SlideShare a Scribd company logo
1 of 61
Download to read offline
Augmen'ng	
  Mobile	
  3G	
  Using	
  WiFi

            By:	
  Aruna	
  Balasubramanian,	
  Ratul	
  Mahajan,	
  Arun	
  
                                  Venkataramani



                                                       Presenter:	
  Ke	
  Huang

Tuesday, April 12, 2011                                                            1
Demand	
  for	
  mobile	
  access	
  growing
                          	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  www.totaltele.com




                                                                                              h+p://www.readwriteweb.com




                                                                                                                           2

Tuesday, April 12, 2011                                                                                                        2
Demand	
  for	
  mobile	
  access	
  growing
                                                                                                                                     	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  www.totaltele.com




                                                                                                                                                                                                                                             h+p://www.readwriteweb.com




   900	
  million	
  mobile	
  broadband	
  subscrip'ons	
  today….
   	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
   www.3gamericas.org
   	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
                                                                                                                                                                                                                        2

Tuesday, April 12, 2011                                                                                                                                                                                                                                                                                                                  2
Mobile	
  demand	
  is	
  projected	
  to	
  far	
  




                                                                    3

Tuesday, April 12, 2011                                                 3
Mobile	
  demand	
  is	
  projected	
  to	
  far	
  
                           www.rysavy.com
 Current	
  spectrum             409.5	
  MHz
 Unallocated	
  spectrum	
   230	
  MHz
 (including	
  whitespaces)
 Projected	
  demand	
  by	
     800	
  MHz	
  –	
  
 2016	
                          1000	
  MHz




                                                                    3

Tuesday, April 12, 2011                                                 3
Mobile	
  demand	
  is	
  projected	
  to	
  far	
  
                                                        www.nyCmes.com
                           www.rysavy.com
 Current	
  spectrum             409.5	
  MHz
 Unallocated	
  spectrum	
   230	
  MHz
 (including	
  whitespaces)
 Projected	
  demand	
  by	
     800	
  MHz	
  –	
  
 2016	
                          1000	
  MHz




                                                                    3

Tuesday, April 12, 2011                                                  3
Mobile	
  demand	
  is	
  projected	
  to	
  far	
  
                                                                                         www.nyCmes.com
                               www.rysavy.com
 Current	
  spectrum                  409.5	
  MHz
 Unallocated	
  spectrum	
   230	
  MHz
 (including	
  whitespaces)
 Projected	
  demand	
  by	
          800	
  MHz	
  –	
  
 2016	
                               1000	
  MHz


                                                                                          www.nyCmes.com
“In	
  light	
  of	
  the	
  limited	
  natural	
  resource	
  of	
  spectrum,	
  we	
  have	
  to	
  look	
  at	
  
   the	
  ways	
  of	
  conserving	
  spectrum”	
  -­‐-­‐	
  Mark	
  Siegel	
  (AT&T)

             Reducing	
  cellular	
  spectrum	
  u'liza'on	
  is	
  key!                                      3

Tuesday, April 12, 2011                                                                                            3
How	
  can	
  we	
  reduce	
  spectrum	
  usage?
                                                blogs.chron.com

     1.	
  Behavioral


     2.	
  Economic



     3.	
  Technical
                                                              4

Tuesday, April 12, 2011                                           4
How	
  can	
  we	
  reduce	
  spectrum	
  usage?
                                                  blogs.chron.com

     1.	
  Behavioral

                                        www.usatoday.com

     2.	
  Economic



     3.	
  Technical
                                                                4

Tuesday, April 12, 2011                                             4
How	
  can	
  we	
  reduce	
  spectrum	
  usage?
                                                  blogs.chron.com

     1.	
  Behavioral

                                        www.usatoday.com

     2.	
  Economic



     3.	
  Technical
                                                                4

Tuesday, April 12, 2011                                             4
Augmen'ng	
  Mobile	
  3G	
  using	
  WiFi

     	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Offload	
  data	
  to	
  WiFi	
  when	
  possible

     	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Focus	
  on	
  vehicular	
  mobility




                                                                                                               5

Tuesday, April 12, 2011                                                                                            5
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Offloading	
  3G	
  data	
  to	
  WiFi




                                                           6

Tuesday, April 12, 2011                                        6
Related	
  work	
  on	
  mul'ple	
  interfaces
     Improving	
  performance	
  using	
  handoffs	
  based	
  on	
  
       current	
  condi'ons
     Reducing	
  power	
  consump'on	
  by	
  switching	
  
       across	
  mul'ple	
  interfaces	
  




                                                                   7

Tuesday, April 12, 2011                                                7
Related	
  work	
  on	
  mul'ple	
  interfaces
     Improving	
  performance	
  using	
  handoffs	
  based	
  on	
  
       current	
  condi'ons
     Reducing	
  power	
  consump'on	
  by	
  switching	
  
       across	
  mul'ple	
  interfaces	
  



    This	
  work:
    1.How	
  much	
  3G	
  data	
  can	
  be	
  offloaded	
  to	
  WiFi?
    2.How	
  to	
  offload	
  without	
  hur'ng	
  applica'ons?
                                                                         7

Tuesday, April 12, 2011                                                      7
Contribu'ons
     Measurement:	
  	
  Joint	
  study	
  of	
  3G	
  and	
  WiFi	
  
      connec'vity
            Across	
  three	
  ci'es:	
  Amherst,	
  Seagle,	
  SFO


     System:	
  Wiffler,	
  to	
  offload	
  3G	
  data	
  to	
  WiFi	
  while	
  
       respec'ng	
  applica'on	
  constraints	
  
            Deployed	
  on	
  20	
  vehicles


                                                                                 8

Tuesday, April 12, 2011                                                              8
Measurement	
  setup
     Testbed:	
  Vehicles	
  with	
  3G	
  and	
  WiFi	
  (802.11b)	
  radios
            Amherst:	
  20	
  buses	
  +	
  1	
  car,	
  Seagle:	
  1	
  car,	
  SFO:	
  1	
  car


     Soiware:	
  Simultaneously	
  probes	
  3G	
  and	
  WiFi	
  for	
  
            Availability,	
  loss	
  rate,	
  throughput


     Dura'on:	
  3000+	
  hours	
  of	
  data	
  over	
  12+	
  days



                                                                                                    9

Tuesday, April 12, 2011                                                                                 9
Open	
  WiFi	
  availability	
  low,	
  but	
  useful
  Availability	
  	
  =	
  frac'on	
  of	
  1-­‐second	
  intervals	
  when	
  at	
  least	
  
  one	
  packet	
  received
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  



          Availability
             (%)




                                                                                            10

Tuesday, April 12, 2011                                                                          10
Open	
  WiFi	
  availability	
  low,	
  but	
  useful
  Availability	
  	
  =	
  frac'on	
  of	
  1-­‐second	
  intervals	
  when	
  at	
  least	
  
  one	
  packet	
  received
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
                          86%


          Availability
             (%)




                                                                                            10

Tuesday, April 12, 2011                                                                          10
Open	
  WiFi	
  availability	
  low,	
  but	
  useful
  Availability	
  	
  =	
  frac'on	
  of	
  1-­‐second	
  intervals	
  when	
  at	
  least	
  
  one	
  packet	
  received
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
                          86%


          Availability
             (%)


                          11%


                                                                                            10

Tuesday, April 12, 2011                                                                          10
Open	
  WiFi	
  availability	
  low,	
  but	
  useful
  Availability	
  	
  =	
  frac'on	
  of	
  1-­‐second	
  intervals	
  when	
  at	
  least	
  
  one	
  packet	
  received
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
                          86%


          Availability
             (%)


                          11%                                                         7%


                                                                                            10

Tuesday, April 12, 2011                                                                          10
Open	
  WiFi	
  availability	
  low,	
  but	
  useful
  Availability	
  	
  =	
  frac'on	
  of	
  1-­‐second	
  intervals	
  when	
  at	
  least	
  
  one	
  packet	
  received
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
                          86%


          Availability
             (%)
    3G+WiFi	
  combinaCon	
  be+er	
  than	
  sum	
  pf	
  parts
                          11%                                                         7%


                                                                                            10

Tuesday, April 12, 2011                                                                          10
WiFi	
  loss	
  rate	
  is	
  higher
  Loss	
  rate	
  =	
  Frac'on	
  of	
  packets	
  lost	
  at	
  10	
  probes/sec




             Cumulative
              fraction                         WiFi

                                          3G




                                                                                    11

Tuesday, April 12, 2011                                                                  11
WiFi	
  loss	
  rate	
  is	
  higher
  Loss	
  rate	
  =	
  Frac'on	
  of	
  packets	
  lost	
  at	
  10	
  probes/sec




             Cumulative
              fraction            28%	
          WiFi

                                            3G




                                                                                    11

Tuesday, April 12, 2011                                                                  11
WiFi	
  loss	
  rate	
  is	
  higher
  Loss	
  rate	
  =	
  Frac'on	
  of	
  packets	
  lost	
  at	
  10	
  probes/sec




             Cumulative
              fraction            28%	
          WiFi

                                   8%	
     3G




                                                                                    11

Tuesday, April 12, 2011                                                                  11
WiFi	
  (802.11b)	
  throughput	
  is	
  lower
        Throughput	
  =	
  Total	
  data	
  received	
  per	
  second


                                             WiFi
                          Cumulative
                           fraction
                                       3G              Upstream



                                                WiFi
                          Cumulative
                           fraction
                                       3G           Downstream


                                                                        12

Tuesday, April 12, 2011                                                      12
WiFi	
  (802.11b)	
  throughput	
  is	
  lower
        Throughput	
  =	
  Total	
  data	
  received	
  per	
  second


                                             WiFi
                          Cumulative
                           fraction
                                       3G              Upstream



                                                WiFi
                          Cumulative
                           fraction
                                       3G           Downstream


                                                                        12

Tuesday, April 12, 2011                                                      12
WiFi	
  (802.11b)	
  throughput	
  is	
  lower
        Throughput	
  =	
  Total	
  data	
  received	
  per	
  second


                                                   WiFi
                          Cumulative
                           fraction
                                              3G             Upstream
                                       0.35



                                                      WiFi
                          Cumulative
                           fraction
                                              3G          Downstream


                                                                        12

Tuesday, April 12, 2011                                                      12
WiFi	
  (802.11b)	
  throughput	
  is	
  lower
        Throughput	
  =	
  Total	
  data	
  received	
  per	
  second


                                                   WiFi
                          Cumulative
                           fraction
                                            3G               Upstream
                                       0.35 0.72



                                                      WiFi
                          Cumulative
                           fraction
                                           3G             Downstream


                                                                        12

Tuesday, April 12, 2011                                                      12
WiFi	
  (802.11b)	
  throughput	
  is	
  lower
        Throughput	
  =	
  Total	
  data	
  received	
  per	
  second


                                                   WiFi
                          Cumulative
                           fraction
                                            3G               Upstream
                                       0.35 0.72



                                                      WiFi
                          Cumulative
                           fraction
                                           3G             Downstream


                                                                        12

Tuesday, April 12, 2011                                                      12
Implica'ons	
  of	
  measurement	
  study
     Strawman	
  augmenta'on:	
  Use	
  WiFi	
  when	
  
       available
            Can	
  offload	
  only	
  ~11%	
  of	
  the	
  'me
            Can	
  hurt	
  applica'ons	
  because	
  of	
  WiFi’s	
  higher	
  loss	
  
              rate	
  and	
  lower	
  throughput




                                                                                          13

Tuesday, April 12, 2011                                                                        13
Key	
  ideas	
  in	
  Wiffler
     Increase	
  savings	
  for	
  delay-­‐   Reduce	
  damage	
  for	
  delay-­‐
       tolerant	
  applica'ons                  sensi've	
  applica'ons
     Problem:	
  Using	
  WiFi	
  only	
      Problem:	
  Using	
  WiFi	
  
       when	
  available	
  saves	
             whenever	
  available	
  can	
  
       ligle	
  3G	
  usage                     hurt	
  applica'on	
  quality
     Solu'on:	
  Exploit	
  delay-­‐          Solu'on:	
  Fast	
  switch	
  to	
  3G	
  
       tolerance	
  to	
  wait	
  to	
  
                                                when	
  WiFi	
  delays	
  
       offload	
  to	
  WiFi	
  when	
  
       availability	
  predicted                exceed	
  threshold



                                                                                      14

Tuesday, April 12, 2011                                                                    14
Predic'on-­‐based	
  offloading
     D	
  =	
  Delay-­‐tolerance	
  threshold	
  (seconds)
     S	
  =	
  Data	
  remaining	
  to	
  be	
  sent	
  (bytes)


     Each	
  second,
     1. If	
  (WiFi	
  available),	
  send	
  data	
  on	
  WiFi	
  


     2. Else	
  if	
  (W(D)	
  <	
  S),	
  send	
  data	
  on	
  3G


     3. Else	
  wait	
  for	
  WiFi.




                                                                       15

Tuesday, April 12, 2011                                                     15
Predic'on-­‐based	
  offloading
     D	
  =	
  Delay-­‐tolerance	
  threshold	
  (seconds)
     S	
  =	
  Data	
  remaining	
  to	
  be	
  sent	
  (bytes)


     Each	
  second,
     1. If	
  (WiFi	
  available),	
  send	
  data	
  on	
  WiFi	
  


     2. Else	
  if	
  (W(D)	
  <	
  S),	
  send	
  data	
  on	
  3G    Predicted	
  WiFi	
  
                                                                       transfer	
  size	
  in	
  
     3. Else	
  wait	
  for	
  WiFi.                                   next	
  D	
  seconds	
  



                                                                                                    15

Tuesday, April 12, 2011                                                                                  15
Predic'ng	
  WiFi	
  capacity
     History-­‐based	
  predic'on	
  of	
  #	
  of	
  APs	
  using	
  last	
  few	
  
       AP	
  encounters	
  
            WiFi	
  capacity	
  =	
  (expected	
  #APs)	
  x	
  (capacity	
  per	
  AP)
     Simple	
  predictor	
  yields	
  low	
  error	
  both	
  in	
  Amherst	
  
       and	
  Seagle




                                                                                          16

Tuesday, April 12, 2011                                                                        16
Predic'ng	
  WiFi	
  capacity
     History-­‐based	
  predic'on	
  of	
  #	
  of	
  APs	
  using	
  last	
  few	
  
       AP	
  encounters	
  
            WiFi	
  capacity	
  =	
  (expected	
  #APs)	
  x	
  (capacity	
  per	
  AP)
     Simple	
  predictor	
  yields	
  low	
  error	
  both	
  in	
  Amherst	
  
       and	
  Seagle


       Negligible	
  benefits	
  with	
  more	
  sophis'cated	
  predic'on,	
  eg	
  
       future	
  loca'on	
  predic'on	
  +	
  AP	
  loca'on	
  database
                                                                                          16

Tuesday, April 12, 2011                                                                        16
Fast	
  switching	
  to	
  3G
     Problem:
            WiFi	
  losses	
  bursty	
  =>	
  high	
  retransmission	
  delay


     Approach:
            If	
  no	
  WiFi	
  link-­‐layer	
  ACK	
  within	
  50ms,	
  switch	
  to	
  3G
            Else,	
  con'nue	
  sending	
  on	
  WiFi




                                                                                               17

Tuesday, April 12, 2011                                                                             17
Wiffler	
  implementa'on



                                                Wiffler	
  
                                                proxy

    § Predic'on-­‐based	
  offloading	
  upstream	
  +	
  downstream
    § 	
  Fast	
  switching	
  only	
  upstream
           Ø   Implemented	
  using	
  signal-­‐upon-­‐ACK	
  in	
  driver


                                                                              18

Tuesday, April 12, 2011                                                            18
Evalua'on	
  Roadmap
     Predic'on-­‐based	
  offloading
            Deployment	
  on	
  20	
  DieselNet	
  buses	
  in	
  150	
  sq.	
  mi	
  
              region	
  around	
  Amherst
            Trace-­‐driven	
  evalua'on	
  using	
  throughput	
  data


     Fast	
  switching
            Deployment	
  on	
  1	
  car	
  in	
  Amherst	
  town	
  center
            Trace-­‐driven	
  evalua'on	
  using	
  measured	
  loss/delay	
  
              trace	
  using	
  VoIP-­‐like	
  probe	
  traffic

                                                                                         19

Tuesday, April 12, 2011                                                                       19
Deployment	
  results
                                                                      Data	
  offloaded	
  to	
  WiFi
       Wiffler’s	
  predic'on-­‐based	
  offloading                                      30%
              WiFi	
  when	
  available                                              10%
                    File	
  transfer	
  size:	
  5MB;	
  Delay	
  tolerance:	
  60	
  secs;	
  	
  
                    Inter-­‐transfer	
  gap:	
  random	
  with	
  mean	
  100	
  secs




                                                                                                      20

Tuesday, April 12, 2011                                                                                    20
Deployment	
  results
                                                                       Data	
  offloaded	
  to	
  WiFi
       Wiffler’s	
  predic'on-­‐based	
  offloading                                       30%
              WiFi	
  when	
  available                                               10%
                    File	
  transfer	
  size:	
  5MB;	
  Delay	
  tolerance:	
  60	
  secs;	
  	
  
                    Inter-­‐transfer	
  gap:	
  random	
  with	
  mean	
  100	
  secs


                                                                    %	
  Cme	
  good	
  voice	
  quality	
  
                Wiffler’s	
  fast	
  switching                                         68%
       WiFi	
  when	
  available	
  (no	
  switching)                                42%
                          VoIP-­‐like	
  traffic:	
  20-­‐byte	
  packet	
  every	
  20	
  ms	
  
                                                                                                               20

Tuesday, April 12, 2011                                                                                             20
Trace-­‐driven	
  evalua'on
   Parameters	
  varied
         Workload,	
  AP	
  density,	
  delay-­‐tolerance,	
  switching	
  threshold


   Strategies	
  compared	
  to	
  predic'on-­‐based	
  offloading:
         WiFi	
  when	
  available
         Adapted-­‐Breadcrumbs:	
  Future	
  loca'on	
  predic'on	
  +	
  AP	
  loca'on	
  
           database
         Oracle	
  (Imprac'cal):	
  Perfect	
  predic'on	
  w/	
  future	
  knowledge



                                                                                        21

Tuesday, April 12, 2011                                                                       21
Wiffler	
  increases	
  data	
  offloaded	
  to	
  WiFi
     Workload:	
  Web	
  traces	
  obtained	
  from	
  commuters	
  




                                                                       22

Tuesday, April 12, 2011                                                     22
Wiffler	
  increases	
  data	
  offloaded	
  to	
  WiFi
     Workload:	
  Web	
  traces	
  obtained	
  from	
  commuters	
  




                                             14%
                                                       WiFi	
  when	
  
                                                       available	
  yields	
  
                                                       ligle	
  savings



                                                                             22

Tuesday, April 12, 2011                                                           22
Wiffler	
  increases	
  data	
  offloaded	
  to	
  WiFi
     Workload:	
  Web	
  traces	
  obtained	
  from	
  commuters	
  


                                                       Wiffler	
  close	
  to	
  
                                             42%       Oracle

                                             14%




                                                                              22

Tuesday, April 12, 2011                                                            22
Wiffler	
  increases	
  data	
  offloaded	
  to	
  WiFi
     Workload:	
  Web	
  traces	
  obtained	
  from	
  commuters	
  


                                                       Wiffler	
  close	
  to	
  
                                             42%       Oracle
                                                       Sophis'cated	
  
                                             14%       predic'on	
  yields	
  
                                                       negligible	
  benefit




                                                                            22

Tuesday, April 12, 2011                                                           22
Wiffler	
  increases	
  data	
  offloaded	
  to	
  WiFi
     Workload:	
  Web	
  traces	
  obtained	
  from	
  commuters	
  


                                                              Wiffler	
  close	
  to	
  
                                                   42%        Oracle
                                                              Sophis'cated	
  
                                                   14%        predic'on	
  yields	
  
                                                              negligible	
  benefit



     Wiffler	
  increases	
  delay	
  by	
  10	
  seconds	
  over	
  Oracle.	
  
                                                                                   22

Tuesday, April 12, 2011                                                                  22
Even	
  more	
  savings	
  in	
  urban	
  centers




                                                      23

Tuesday, April 12, 2011                                    23
Fast	
  switching	
  improves	
  quality	
  of	
  
                delay-­‐sensi've	
  applica'ons




                                                                  24

Tuesday, April 12, 2011                                                24
Fast	
  switching	
  improves	
  quality	
  of	
  
                delay-­‐sensi've	
  applica'ons




                                                  58%




                                                                  24

Tuesday, April 12, 2011                                                24
Fast	
  switching	
  improves	
  quality	
  of	
  
                delay-­‐sensi've	
  applica'ons




                                                  58%
                                                  40%




                                                                  24

Tuesday, April 12, 2011                                                24
Fast	
  switching	
  improves	
  quality	
  of	
  
                delay-­‐sensi've	
  applica'ons


                             73%

                                                  58%
                                                  40%




                                                                  24

Tuesday, April 12, 2011                                                24
Fast	
  switching	
  improves	
  quality	
  of	
  
                delay-­‐sensi've	
  applica'ons


                              73%

                                                          58%
                                                          40%




  30%	
  data	
  offloaded	
  to	
  WiFi	
  with	
  40ms	
  switching	
  threshold
                                                                              24

Tuesday, April 12, 2011                                                            24
Future	
  work
     Reduce	
  energy	
  to	
  search	
  for	
  usable	
  WiFi

     Improve	
  performance/usage	
  by	
  predic'ng	
  user	
  
       accesses	
  to	
  prefetch	
  over	
  WiFi

     Incorporate	
  evolving	
  metrics	
  of	
  cost	
  for	
  3G	
  and	
  
       WiFi	
  usage


                                                                                25

Tuesday, April 12, 2011                                                              25
Summary
     Augmen'ng	
  3G	
  with	
  WiFi	
  can	
  reduce	
  pressure	
  on	
  
       cellular	
  spectrum

     Measurement	
  in	
  3	
  ci'es	
  confirms	
  WiFi	
  availability	
  and	
  
      performance	
  poorer,	
  but	
  poten'ally	
  useful

     Wiffler:	
  Predic'on-­‐based	
  offloading	
  and	
  fast	
  switching	
  
      to	
  offload	
  without	
  hur'ng	
  applica'ons



                                                                                     26

Tuesday, April 12, 2011                                                                   26
Summary
     Augmen'ng	
  3G	
  with	
  WiFi	
  can	
  reduce	
  pressure	
  on	
  
       cellular	
  spectrum

     Measurement	
  in	
  3	
  ci'es	
  confirms	
  WiFi	
  availability	
  and	
  
      performance	
  poorer,	
  but	
  poten'ally	
  useful

     Wiffler:	
  Predic'on-­‐based	
  offloading	
  and	
  fast	
  switching	
  
      to	
  offload	
  without	
  hur'ng	
  applica'ons

                              Questions?
                                                                                     26

Tuesday, April 12, 2011                                                                   26
Thank you!




Tuesday, April 12, 2011                27

More Related Content

Viewers also liked

Capacity planning in mobile data networks experiencing exponential growth in ...
Capacity planning in mobile data networks experiencing exponential growth in ...Capacity planning in mobile data networks experiencing exponential growth in ...
Capacity planning in mobile data networks experiencing exponential growth in ...Dr. Kim (Kyllesbech Larsen)
 
Chapter 4 - Digital Transmission
Chapter 4 - Digital TransmissionChapter 4 - Digital Transmission
Chapter 4 - Digital TransmissionWayne Jones Jnr
 
Throughput Calculation for LTE TDD and FDD System
Throughput Calculation for  LTE TDD and FDD SystemThroughput Calculation for  LTE TDD and FDD System
Throughput Calculation for LTE TDD and FDD SystemSukhvinder Singh Malik
 
How to dimension user traffic in LTE
How to dimension user traffic in LTEHow to dimension user traffic in LTE
How to dimension user traffic in LTEAlthaf Hussain
 
Chap 2. lte channel structure .eng
Chap 2. lte  channel structure .engChap 2. lte  channel structure .eng
Chap 2. lte channel structure .engsivakumar D
 
Chapter 3 - Data and Signals
Chapter 3 - Data and SignalsChapter 3 - Data and Signals
Chapter 3 - Data and SignalsWayne Jones Jnr
 

Viewers also liked (9)

Capacity planning in mobile data networks experiencing exponential growth in ...
Capacity planning in mobile data networks experiencing exponential growth in ...Capacity planning in mobile data networks experiencing exponential growth in ...
Capacity planning in mobile data networks experiencing exponential growth in ...
 
Fundamentals of Mobile Network Sharing
Fundamentals of Mobile Network SharingFundamentals of Mobile Network Sharing
Fundamentals of Mobile Network Sharing
 
Chapter 4 - Digital Transmission
Chapter 4 - Digital TransmissionChapter 4 - Digital Transmission
Chapter 4 - Digital Transmission
 
Throughput Calculation for LTE TDD and FDD System
Throughput Calculation for  LTE TDD and FDD SystemThroughput Calculation for  LTE TDD and FDD System
Throughput Calculation for LTE TDD and FDD System
 
How to dimension user traffic in LTE
How to dimension user traffic in LTEHow to dimension user traffic in LTE
How to dimension user traffic in LTE
 
The Economics of 5G
The Economics of 5GThe Economics of 5G
The Economics of 5G
 
LTE Planning
LTE PlanningLTE Planning
LTE Planning
 
Chap 2. lte channel structure .eng
Chap 2. lte  channel structure .engChap 2. lte  channel structure .eng
Chap 2. lte channel structure .eng
 
Chapter 3 - Data and Signals
Chapter 3 - Data and SignalsChapter 3 - Data and Signals
Chapter 3 - Data and Signals
 

More from Beibei Yang

Hubway Half a Million Trip Data
Hubway Half a Million Trip DataHubway Half a Million Trip Data
Hubway Half a Million Trip DataBeibei Yang
 
Semantic Relatedness for Evaluation of Course Equivalencies
Semantic Relatedness for Evaluation of Course EquivalenciesSemantic Relatedness for Evaluation of Course Equivalencies
Semantic Relatedness for Evaluation of Course EquivalenciesBeibei Yang
 
91.650 Paper Presentation
91.650 Paper Presentation91.650 Paper Presentation
91.650 Paper PresentationBeibei Yang
 
Google Kernel Function
Google Kernel FunctionGoogle Kernel Function
Google Kernel FunctionBeibei Yang
 
Class Project Showcase: DNS Spoofing
Class Project Showcase: DNS SpoofingClass Project Showcase: DNS Spoofing
Class Project Showcase: DNS SpoofingBeibei Yang
 
Localization in HCI: Yahoo (US vs. China)
Localization in HCI: Yahoo (US vs. China)Localization in HCI: Yahoo (US vs. China)
Localization in HCI: Yahoo (US vs. China)Beibei Yang
 

More from Beibei Yang (6)

Hubway Half a Million Trip Data
Hubway Half a Million Trip DataHubway Half a Million Trip Data
Hubway Half a Million Trip Data
 
Semantic Relatedness for Evaluation of Course Equivalencies
Semantic Relatedness for Evaluation of Course EquivalenciesSemantic Relatedness for Evaluation of Course Equivalencies
Semantic Relatedness for Evaluation of Course Equivalencies
 
91.650 Paper Presentation
91.650 Paper Presentation91.650 Paper Presentation
91.650 Paper Presentation
 
Google Kernel Function
Google Kernel FunctionGoogle Kernel Function
Google Kernel Function
 
Class Project Showcase: DNS Spoofing
Class Project Showcase: DNS SpoofingClass Project Showcase: DNS Spoofing
Class Project Showcase: DNS Spoofing
 
Localization in HCI: Yahoo (US vs. China)
Localization in HCI: Yahoo (US vs. China)Localization in HCI: Yahoo (US vs. China)
Localization in HCI: Yahoo (US vs. China)
 

Recently uploaded

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsJoaquim Jorge
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?Antenna Manufacturer Coco
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking MenDelhi Call girls
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc
 
The Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxThe Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxMalak Abu Hammad
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking MenDelhi Call girls
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationSafe Software
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Miguel Araújo
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...Martijn de Jong
 

Recently uploaded (20)

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
The Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxThe Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptx
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...
 

Augmenting mobile 3 g using wifi

  • 1. Augmen'ng  Mobile  3G  Using  WiFi By:  Aruna  Balasubramanian,  Ratul  Mahajan,  Arun   Venkataramani Presenter:  Ke  Huang Tuesday, April 12, 2011 1
  • 2. Demand  for  mobile  access  growing                        www.totaltele.com h+p://www.readwriteweb.com 2 Tuesday, April 12, 2011 2
  • 3. Demand  for  mobile  access  growing                        www.totaltele.com h+p://www.readwriteweb.com 900  million  mobile  broadband  subscrip'ons  today….                                                                                                                                                         www.3gamericas.org                                                       2 Tuesday, April 12, 2011 2
  • 4. Mobile  demand  is  projected  to  far   3 Tuesday, April 12, 2011 3
  • 5. Mobile  demand  is  projected  to  far   www.rysavy.com Current  spectrum 409.5  MHz Unallocated  spectrum   230  MHz (including  whitespaces) Projected  demand  by   800  MHz  –   2016   1000  MHz 3 Tuesday, April 12, 2011 3
  • 6. Mobile  demand  is  projected  to  far   www.nyCmes.com www.rysavy.com Current  spectrum 409.5  MHz Unallocated  spectrum   230  MHz (including  whitespaces) Projected  demand  by   800  MHz  –   2016   1000  MHz 3 Tuesday, April 12, 2011 3
  • 7. Mobile  demand  is  projected  to  far   www.nyCmes.com www.rysavy.com Current  spectrum 409.5  MHz Unallocated  spectrum   230  MHz (including  whitespaces) Projected  demand  by   800  MHz  –   2016   1000  MHz www.nyCmes.com “In  light  of  the  limited  natural  resource  of  spectrum,  we  have  to  look  at   the  ways  of  conserving  spectrum”  -­‐-­‐  Mark  Siegel  (AT&T) Reducing  cellular  spectrum  u'liza'on  is  key! 3 Tuesday, April 12, 2011 3
  • 8. How  can  we  reduce  spectrum  usage? blogs.chron.com 1.  Behavioral 2.  Economic 3.  Technical 4 Tuesday, April 12, 2011 4
  • 9. How  can  we  reduce  spectrum  usage? blogs.chron.com 1.  Behavioral www.usatoday.com 2.  Economic 3.  Technical 4 Tuesday, April 12, 2011 4
  • 10. How  can  we  reduce  spectrum  usage? blogs.chron.com 1.  Behavioral www.usatoday.com 2.  Economic 3.  Technical 4 Tuesday, April 12, 2011 4
  • 11. Augmen'ng  Mobile  3G  using  WiFi                            Offload  data  to  WiFi  when  possible                              Focus  on  vehicular  mobility 5 Tuesday, April 12, 2011 5
  • 12. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 13. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 14. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 15. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 16. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 17. Offloading  3G  data  to  WiFi 6 Tuesday, April 12, 2011 6
  • 18. Related  work  on  mul'ple  interfaces Improving  performance  using  handoffs  based  on   current  condi'ons Reducing  power  consump'on  by  switching   across  mul'ple  interfaces   7 Tuesday, April 12, 2011 7
  • 19. Related  work  on  mul'ple  interfaces Improving  performance  using  handoffs  based  on   current  condi'ons Reducing  power  consump'on  by  switching   across  mul'ple  interfaces   This  work: 1.How  much  3G  data  can  be  offloaded  to  WiFi? 2.How  to  offload  without  hur'ng  applica'ons? 7 Tuesday, April 12, 2011 7
  • 20. Contribu'ons Measurement:    Joint  study  of  3G  and  WiFi   connec'vity Across  three  ci'es:  Amherst,  Seagle,  SFO System:  Wiffler,  to  offload  3G  data  to  WiFi  while   respec'ng  applica'on  constraints   Deployed  on  20  vehicles 8 Tuesday, April 12, 2011 8
  • 21. Measurement  setup Testbed:  Vehicles  with  3G  and  WiFi  (802.11b)  radios Amherst:  20  buses  +  1  car,  Seagle:  1  car,  SFO:  1  car Soiware:  Simultaneously  probes  3G  and  WiFi  for   Availability,  loss  rate,  throughput Dura'on:  3000+  hours  of  data  over  12+  days 9 Tuesday, April 12, 2011 9
  • 22. Open  WiFi  availability  low,  but  useful Availability    =  frac'on  of  1-­‐second  intervals  when  at  least   one  packet  received                                             Availability (%) 10 Tuesday, April 12, 2011 10
  • 23. Open  WiFi  availability  low,  but  useful Availability    =  frac'on  of  1-­‐second  intervals  when  at  least   one  packet  received                                             86% Availability (%) 10 Tuesday, April 12, 2011 10
  • 24. Open  WiFi  availability  low,  but  useful Availability    =  frac'on  of  1-­‐second  intervals  when  at  least   one  packet  received                                             86% Availability (%) 11% 10 Tuesday, April 12, 2011 10
  • 25. Open  WiFi  availability  low,  but  useful Availability    =  frac'on  of  1-­‐second  intervals  when  at  least   one  packet  received                                             86% Availability (%) 11% 7% 10 Tuesday, April 12, 2011 10
  • 26. Open  WiFi  availability  low,  but  useful Availability    =  frac'on  of  1-­‐second  intervals  when  at  least   one  packet  received                                             86% Availability (%) 3G+WiFi  combinaCon  be+er  than  sum  pf  parts 11% 7% 10 Tuesday, April 12, 2011 10
  • 27. WiFi  loss  rate  is  higher Loss  rate  =  Frac'on  of  packets  lost  at  10  probes/sec Cumulative fraction WiFi 3G 11 Tuesday, April 12, 2011 11
  • 28. WiFi  loss  rate  is  higher Loss  rate  =  Frac'on  of  packets  lost  at  10  probes/sec Cumulative fraction 28%   WiFi 3G 11 Tuesday, April 12, 2011 11
  • 29. WiFi  loss  rate  is  higher Loss  rate  =  Frac'on  of  packets  lost  at  10  probes/sec Cumulative fraction 28%   WiFi 8%   3G 11 Tuesday, April 12, 2011 11
  • 30. WiFi  (802.11b)  throughput  is  lower Throughput  =  Total  data  received  per  second WiFi Cumulative fraction 3G Upstream WiFi Cumulative fraction 3G Downstream 12 Tuesday, April 12, 2011 12
  • 31. WiFi  (802.11b)  throughput  is  lower Throughput  =  Total  data  received  per  second WiFi Cumulative fraction 3G Upstream WiFi Cumulative fraction 3G Downstream 12 Tuesday, April 12, 2011 12
  • 32. WiFi  (802.11b)  throughput  is  lower Throughput  =  Total  data  received  per  second WiFi Cumulative fraction 3G Upstream 0.35 WiFi Cumulative fraction 3G Downstream 12 Tuesday, April 12, 2011 12
  • 33. WiFi  (802.11b)  throughput  is  lower Throughput  =  Total  data  received  per  second WiFi Cumulative fraction 3G Upstream 0.35 0.72 WiFi Cumulative fraction 3G Downstream 12 Tuesday, April 12, 2011 12
  • 34. WiFi  (802.11b)  throughput  is  lower Throughput  =  Total  data  received  per  second WiFi Cumulative fraction 3G Upstream 0.35 0.72 WiFi Cumulative fraction 3G Downstream 12 Tuesday, April 12, 2011 12
  • 35. Implica'ons  of  measurement  study Strawman  augmenta'on:  Use  WiFi  when   available Can  offload  only  ~11%  of  the  'me Can  hurt  applica'ons  because  of  WiFi’s  higher  loss   rate  and  lower  throughput 13 Tuesday, April 12, 2011 13
  • 36. Key  ideas  in  Wiffler Increase  savings  for  delay-­‐ Reduce  damage  for  delay-­‐ tolerant  applica'ons sensi've  applica'ons Problem:  Using  WiFi  only   Problem:  Using  WiFi   when  available  saves   whenever  available  can   ligle  3G  usage hurt  applica'on  quality Solu'on:  Exploit  delay-­‐ Solu'on:  Fast  switch  to  3G   tolerance  to  wait  to   when  WiFi  delays   offload  to  WiFi  when   availability  predicted exceed  threshold 14 Tuesday, April 12, 2011 14
  • 37. Predic'on-­‐based  offloading D  =  Delay-­‐tolerance  threshold  (seconds) S  =  Data  remaining  to  be  sent  (bytes) Each  second, 1. If  (WiFi  available),  send  data  on  WiFi   2. Else  if  (W(D)  <  S),  send  data  on  3G 3. Else  wait  for  WiFi. 15 Tuesday, April 12, 2011 15
  • 38. Predic'on-­‐based  offloading D  =  Delay-­‐tolerance  threshold  (seconds) S  =  Data  remaining  to  be  sent  (bytes) Each  second, 1. If  (WiFi  available),  send  data  on  WiFi   2. Else  if  (W(D)  <  S),  send  data  on  3G Predicted  WiFi   transfer  size  in   3. Else  wait  for  WiFi. next  D  seconds   15 Tuesday, April 12, 2011 15
  • 39. Predic'ng  WiFi  capacity History-­‐based  predic'on  of  #  of  APs  using  last  few   AP  encounters   WiFi  capacity  =  (expected  #APs)  x  (capacity  per  AP) Simple  predictor  yields  low  error  both  in  Amherst   and  Seagle 16 Tuesday, April 12, 2011 16
  • 40. Predic'ng  WiFi  capacity History-­‐based  predic'on  of  #  of  APs  using  last  few   AP  encounters   WiFi  capacity  =  (expected  #APs)  x  (capacity  per  AP) Simple  predictor  yields  low  error  both  in  Amherst   and  Seagle Negligible  benefits  with  more  sophis'cated  predic'on,  eg   future  loca'on  predic'on  +  AP  loca'on  database 16 Tuesday, April 12, 2011 16
  • 41. Fast  switching  to  3G Problem: WiFi  losses  bursty  =>  high  retransmission  delay Approach: If  no  WiFi  link-­‐layer  ACK  within  50ms,  switch  to  3G Else,  con'nue  sending  on  WiFi 17 Tuesday, April 12, 2011 17
  • 42. Wiffler  implementa'on Wiffler   proxy § Predic'on-­‐based  offloading  upstream  +  downstream §  Fast  switching  only  upstream Ø Implemented  using  signal-­‐upon-­‐ACK  in  driver 18 Tuesday, April 12, 2011 18
  • 43. Evalua'on  Roadmap Predic'on-­‐based  offloading Deployment  on  20  DieselNet  buses  in  150  sq.  mi   region  around  Amherst Trace-­‐driven  evalua'on  using  throughput  data Fast  switching Deployment  on  1  car  in  Amherst  town  center Trace-­‐driven  evalua'on  using  measured  loss/delay   trace  using  VoIP-­‐like  probe  traffic 19 Tuesday, April 12, 2011 19
  • 44. Deployment  results Data  offloaded  to  WiFi Wiffler’s  predic'on-­‐based  offloading 30% WiFi  when  available 10% File  transfer  size:  5MB;  Delay  tolerance:  60  secs;     Inter-­‐transfer  gap:  random  with  mean  100  secs 20 Tuesday, April 12, 2011 20
  • 45. Deployment  results Data  offloaded  to  WiFi Wiffler’s  predic'on-­‐based  offloading 30% WiFi  when  available 10% File  transfer  size:  5MB;  Delay  tolerance:  60  secs;     Inter-­‐transfer  gap:  random  with  mean  100  secs %  Cme  good  voice  quality   Wiffler’s  fast  switching 68% WiFi  when  available  (no  switching) 42% VoIP-­‐like  traffic:  20-­‐byte  packet  every  20  ms   20 Tuesday, April 12, 2011 20
  • 46. Trace-­‐driven  evalua'on Parameters  varied Workload,  AP  density,  delay-­‐tolerance,  switching  threshold Strategies  compared  to  predic'on-­‐based  offloading: WiFi  when  available Adapted-­‐Breadcrumbs:  Future  loca'on  predic'on  +  AP  loca'on   database Oracle  (Imprac'cal):  Perfect  predic'on  w/  future  knowledge 21 Tuesday, April 12, 2011 21
  • 47. Wiffler  increases  data  offloaded  to  WiFi Workload:  Web  traces  obtained  from  commuters   22 Tuesday, April 12, 2011 22
  • 48. Wiffler  increases  data  offloaded  to  WiFi Workload:  Web  traces  obtained  from  commuters   14% WiFi  when   available  yields   ligle  savings 22 Tuesday, April 12, 2011 22
  • 49. Wiffler  increases  data  offloaded  to  WiFi Workload:  Web  traces  obtained  from  commuters   Wiffler  close  to   42% Oracle 14% 22 Tuesday, April 12, 2011 22
  • 50. Wiffler  increases  data  offloaded  to  WiFi Workload:  Web  traces  obtained  from  commuters   Wiffler  close  to   42% Oracle Sophis'cated   14% predic'on  yields   negligible  benefit 22 Tuesday, April 12, 2011 22
  • 51. Wiffler  increases  data  offloaded  to  WiFi Workload:  Web  traces  obtained  from  commuters   Wiffler  close  to   42% Oracle Sophis'cated   14% predic'on  yields   negligible  benefit Wiffler  increases  delay  by  10  seconds  over  Oracle.   22 Tuesday, April 12, 2011 22
  • 52. Even  more  savings  in  urban  centers 23 Tuesday, April 12, 2011 23
  • 53. Fast  switching  improves  quality  of   delay-­‐sensi've  applica'ons 24 Tuesday, April 12, 2011 24
  • 54. Fast  switching  improves  quality  of   delay-­‐sensi've  applica'ons 58% 24 Tuesday, April 12, 2011 24
  • 55. Fast  switching  improves  quality  of   delay-­‐sensi've  applica'ons 58% 40% 24 Tuesday, April 12, 2011 24
  • 56. Fast  switching  improves  quality  of   delay-­‐sensi've  applica'ons 73% 58% 40% 24 Tuesday, April 12, 2011 24
  • 57. Fast  switching  improves  quality  of   delay-­‐sensi've  applica'ons 73% 58% 40% 30%  data  offloaded  to  WiFi  with  40ms  switching  threshold 24 Tuesday, April 12, 2011 24
  • 58. Future  work Reduce  energy  to  search  for  usable  WiFi Improve  performance/usage  by  predic'ng  user   accesses  to  prefetch  over  WiFi Incorporate  evolving  metrics  of  cost  for  3G  and   WiFi  usage 25 Tuesday, April 12, 2011 25
  • 59. Summary Augmen'ng  3G  with  WiFi  can  reduce  pressure  on   cellular  spectrum Measurement  in  3  ci'es  confirms  WiFi  availability  and   performance  poorer,  but  poten'ally  useful Wiffler:  Predic'on-­‐based  offloading  and  fast  switching   to  offload  without  hur'ng  applica'ons 26 Tuesday, April 12, 2011 26
  • 60. Summary Augmen'ng  3G  with  WiFi  can  reduce  pressure  on   cellular  spectrum Measurement  in  3  ci'es  confirms  WiFi  availability  and   performance  poorer,  but  poten'ally  useful Wiffler:  Predic'on-­‐based  offloading  and  fast  switching   to  offload  without  hur'ng  applica'ons Questions? 26 Tuesday, April 12, 2011 26