,000  Disruption tolerant networks: 206,000  Disconnection tolerant networks: 99,800  Opportunistic networks: 978,000  Magna Carta (1215) 2,630,000  United states constitution (1787) 5,790,000  Paris Hilton 68,800,000  Computer architecture 21,400,000  802.11 66,000,000

21. 21 Research Activity Most activity seems to be in  Mobile ad hoc networks No infrastructure Large diameter High mobility  Sensor networks Low power Large diameter Extreme assumptions make the problem exciting  But what about relevance ?

22. 22 Relevance ?

23. 23 Relevance  Not all networks are made equal …  Some are likely to be commonplace others limited to niche scenarios Relevance In increasing order of relevance …

24. 24 Delay Tolerant Networks  Limited to niche scenarios

25. 25 Interesting Variation  Wireless Graffiti  Microblogs  “Sticky notes in-the-air” Users leave information “in the air” at some location  Others can retrieve later from there May be viewed as opportunistic communication (Not quite the same as DTN)

26. 26 Mobile Ad Hoc Networks  Why design networks without infrastructure ?  Possible to deploy some infrastructure in most environments

27. 27 Sensor Networks  Wireless sensors are important  Important to network the sensors  Sensors + Network ≠ Large diameter

28. 28 Infrastructure Extension  Most compelling reason for multi-hop wireless  Only a small number of hops! AP Relay

29. 29 Infrastructure Extension  Mesh (Wireless “infrastructure”) internet

30. 30 Summary: Most Appealing Scenario  Some wired infrastructure  Capacity scales with the infrastructure  Small diameter wireless extension for the infrastructure Using relays or peer-to-peer  Better reachability  Low-power operation  Reduced capacity loss

31. 31 If only small diameter networks matter, did we waste our time ? Not quite …  Interference management and MAC-related issues somewhat independent of network diameter

32. 32 State of Multi-Hop Wireless Very large volume of activity  Beautiful theory Asymptotic Capacity Throughput-optimal scheduling Network utility optimization Network coding Cooperative relaying

33. 33 State of Multi-Hop Wireless Very large volume of activity  Practical protocols & deployments Many wireless standards And many more MAC & routing protocols Many experimental deployments Mesh devices Sensor devices Start-ups

34. 34 State of Multi-Hop Wireless Very large volume of activity  (Too) Many conferences and workshops  Plenty of research funding Compared to many other areas

35. 35 State of Multi-Hop Wireless Despite the volume of activity  Difficult to enumerate core set of principles for wireless network design What should we teach in an undergraduate wireless networks class ?

36. 36 State of Multi-Hop Wireless Despite the volume of activity  Theoretical developments haven’t been translated to practice  Much protocol design ignores physical layer issues Much talk of cross-layer design, but progress not impressive

37. 37 What is Lacking ? Meaningful contact between  Practice  Networking  Theory  Comm Picture from Wikipedia

38. 38 Net-X Theory to Practice Multi-channel protocol Channel Abstraction Module IP Stack Interface Device Driver User Applications ARP Interface Device Driver OS improvements Software architecture Capacity bounds channels capacity Net-X testbed CSL A B C D E F Fixed Switchable Insights on protocol design Linux box

39. 39 Things I Wish I Had Learned in Kindergarten

40. 40 Those who cannot learn from history are doomed to repeat it With apologies to George Santayana outgrow 1

41. 41  Relaying : Multi-hop routes (store-and-forward) Pre-History of Wireless Communication: Smoke Signals, Fires, Semaphore

42. 42 Pre-History of Wireless Communication: Homing Pigeons  Exploiting mobility

43. 43 Reusing Ideas Reasonable, but Need to Explore Better Alternatives No wired-equivalent for wireless networks No links !

44. 44 Wireless Channel Offers Rich Diversity Current protocols exploit diversity only to a limited extent The vanishing link : Diversity confuses the notion of a link Layer 1 : 2+ gap

45. 45 Interference is Information 2

46. 46 Interference is Information A B D C Signal Interference

47. 47 Bits Are Not Automobiles 3

48. 48 Bits Are Not Automobiles  We treat information networks same as physical transportation networks Planes, Trains and Automobiles  Bits can be combined (encoded) and separated, unlike physical objects

49. 49 Network Coding ACB P P Q Q

50. 50 Network Coding ACB P Q P +Q Q + Q P

51. 51 Physics Does Not Know Layers 4

52. 52 Physics Does Not Know Layers  Layering is an abstraction, not a theorem  Backpressure scheduler ( “ throughput-optimal ” ) spans traditional layers 1 through 3: arg max ∑ W( l ) r( l ) r Є Rate l Region

53. 53 Physics Does Not Know Layers  Layering is useful, but need a principled approach to identifying appropriate cross-layer exchange  Great start towards this: Network utility optimization »Queue as price  Shortcomings: »Not all requirements easy to capture as concave utility »Framework does not (yet) yield enough insight on practical “scheduling/routing”

54. 54 Opportunism Pays 5

55. 55 Opportunism Pays  Channel variations make it difficult to predict short-term optimal in advance  Late binding can work better –Opportunistic beamforming –Opportunistic routing –MAC-Layer anycasting –…

56. 56 State of Multi-Hop Wireless  Theoretical developments haven’t been translated to practice  Much protocol design ignores physical layer issues  Much talk of cross-layer design, but progress not impressive

57. 57 State of Multi-Hop Wireless Despite the volume of activity  Theoretical developments haven’t been translated to practice  Much protocol design ignores physical layer issues  Much talk of cross-layer design, but progress not impressive

58. 58 What Now ? Four-Point Agenda

59. 59  Reduce the unknown unknowns  Increase phy content in CS/CE networking courses –Awareness of phy necessary to ask better questions –Phy community should help  Educate phy students about higher layer issues 1. Educate Better Ourselves & Next Generation

60. 60 If you have influence at funding agencies …  Resist temptation to create new networking programs Partitioning of resources creates false demand –Remove existing partitions Possible to encourage research without these –Past examples: NOSS, FIND? 2. Fewer Research Programs

61. 61 3. Fewer “Better” Conferences  Increase venues that encourage diverse community interactions (phy-networking, theory-applied) More Workshops, fewer “selective” conferences, (fewer papers!) Co-located conferences Tutorials  Eliminate most (wireless) networking conferences Emulate Info Theory model ?

62. 62 4. Greater Industry/User Feedback  What are the industry-perceived long-term challenges ?  What do they need from us ?  Not everything needs to be dictated by industry, but practical insights can benefit academic research –Problem formulations constrained by reality

63. 63 Summary: Multi-Hop Wireless Networks  Enormous progress in past 15 years  But potential for much more impact  Need greater attention to cross-layer design  Improved education a prerequisite

64. 64 Advertisement

65. 65 Illinois Wireless Summer School  August 3-7, 2009  Illinois Center for Wireless Systems (ICWS) at the University of Illinois at Urbana-Champaign  Lectures ranging antennas-to-applications  Opportunities for students to interact  Sponsorships welcome !

66. 66 Thanks!

67. 67 Thanks!