Cognitive Wireless Networking in the TV Bands by ecj13059

VIEWS: 12 PAGES: 71

									    Networking Devices
    over White Spaces
                 Ranveer Chandra

                      Collaborators:
Thomas Moscibroda, Rohan Murty, Victor Bahl, Srihari Narlanka
           Wi-Fi’s Success Story
• Wi-Fi is extremely popular (billion $$ business)
   – Enterprise/campus LANs, Home networks, Hotspots


• Why is Wi-Fi successful
   – Wireless connectivity: no wires, increased reach
   – Broadband speeds: 54 Mbps (11a/g), 200 Mbps (11n)
   – Free: operates in unlicensed bands, in contrast to
     cellular
          Problems with Wi-Fi
• Poor performance:
  – Contention with Wi-Fi devices
  – Interference from other devices in 2.4 GHz, such
    as Bluetooth, Zigbee, microwave ovens, …


• Low range:
  – Can only get to a few 100 meters in 2.4 GHz
  – Range decreases with transmission rate
   Overcoming Wi-Fi’s Problems
• Poor performance:
  – Fix Wi-Fi protocol – several research efforts (11n,
    MIMO, interference cancellation, …)
  – Obtain new spectrum?


• Low range:
  – Operate at lower frequencies?
                   Analog TV  Digital TV
Higher Frequency




                         USA (2009)
                        Japan (2011)
                       Canada (2011)
                               Broadcast TV
                          UK (2012)
                        China (2015)
                             ….     Wi-Fi (ISM)
                             ….
                             …..
                                                  5
             What are White Spaces?
             TV           Wireless Mic                         ISM (Wi-Fi)

 0 54-88 170-216 470     700                2400 2500           5180   5300          7000
MHz•50 TV Channels   -60                                                             MHz
                                  “White spaces”

   •Each channel is 6 MHz wide
                    dbm

   •FCC Regulations*                                    TV Stations in America

       •Sense TV stations and Mics
                -100
       •Portable devices on channels 21 - 51
                     470 MHz Frequency 700 MHz


            White Spaces are Unoccupied TV Channels                              6
Why should we care
about White Spaces?



                      7
         The Promise of White Spaces
               TV          Wireless Mic                     ISM (Wi-Fi)

 0 54-90 174-216 470       700             2400 2500         5180   5300       7000
MHz                                                                            MHz

                           Up to 3x of 802.11g




                                                  }
                      More                              Potential Applications

                    Spectrum                            Rural wireless broadband
                                                        City-wide mesh
                                                            ……..

                       Longer                               ……..

                       Range     at least 3 - 4x of Wi-Fi                  8
Goal: Deploy Wireless Network




            Base Station
                (BS)

      Good throughput for all nodes

     Avoid interfering with incumbents
                                         9
 Why not reuse Wi-Fi
based solutions, as is?



                          10
                                 White Spaces Spectrum Availability
                                 0.8
                                                                  Urban
 Fraction of Spectrum Segments




                                 0.7                                            Differences from ISM(Wi-Fi)
                                 0.6                              Suburban
                                                                                Fragmentation
                                 0.5                                                 Variable channel widths
                                                                  Rural
                                 0.4
                                 0.3
                                 0.2

1 20.13 4 5                                              1 2 3 4 5
                                  0
                                       1   2       3    4     5        6   >6
                                               # Contiguous Channels
                    Each TV Channel is 6 MHz wide  Use multiple channels for more bandwidth
                                             Spectrum is Fragmented


                                                                                                         11
  White Spaces Spectrum Availability
                                                    Differences from ISM(Wi-Fi)
                                                    Fragmentation
                                                          Variable channel widths

                                                    Spatial Variation
                                                          Cannot assume same
                                                          channel free everywhere


1 2 3 4 5               1 2 3 4 5
                                            TV
                                          Tower


 Location impacts spectrum availability  Spectrum exhibits spatial variation


                                                                              12
  White Spaces Spectrum Availability
                                                 Differences from ISM(Wi-Fi)
                                                 Fragmentation
                                                       Variable channel widths

                                                 Spatial Variation
                                                       Cannot assume same
                                                       channel free everywhere

                                                 Temporal Variation
1 2 3 4 5                1 2 3 4 5                     Same Channel will
                                                       not always be free
                                                       Any connection can be
                                                       disrupted any time

Incumbents appear/disappear over time  Must reconfigure after disconnection

                                                                            13
         Cognitive (Smart) Radios
1. Dynamically identify currently unused portions of spectrum
2. Configure radio to operate in available spectrum band
      take smart decisions how to share the spectrum
         Signal Strength




                                         Signal Strength
                           Frequency
                                                           Frequency
                  Networking Challenges
       The KNOWS Project (Cogntive Radio Networking)

   How should nodes connect?              How should they discover
                                          one another?


Which spectrum-band should two
cognitive radios use for transmission?
    1. Frequency…?
    2. Channel Width…?
                                                        Need analysis tools to
    3. Duration…?                                       reason about capacity &
                                                        overall spectrum
                                                        utilization




                             Which protocols should we use?
                MSR KNOWS Program
                              Prototypes
• Version 1: Ad hoc networking in white spaces
   – Capable of sensing TV signals, limited hardware functionality, analysis of
     design through simulations



• Version 2: Infrastructure based networking (WhiteFi)
   – Capable of sensing TV signals & microphones, deployed in lab



• Version 3: Campus-wide backbone network (WhiteFi +
  Geolocation)
   – Deployed on campus, and provide coverage in MS Shuttles
   Version 2: WhiteFi System
Prototype Hardware Platform
 Base Stations and Clients
Algorithms and Implementation
 Discovery
 Spectrum Assignment
 Handling Disconnections
Evaluation
 Deployment of prototype nodes
 Simulations                     17
               Hardware Design
• Send high data rate signals in TV bands
   – Wi-Fi card + UHF translator
• Operate in vacant TV bands
   – Detect TV transmissions using a scanner
• Avoid hidden terminal problem
   – Detect TV transmission much below decode threshold
• Signal should fit in TV band (6 MHz)
   – Modify Wi-Fi driver to generate 5 MHz signals
• Utilize fragments of different widths
   – Modify Wi-Fi driver to generate 5-10-20-40 MHz signals
KNOWS Platform: Salient Features
• Can dynamically adjust channel-width and
  center-frequency.
• Low time overhead for switching
   can change at fine-grained time-scale
                                    Transceiver can tune
                                  to contiguous spectrum
                                        bands only!




                      Frequency
           Changing Channel Widths
Scheme 1: Turn off certain subcarriers ~ OFDMA




                           10 MHz
                           20


Issues: Guard band? Pilot tones? Modulation scheme?
           Changing Channel Widths
Scheme 2: reduce subcarrier spacing and width!
 Increase symbol interval




                           10 MHz
                           20


   Properties: same # of subcarriers, same modulation
        Adaptive Channel-Width
                                                     20Mhz
                                              5Mhz
• Why is this a good thing…?

                                             Frequency
1. Fragmentation
    White spaces may have different sizes
    Make use of narrow white spaces if necessary


2. Opportunistic, load-aware channel allocation
    Few nodes: Give them wider bands!
    Many nodes: Partition the spectrum in narrower bands
KNOWS White Spaces Platform

         Windows PC
                                       Scanner (SDR)
            TV/MIC
                             FFT               UHF RX
           detection               FPGA
                                            Daughterboard



  Net
 Stack
                                   Whitespace Radio
           Connection Manager

                                   Wi-Fi            UHF
           Atheros Device Driver   Card          Translator
              Variable Channel
               Width Support



                                                              25
        WhiteFi System Challenges
Fragmentation    Spatial    Temporal       Impact
                Variation   Variation


                                          Discovery

                                          Spectrum
                                         Assignment

                                        Disconnection


                                                    26
      Discovering a Base Station


                   Discovery Problem

    1 2 3 Goal
          4 5                              BS is 3 4
                     Quickly find channels 1 2 using 5



Discovery Time = (B x W)

       BS we  Try new client channel and
       How does themust use same channels
  Fragmentation Clients different center discover widths
      Can and optimize this discovery time?
            channels used by the BS?
                                                           27
Whitespaces Platform: Adding SIFT

              PC
                                       Scanner (SDR)
            TV/MIC
                            FFT                UHF RX
           detection               FPGA
                                            Daughterboard


            Temporal Analysis
    Net          (SIFT)            Whitespace Radios
   Stack

           Connection Manager

                                   Wi-Fi            UHF
           Atheros Device Driver   Card          Translator




SIFT: Signal Interpretation before Fourier Transform
                                                              28
 SIFT, by example
           5 MHz
           10MHz



                                                        ADC             SIFT




           SIFT                                 Data
                                           Beacon              ACK
                                                               Beacon
                               Amplitude

  Does not decode packets                               SIFS
Pattern match in time domain


                                                 Time
                                                                         29
BS Discovery: Optimizing with SIFT



             1 2 3 4 5                              1 2 3 4 5
                                                        18 MHz
 Amplitude




                       Matched against 18 MHz packet signature



                Time

              SIFT enables faster discovery algorithms
                                                                 30
BS Discovery: Optimizing with SIFT
Linear SIFT (L-SIFT)

                          1 2 3 4 5


Jump SIFT (J-SIFT)

                       1 2 3 4 5 6 7 8



                                         31
     Discovery: Comparison to Baseline
                          1               Baseline =(B x W)
                         0.9                L-SIFT = (B/W)                    Linear-SIFT
                                            J-SIFT = (B/W)
(compared to baseline)




                         0.8
  Discovery Time Ratio




                                                                               Jump-SIFT
                         0.7
                                              2X reduction
                         0.6
                         0.5
                         0.4
                         0.3
                         0.2
                         0.1
                          0
                               0   30        60         90         120           150              180
                                        White Space - Contiguous Width (MHz)
                                                                                             32
        WhiteFi System Challenges
Fragmentation    Spatial    Temporal       Impact
                Variation   Variation


                                           Discovery

                                          Spectrum
                                         Assignment

                                        Disconnection


                                                    33
   Channel Assignment in Wi-Fi




         1   6   11               1    6   11



Fixed Width Channels  Optimize which channel to use

                                                       34
  Spectrum Assignment in WhiteFi

             Spectrum Assignment Problem

               Goal        Maximize Throughput

              Include       Spectrum at clients
         1 2 3 4 5                  1 2 3 4 5
                              Center Channel
              Assign                &
                                  Width

Fragmentation  Optimize for both, center channel and width
    Spatial Variation  BS must use channel iff free at client
                                                             35
    Accounting for Spatial Variation



 1 2 3 4 5              1 2 3 4 5               1 2 3 4 5




1 2 3 4 5      1 2 3 4 5      1 2 3 4 5   =   1 2 3 4 5



                                                       36
                                Intuition
Intuition
Use widest possible channel
                                                       BS
But
Limited by most busy channel
                                            1 2 3 4 5


 Carrier Sense Across All Channels

 All channels must be free
      ρBS(2 and 3 are free) = ρBS(2 is free) x ρBS(3 is free)
                    Tradeoff between wider channel widths
                  and opportunity to transmit on each channel    37
                        Multi Channel Airtime Metric (MCham)
                           3.5
                             3         20 Mhz     10 MHz
Throughput (Mbps)




                           2.5         5 MHz
                             2
                           1.5
                             1
                                                        W             BS
                                                                    W) n ( c )
                           0.5
                             0
                                  MChamn (F, W) =
                                                    5 Mhz c( F ,
                                 0      10           20             30         40         50
                                                        1 2 Packet 5
                                              Background traffic - 3 4 delay (ms)
                  Pick (F, W) that maximizes
                          2.5
                      20 Mhz     10 MHz
            2            *
                    (NMHzMChamBS + ΣnMChamn)
              ρn(c) = Approx.(2) Time1on Channel 2
                ρBS(2)  Free opportunity node n will
                      5
                             ρBS Air
          MCham-value




          1.5
  ρBS(2) = Max (Free Air Time on channel 2, 1/Contention)
                                      Contention
                         get to transmit on channel c
            1
                          0.5
                            0
                                 0          10          20             30            40   50
                                                 Background traffic - Packet delay (ms)
                                                                                               38
                    WhiteFi Prototype Performance
                              25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

                      5
                    4.5                        WhiteFi              OPT
Throughput (Mbps)




                      4
                    3.5
                      3
                    2.5
                      2
                    1.5
                      1
                    0.5
                      0
                          0    25   50   75   100     125     150   175   200   225   250
                                                    Seconds
                                                                                            39
        WhiteFi System Challenges
Fragmentation    Spatial    Temporal       Impact
                Variation   Variation


                                           Discovery

                                          Spectrum
                                         Assignment

                                        Disconnection


                                                    40
                MSR KNOWS Program
                              Prototypes
• Version 1: Ad hoc networking in white spaces
   – Capable of sensing TV signals, limited hardware functionality, analysis of
     design through simulations



• Version 2: Infrastructure based networking (WhiteFi)
   – Capable of sensing TV signals & microphones, deployed in lab



• Version 3: Campus-wide backbone network (WhiteFi +
  Geolocation)
   – Deployed on campus, and provide coverage in MS Shuttles
Geo-location Service
     Shuttle Deployment
 World’s first urban white space network!
Goal: Provide free Wi-Fi Corpnet access in MS shuttles
    • Use white spaces as backhaul, Wi-Fi inside shuttle
    • Obtained FCC Experimental license for MS Campus
    • Deployed antenna on rooftop, radio in building & shuttle
    • Protect TVs and mics using geo-location service & sensing
       Some Results
Demo
       Summary & On-going Work
• White Spaces enable new networking scenarios
• KNOWS project researched networking problems:
  –   Spectrum assignment: MCham
  –   Spectrum efficiency: variable channel widths
  –   Network discovery: using SIFT
  –   Network Agility: Ability to handle disconnections
• Ongoing work:
  – MIC sensing, mesh networks, co-existence among
    white space networks, …

                                                          45
Questions
SIGCOMM 2008 Talk
       A Case for
 Adapting Channel Width
  in Wireless Networks

  Ranveer Chandra, Ratul Mahajan,
  Thomas Moscibroda, Victor Bahl
        Microsoft Research
         Ramya Raghavendra
University of California, Santa Barbara
     Adaptation in Wireless Networks
• Existing knobs:
     – Transmit rate/Modulation: auto rate algorithms
        • Adapt how tightly bits are packed in spectrum
     – Transmit power: TPC algorithms
        • Adapt tx power for connectivity, spectrum reuse
     –…

• This paper:
     – Channel Width: how & why?

49
              Channelization in IEEE 802.11
               802.11 uses 20 MHz wide channels

                                  70 MHz

2402 MHz                     2427 MHz       2452 MHz   2472 MHz

              2412 MHz

           1      2      3    6                11

   2407 MHz

                                   20 MHz


   50
         Why Adapt Channel Widths?
                              One Scenario


  More spectrum 
       + more capacity (Shannon’s)
       – higher idle power consumption (coming up)


       Challenge:
       Dynamically determine app demand
               20 MHz
              40 MHz
               5

       & adapt channel width

         When intensive apps, go least for best data rate
For throughput idle, go narrow for widerpower consumption
  51
                Our Contributions

• Demonstrate feasibility of dynamic channel
  width adaptation on off-the-shelf hardware

• Characterize properties of channel widths
     – Throughput, range, energy consumption


• SampleWidth to dynamically select best
  channel width
52
        Implementing Variable Widths
                                                                         Antenna


                 Typical Wireless Card

                  Baseband/MAC                     RF Component
                    (coding/decoding,              (PLLs, upconverters
                   timing, encryption)              Power Amplifiers)



                                     REF CLOCK

Modify driver to programmatically tune clock frequency

                Channel width proportional to clock frequency
  53
         Variable Channel Widths in OFDM
In 802.11: 48 data subcarriers, 4 pilots
           Pilot tone

                                                Data
                                                Subcarriers



                                       20 MHz


     Subcarrier Spacing: 0.3125 MHz

     At 20 MHz:
     Guard Interval: 0.8 s
     Symbol Period = 1/0.3125 s + GI = 4 s

54
         Variable Channel Widths in OFDM
To reduce width to 10 MHz, halve the clock frequency
           Pilot tone

                                                   Data
                                                   Subcarriers



                                        20 MHz
                                        10


     Subcarrier Spacing: 0.3125/2 MHz

     At 10 MHz:
     Guard Interval: 0.8*2 s
     Symbol Period = (1/0.3125 s + GI)*2 = 8 s

55
               Our Implementation
• Using Atheros cards on Windows
     – Implemented 5, 10, 20, 40 MHz

     – MAC parameters scale with clock
        • e.g. SIFS: 20 s at 20 MHz, 40 s at 10 MHz


     – We keep 802.11 slot time constant for interop



56
     Properties of Channel Widths
Impact on:

• Throughput

• Transmission Range

• Battery Power


57
             Experimental Setup
• Conducted (clean) experiment
     – Using attenuator & CMU emulator




• Indoor experiments at MSR & UCSB
• Outdoor experiments in large park

58
                                           Throughput
• Throughput increases with channel width
                      – (Shannon’s) Capacity = Bandwidth * log (1 + SNR)
                      – In practice, protocol overheads come into play
                          • Twice bandwidth has less than double throughput
                                                                       Actual Data Rate:
     UDP Throughput (in Mbps)




                                  5 MHz       10 MHz

                                  20 MHz      40 MHz                   108 Mbps@40 MHz

                                                                       54 Mbps@20 MHz

                                                                       27 Mbps@10 MHz

                                                                       13.5 Mbps@5MHz

                                            Modulation
59
                         Transmission Range

• Reducing channel width increases range
            – Narrow channel widths have same signal energy but lesser noise
                                               ~ 3 dB
               better SNR
                              5MHz
                              10MHz
                              20MHz
     Loss Rate




                              40MHz




60
                                Attenuation (dB)
                     Impact of Guard Interval

• Reducing width increases guard interval
 more resilience to delay spread (more range)
     Loss Rate (%)




                                                  5MHz
                                                  10MHz
                                                  20MHz
                                                  40MHz

                           Delay Spread (in ns)

61
         Need for Width Adaptation
With auto rate:

                      40 MHz

                               20 MHz



                                        10 MHz
                                                 5 MHz




          There is no single best channel width!
62
               Energy Consumption

• Lower channel widths consume less power
   – Similar to CPU clock scaling
               5MHz    10MHz      20MHz     40MHz
      Send      1.92    1.98       2.05      2.17
       Idle     1.00    1.11       1.25      1.41
     Receive    1.01    1.13       1.27      1.49
• When idle, lowest channel width is best
• During send/receive, best energy/bit width depends on
  distance

63
        Recap: Channel Width Properties


   • When nodes are near, higher channel widths have
     more throughput

   • Lower channel widths have more range
        – Better SNR, resilience to delay spread


   • Lower channel widths consume less power
Lower widths increase range while consuming less power!

   64
          Application: Song Sharing



Zune Social over Wi-Fi
1. Zunes advertise (periodically beacon) their song list
2. Interested Zunes download songs from peers

Issues: throughput, power!

  Our Solution: Adapt channel width based on traffic
  (SampleWidth)
  65
       SampleWidth for Throughput
Goal: Use minimum width that satisfies demand




• Algorithm:
     – Start at minimum width – best energy, range
     – When interface queue is full, probe higher width
        • During song transfer
     – Periodically probe adjacent (higher/lower) widths
     – Return to minimum width when no traffic
66
       SampleWidth Evaluation
• SampleWidth adapts to best throughput width




67
     Reducing MB
          Start 20
                   Power Consumption
                       file transfer
                       @ 25 sec
                                             5MHz    10MHz
                                             20MHz   40MHz
      Energy (Joule)




                                       Seconds
68
                             SampleWidth for Energy
     Total Energy (Joules)


                                                      ~ 25%
                                                      savings




69
         Application Scenarios

1. Throughput/energy-aware song sharing

2. Load aware spectrum allocation in WLANs

3. Improved capacity in 802.11

4. Cognitive (DSA-based) networking

70
                           Summary
• Channel width can be adapted
     – On off-the-shelf hardware
     – To improve application performance
     – To design better, more efficient networks


• Future work
     – Explore other channel width strategies
        • e.g. modifying number of subcarriers
     – Communication across channel widths
        • Nodes on different widths cannot communicate
     – Build larger systems using adaptive channel widths
71
                 Questions?




http://research.microsoft.com/netres/projects/spawn/




72
73

								
To top