24
ARUBA 802.11AC LAB TESTING PROJECT REPORT May 3, 2014 Product Category: Wireless Access Point Vendor Tested: Aruba: Products Tested: Aruba Networks AP225 ABSTRACT This is a report to understand the parameters to implement a wireless network in an enterprise. A performed set of experiments is used to explain few important parameters and an attempt is made to understand the feasibility of a network. Aniruddha Nandam Bhavik N. Patel Manish Sadhwani IST 648 Enterprise Wireless Network

IST 648 Aruba Wireless Project Report - Manish · implement!a!wireless!network!in!an!enterprise.!A! ... network,!a!quantifiedset!of!readings!canhelpunderstandthe!placement ... includedthechannel!width!and!upstream

  • Upload
    haxuyen

  • View
    217

  • Download
    1

Embed Size (px)

Citation preview

   

 

ARUBA  802.11AC    LAB  TESTING  PROJECT  REPORT  

May  3,  2014  

 

Product  Category:  Wireless  Access  Point      

Vendor  Tested:  Aruba:    

 Products  Tested:  Aruba  Networks  AP-­‐225    

 

ABSTRACT  This  is  a  report  to  understand  the  parameters  to  implement  a  wireless  network  in  an  enterprise.  A  performed  set  of  experiments  is  used  to  explain  few  important  parameters  and  an  attempt  is  made  to  understand  the  feasibility  of  a  network.  

   Aniruddha  Nandam  Bhavik  N.  Patel  Manish  Sadhwani    IST  648  Enterprise  Wireless  Network  

 

EXECUTIVE  SUMMARY:    

Most  of  the  enterprises  today  are  implementing  wireless  networks  for  their  operation  or  are  in  the  phase  of  implementing  such.  While  wireless  networks  bring  about  mobility  in  such  deployment  situations,  they  are  also  not  as  efficient  in  terms  of  throughput  as  compared  to  wired  networks.  However,  considering  the  present  technology  it  might  not  be  huge  of  a  problem  depending  on  the  method  in  which  it  is  being  implemented.  This  is  when  the  complexity  of  implementing  wireless  networks  enters.  While  wireless  networks  seem  to  be  a  transparent  ease  of  use  implementation  in  terms  of  users,  it  is  much  harder  for  network  administrators  to  implement  and  monitor  such  networks.  It  is  further  difficult  to  understand  their  ideal  configuration  or  best-­‐suited  requirement  configuration.  This  paper  targets  this  issue  in  not  all  but  few  of  the  most  important  areas  of  implementation.    

While  implementing  wireless  in  an  enterprise  network  it  is  important  to  understand  the  feasibility  of  the  network  and  to  understand  the  specifications,  which  might  be  the  best  to  implement  the  network  considering  the  requirement  of  the  enterprise.  This  includes  the  channel  width  to  be  implemented,  primary  channel  to  implement  the  network,  other  specifications  to  be  decided  considering  the  vendor  whose  products  are  to  be  implemented.  More  important  is  to  know  the  range  of  the  AP  to  understand  the  coverage  of  the  network,  effect  of  the  type  and  number  of  clients,  which  are  accessing  the  network.  Finally  it  is  important  to  understand  the  performance  of  the  network  and  monitor  it  in  real  time.  In  this  project  we  have  tried  to  understand  the  effect  of  these  parameters  using  various  tests  and  documented  these  results  in  the  form  of  graphs  for  easy  visual  realization.    

While  this  report  consists  of  major  issues  to  be  considered  while  implementing  wireless  networks,  only  the  configuration  related  to  the  test  scenario  is  modified  while  maintaining  other  specifications  as  default.  This  might  not  be  perfect  implementation  of  an  enterprise  network  and  would  vary  from  situation  to  situation  depending  on  the  requirements  of  the  network.  

 

Test  Objectives:    

In  this  report  we  have  included  the  results  on  tests  performed  to  understand  the  effect  of  few  parameters  related  to  implementation  of  enterprise  wireless  network.    

In  the  experiments  that  we  performed  we  considered  the  following  specifics:  

• Rate  Verses  Range:  It  is  important  to  understand  the  range  of  the  AP  such  that  it  would  provide  the  throughput  requirements  of  the  network  users.  While  this  explains  the  range  of  the  network,  a  quantified  set  of  readings  can  help  understand  the  placement  of  the  AP.  The  set  of  experiments  we  performed  also  tried  to  understand  the  effect  of  intrusion  of  different  type  of  devices  and  trying  to  explain  the  effect  of  throughput  received  on  one  client  while  increasing  the  number  and  type  of  client.      

• Multi-­‐client  testing:  This  set  of  testing  was  performed  to  understand  the  situation  in  which  the  increasing  number  of  clients  would  behave  in  an  environment.  This  would  be  an  ideal  situation  in  which  one  of  the  clients  is  associated  with  and  AP,  However  as  newer  clients  are  added  to  the  network  the  throughput  would  be  affected  due  to  sharing  of  radio  resources  and  effect  of  channel  share  time  on  these  devices  leading  to  change  in  throughput  received  by  each  client.    

• Aruba  Adaptive  Radio  Management:  Aruba  provides  a  smart  radio  management  and  intelligence,  which  is  used  to  monitor  and  manage  radio  transmission  parameters.  An  attempt  to  understand  this  feature  is  done  in  this  report.  

 

While  the  above  set  of  core  parameters  are  tried  to  be  understood  in  the  project,  we  performed  a  basic  comparative  set  of  tests  to  decide  the  specifics  of  configuring  the  AP  to  perform  the  above  tests.  This  included  the  channel  width  and  upstream/downstream  behavior.    

TESTING  TOOLS:    

Hardware  Used:  

Aruba  AP  225:  

• Has  maximum  data  rate  of  1.3  Gbps  in  the  5  GHz  band  &  600  Mbps  in  the  2.4  GHz  band  • Eliminates  sticky  clients  with  the  use  of  Client  Match  Technology    • Can  function  on  802.11af  i.e.  POE.  

 

Aruba  3600  controller:  

• Can  support  up  to  128  APs  or  Mobility  access  switches  • Manages  authentication,  encryption  and  VPN  connections  and  Layer  3  networking  • Contains  Aruba  Policy  Enforcement  Firewall  (PEF),  Aruba  Adaptive  Radio  Management  

(ARM),  and  Aruba  RF  Protect™  Spectrum  Analysis  • Contains  Airwave  that  provides  real-­‐time  monitoring,  reporting  and  troubleshooting.    

 

Meraki  MS22P  Switch:  

• Contains  24  Gigabit  Ethernet  ports.  • Supports  POE+  i.e.  80.11at  • Supports  voice  and  video  QOS  • Real-­‐time  troubleshooting  

 

Cisco  3760G  Switch:  

• Password  protected  for  secure  access  • Supports  POE  and  POE+  • Supports  Auto-­‐QOS  for  easy  QOS  implementation  • Supports  Cisco  Energy  wise  for  energy  management  of  POE  

 

Dell  PowerEdge  R200:  

• Provides  flexibility  in  performance  depending  on  the  load  • The  server  provides  scalability  and  can  be  easily  upgraded    • Contains  management  tools  for  troubleshooting    

Software  Used:  

iPerf/jPerf  software  was  used  for  obtaining  the  throughput  for  single  client  range  vs.  rate  tests  as  well  as  for  multi  client  tests.  iPerf  is  a  tool  to  measure  network  performance  for  measuring  bandwidth  provided  by  the  network.  iPerf  is  a  tool  to  measure  maximum  TCP  bandwidth,  allowing  the  tuning  of  various  parameters  and  UDP  characteristics.    

 

 

The  screenshot  above  shows  the  jPerf  client  setup  for  executing  the  test  scenarios.  As  seen  in  the  screenshot,  on  the  client  side,  the  IP  address  of  server  was  entered  keeping  the  port  no  5001  and  number  of  parallel  streams  5  or  6  as  appropriate  for  that  specific  test  scenario.  Each  test  run  was  conducted  for  30  seconds  and  the  output  was  set  to  be  displayed  in  Mbps.  

For  measuring  the  signal  levels  at  different  locations  on  the  floor  and  to  identify  the  number  of  channels  for  free  transmission,  we  used  spectrum  analysis  tools  called  Chanalyzer  Pro  and  InSSIDer  installed  on  a  Dell  laptop.  From  the  spectrum  analysis,  RSSI  was  observed  at  various  locations  and  finally  seven  locations  were  selected  for  testing.  Also,  it  was  decided  to  run  all  the  tests  with  the  AP  transmitting  on  the  channel  149  +  151.  

   

TESTING  ENVIRONMENT  AND  CONFIGURATION:    

Network  Diagram:  

 

 

 

The  figure  above  shows  the  network  setup  for  testing  that  comprised  of  an  Aruba  Wireless  LAN  Controller,  Layer  3  Cisco  switch,  DHCP  server,  Aruba  Access  point  and  single/multiple  clients.  Aruba  AP  225  can  be  exploited  completely  on  its  full  power  mode  only  via  POE+  cable.  Hence,  another  Meraki  switch  was  also  introduced  in  the  network  to  provide  POE+  support  to  the  AP.  We  used  an  HP  ENVY  m7-­‐j010dx  Notebook  with  10/100/1000  Gigabit  Ethernet  LAN  (RJ-­‐45  connector)  as  the  server  for  testing  with  jPerf.  A  jPerf/iPerf  server  was  created  on  this  machine  connected  to  the  Aruba  Wireless  3600  series  controller  using  a  CAT  6e  cable  via  Meraki  MS22P  Switch  provided  with  DHCP  services  from  Dell  PowerEdge  R200  acting  as  DHCP  server.  The  wireless  clients  were  connected  to  the  Aruba  AP  225,  also  connected  to  the  Aruba  Wireless  

3600  series  controller  using  similar  arrangement  as  HP  laptop  obtaining  IP  address  from  DHCP  server  thereby  getting  connected  to  the  controller.  

 

Basic  Testing  for  Test  Configurations:  

In  order  to  implement  a  Wireless  network  it  is  important  to  decide  the  parameters  of  implementing  the  wireless  network.  This  includes  the  width  of  channel  on  which  the  wireless  network  is  being  broadcasted.  This  can  be  20  MHz,  40  MHz  or  80  MHz.  As  also  we  need  to  understand  the  traffic  flow  requirements  of  the  network,  which  can  be  upstream/downstream.  In  a  real  life  situation  the  amount  of  downstream  traffic  is  generally  higher  than  the  upstream  traffic  and  it  would  be  important  to  know  the  throughput,  which  can  be  delivered  in  either  case  or  a  case  in  which  there  is  upstream  along  with  downstream  traffic.  Hence,  a  small  set  of  tests  was  performed  to  understand  the  effect  of  these  parameters  on  the  network  to  be  setup.  

 20/40/80  MHz  Comparison:    

In  order  understand  the  effect  of  channel  width  we  configured  the  access  point  to  transmit  in  the  default  specifications  of  channel  149  for  a  20  MHz  channel  with  the  other  options  set  to  default.  This  included  selection  of  options  for  High  throughput  enable  (radio)    and  Very  high  throughput  enable  (radio)  mode.  Then  a  throughput  test  was  performed  to  observe  the  performance  of  network  in  downstream  mode.  We  received  the  following  set  of  readings  for  the  same.  

Channel  width   20  MHz   40  MHz   80  MHz  Reading  1   368   219   427  Reading  2   366   236   311  Reading  3   274   231   325  Average   335   230   355  

 

The  above  set  of  readings  were  received  at  a  distance  of  20  feet,  where  we  received  a  signal  strength  of  -­‐62  dBm  and  the  transmit  rate  received  was  527  Mbps  for  all  of  the  above  cases.    

As  you  can  see  from  the  above  set  of  readings  we  can  see  that  the  effect  of  channel  width  is  not  as  expected  from  theory.  This  can  be  reasoned  to  the  default  selection  of  Aruba  Radio  Management  (ARM)  and  selection  of  Very  high  throughput  enable  (radio)  mode.  This  might  be  responsible  to  override  the  channel  width  setting  for  providing  throughput  according  to  the  nature  of  the  client.  The  Aruba  Web  interface  displays  dashboard,  which  displays  the  clients,  connected  to  the  network  and  the  capability  of  it.  As  an  example  the  MacBook  pro  client  to  the  

network  was  recognized  as  802.11ac  device  of  80  MHz  channel  capacity.  Hence,  we  observe  a  similar  set  of  readings  for  either  of  the  selection  of  channel  width  of  20/40/80  MHz.  

 

Upstream/Downstream  Comparison:  

 

It  is  also  notable  to  observe  the  type  of  traffic  and  corresponding  effect  on  the  performance  of  the  network.  In  order  to  perform  this  we  connected  a  HP  Laptop  via  Ethernet  to  serve  as  a  JPerf/iPerf  client  at  first  to  transmit  data  to  server,  which  gets  connected  to  the  network  via  Wireless  connectivity,  this  simulates  a  typical  downstream  flow  of  traffic  on  the  network.  This  arrangement  was  then  modified  a  bit  to  observe  upstream  traffic.  This  was  done  by  keeping  the  device  connectivity  as  the  same,  However  changing  the  JPerf  setting  on  HP  Laptop  to  serve  as  server  and  setting  MacBook  Pro  802.11ac  client  to  client  to  transmit  data  from  wireless  medium  to  the  wired  connected  server.  Finally,  the  parameter  on  the  client  (HP  Laptop-­‐Wired)  is  set  to  –d,  which  corresponds  to  demonstrate  both  way  traffic  and  displays  both  upstream  and  downstream  traffic  behavior.  This  can  be  represented  using  the  table:  

 

Only  Upstream   Only  Downstream  Both  way  traffic  

Upstream   Downstream  219   313   126   228  236   299   153   228  231   348   154   209  

 

The  above  set  of  readings  was  received  at  same  distance  as  above  for  signal  strength  of  -­‐63  dBm  and  a  transmit  rate  received  was  527  Mbps.  The  following  graph  illustrates  a  graphical  inference  that  we  understand  as  explained  below.  

 

Graph  1:  Throughput  Comparison  of  Upstream  and  Downstream  Traffic  

As  you  can  see  from  the  above  readings  we  get  a  higher  downstream  rate  as  expected  also  when  we  have  dedicated  server-­‐client  scenario  and  when  we  have  a  both  way  traffic  flow  between  the  server  and  the  client.  

 

AP  configuration:  

Based  on  the  readings  obtained  in  the  above-­‐performed  tests,  we  configured  the  access  point  as  described  below.  

• Tests  performed  in  the  2.4  GHz  band  used  20  MHz  channels,  while  tests  performed  in  the  5  GHz  band  used  40  MHz  channels.  

• AP  channel  selection  was  static  set  to  channel  number  149  +  151  for  5  GHz  and  channel  number  6  for  2.4  GHz.  

• For  5  GHz,  ‘High  Throughput’  and  ‘Very  High  Throughput’  radios  were  enabled.  

 Note:  All  the  readings  were  taken  in  Mbps.  All  the  values  in  the  graphs  are  represented  in  

Mbps.    

   

0  

50  

100  

150  

200  

250  

300  

350  

400  

Upstream   Downtream  

Only  Upstream   Only  Downstream   Both  way  traffic  

Upstream  vs.  Downstream  

Reading  1   Reading  2   Reading  3  

Floor  Plan:  

 

Based  on  the  RSSI  values  obtained  during  a  site  survey  of  the  second  floor  of  Hinds  Hall,  Syracuse  University,  six  test  locations  were  finalized  as  shown  in  the  figure  above.  One  more  test  location  was  added  other  than  those  described  in  the  figure  above.  This  was  just  below  the  AP  to  minimize  the  interference  effect.  The  description  of  all  the  locations  is  as  follows:  

1. Just  Below  AP  (0  feet)  –  to  observe  throughput  with  minimum  interference.  2. Hallway  door  –  considered  to  be  in  direct  LOS  of  the  AP.  RSSI  value:  -­‐36  dBm  3. Hallway  (near  Room  207)  –  Another  location  in  direct  LOS  but  at  a  farther  distance.  

RSSI  value:  -­‐45  dBm  4. Near  Room  210  –  Not  in  LOS,  but  with  minimum  interference  outside  LOS.  

RSSI  value:  -­‐51  dBm  5. Conference  Room  216  –  Going  farther  away  from  access  point  introducing  more  

interference.  RSSI  value:  -­‐62  dBm  

6. Near  Room  224  –  Interference  more  intense  due  to  intervening  walls.  RSSI  value:  -­‐73  dBm  

7. Near  Room  232  –  basically  to  the  end  of  hallway  providing  maximum  interference  on  the  same  floor.  RSSI  value:  -­‐80  dBm    

RATE  VS.  RANGE  TESTS:  

Test  1  –  Single  Client,  Single  AP  Throughput  Test:  First  we  started  with  basic  testing  by  accessing  Aruba  AP  from  just  one  client  at  a  time.  The  purpose  of  these  single  client  tests  was  not  any  way  related  to  frame  a  real  world  scenario  for  testing  the  AP.  Instead  the  purpose  of  this  testing  was  to  obtain  the  actual  throughput  value  of  the  Aruba  AP  to  form  a  basic  ground  for  our  further  testing.  As  we  all  know,  there  are  some  differences  between  the  data  rates  of  an  AP  and  its  throughput.  In  these  tests  we  had  just  one  client  present  in  AP’s  network  and  trying  to  capitalize  optimally  by  transmitting  all  the  parallel  streams  over  the  complete  channel  available  for  communication.  Rate  vs.  range  tests  would  give  us  a  better  picture  regarding  the  throughput  capabilities  of  the  AP.  We  selected  all  802.11ac  clients  for  these  test  scenarios.  Following  were  the  clients  used  for  single  client  throughput  test:  

-­‐ New  MacBook  Pro  with  Retina  display  that  is  802.11ac  compatible  and  3x3  MIMO  -­‐ Same  MacBook  Pro  with  Asus  Wi-­‐Fi  Adapter  that  is  2x2  MIMO  -­‐ Samsung  Note  3  with  802.11ac  support  and  1x1  MIMO  

We  chose  MacBook  for  these  tests  because  they  support  802.11ac  standard  and  provide  slightly  higher  throughput  when  compared  to  other  Dell/Intel  laptops.  Asus  wireless  adapter  is  a  USB  version  3.0  adapter.  Hence  it  would  be  the  best  option  to  choose  a  device  that  has  USB  version  3.0  for  testing  the  Asus  adapter,  as  it  would  exploit  it  completely  to  provide  appropriate  readings  of  throughput  for  a  single  client  test.  New  MacBook  Pro  satisfied  that  requirement  and  was  apt  for  such  testing.  

We  conducted  these  tests  at  all  the  testing  locations  described  in  the  floor  plan.  The  RSSI  levels  at  each  location  were  measured  each  time  a  test  was  executed  and  accordingly  decision  was  made  on  client’s  jPerf  regarding  the  number  of  parallel  streams  to  be  transmitted  while  running  the  test.  The  number  of  parallel  streams  was  either  5  or  6.  Set  of  three  readings  was  carried  out  on  each  client  at  each  location  and  average  of  these  reading  was  calculated.  This  avoided  any  sudden  changes  in  the  environment  that  may  result  into  abrupt  drop  or  rise  in  the  throughput.  The  readings  were  captured  till  the  time  they  were  stable  and  consistent.  

   

Results:  

 

 

Graph  2:  Throughput  results  for  Rate  vs.  Range  test  on  New  MacBook  Pro  Client  (3x3)  

 

 

Graph  3:  Throughput  results  for  Rate  vs.  Range  test  on  MacBook  Pro  with  Asus  Adapter  (2x2)  

0  

50  

100  

150  

200  

250  

300  

350  

400  

450  

Just  below  AP   Hallway  door   Hallway  (Room  207)  

Near  Room  210   Conference  room  216  

Near  Room  224   Near  Room  232  

New  MBP  (3x3)  

0  

50  

100  

150  

200  

250  

Just  below  AP   Hallway  door   Hallway  (Room  207)  

Near  Room  210   Conference  room  216  

Near  Room  224   Near  Room  232  

MBP  with  ASUS  adapter  (2x2)  

 

Graph  4:  Throughput  results  for  Rate  vs.  Range  test  on  Samsung  Note  3  (1x1)  

 

As  expected  for  all  the  three  clients,  the  maximum  throughput  was  obtained  when  the  clients  were  in  direct  Line  of  Sight  (LOS)  of  the  AP.  The  throughput  obtained  for  New  MBP  (3x3  client)  is  highest  of  all  three  clients  just  below  the  AP  with  a  value  of  408  Mbps,  whereas  maximum  throughput  obtained  for  MBP  with  Asus  adapter  and  Note  3  is  210  Mbps  and  185  Mbps  respectively.  

But,  as  the  distance  between  the  client  and  the  AP  increased,  the  throughput  decreased.  This  drop  in  the  throughput  was  not  only  due  to  the  increase  in  the  distance  between  client  and  AP  but  also  due  to  the  interference  caused  by  the  intervening  walls.  It  can  be  seen  from  the  graphs  of  first  two  clients  (MBP  and  MBP  with  Asus  adapter),  that  there  is  sudden  drop  in  the  throughput  when  the  client  moves  near  the  conference  room  216.  This  drop  in  the  throughput  is  not  so  abrupt  when  the  client  was  Samsung  Note.  Although  Samsung  Note  (1x1  client)  gives  a  lower  throughput  value  even  at  LOS  locations,  there  is  not  much  variation  in  throughput  with  the  variation  in  distance  from  the  AP.  It’s  a  gradual  decrease  in  throughput  when  compared  to  other  two  clients  (3x3  and  2x2).  This  can  be  clearly  observed  in  the  following  graph  that  compares  the  throughput  values  of  all  the  three  clients  obtained  at  different  locations.  

 

0  20  40  60  80  100  120  140  160  180  200  

Just  below  AP   Hallway  door   Hallway  (Room  207)  

Near  Room  210   Conference  room  216  

Near  Room  224   Near  Room  232  

Note3  (1x1)  

 

Graph  5:  Comparison  of  Throughput  results  for  Rate  vs.  Range  test  on  three  802.11ac  clients  

 

Since  the  RSSI  value  was  very  low  at  the  farthest  location  (near  Room  232),  the  throughput  values  obtained  at  this  location  were  very  less  (almost  negligible  for  3x3  and  1x1  clients).  But  still  the  readings  could  be  taken  and  displayed  on  the  graph  as  the  clients  could  find  the  AP  successfully  and  the  signal  was  not  dead.  

     

0  50  100  150  200  250  300  350  400  450  

Just  below  AP   Hallway  door   Hallway  (Room  207)  

Near  Room  210   Conference  room  216  

Near  Room  224   Near  Room  232  

Throughput  Comparison  of  3  clients  

New  MBP  (3x3)   MBP  with  ASUS  adapter  (2x2)   Note3  (1x1)  

Test  2  –  Dual  Client,  Single  AP  Throughput  Test:  

In  this  test  scenario,  we  extended  our  Rate  vs.  Range  test  from  a  single  client  to  dual  client.  Here,  two  clients  were  simultaneously  accessing  the  AP  and  jPerf  test  was  run  on  both  the  clients  together.  The  clients  used  for  this  test  case  were:  

-­‐ New  MacBook  Pro  with  Retina  display  that  is  802.11ac  compatible  and  3x3  MIMO  -­‐ MacBook  Pro  with  Asus  Wi-­‐Fi  Adapter  that  is  2x2  MIMO  

For  Asus  adapter,  MacBook  Pro  was  used  because  it  has  USB  version  3.0  support  that  has  the  optimum  compatibility  with  Asus  adapter.  Both  the  clients  were  given  the  same  server  address  on  jPerf  accessing  the  same  AP.  Test  run  were  executed  on  both  the  clients  simultaneously  for  the  same  period  of  time.  Similar  to  single  client  testing,  set  of  multiple  readings  was  taken  on  both  the  clients  to  obtain  a  consistent  result  and  average  of  all  these  reading  was  calculated  and  plotted  on  a  graph.  The  readings  were  taken  at  all  the  locations  mentioned  in  the  floor  plan  as  done  in  single  client  testing  except  the  last  location  –  near  Room  232.  The  signal  level  at  this  location  dropped  to  a  very  low  level  and  the  clients  were  unable  to  reach  the  AP.  

 

Graph  6:  Throughput  results  for  Rate  vs.  Range  test  –  Dual  Client  accessing  simultaneously  

 

From  the  graph,  it  can  be  seen  that  the  throughput  values  have  declined  compare  to  the  values  obtained  for  single  client  testing  using  the  same  clients.  This  is  because  of  channel  sharing  between  both  the  clients  accessing  the  same  AP  at  the  same  time.  As  expected,  the  throughput  obtained  for  a  3x3  MIMO  client  is  higher  than  that  obtained  for  2x2  MIMO  client  at  each  location.  The  difference  between  the  throughput  values  of  both  the  clients  increases  with  the  increase  in  the  distance  from  the  AP.  

   

0  20  40  60  80  100  120  140  160  180  

Just  below  AP   Hallway  door   Hallway  (Room  207)  

Near  Room  210   Conference  room  216  

Near  Room  224  

Dual  Client  Throughput  Test  

MBP  3x3   MBP  with  ASUS  2x2  

Test  3  –  Triple  Client,  Single  AP  Throughput  Test:  

Similar  to  the  tests  performed  previously,  we  further  extended  our  parameters  involved  in  a  rate  vs.  range  test  by  further  increasing  the  number  of  clients  to  three.  In  this  case,  three  802.11ac  clients  will  access  the  AP  simultaneously  and  throughput  values  obtained  for  all  the  three  clients  will  be  observed.  For  this  we  used  the  same  three  clients  used  for  single  client  testing,  the  only  difference  being  that  this  time  all  the  clients  were  running  the  tests  on  their  jPerf  simultaneously  for  the  same  period  of  time.  List  of  clients  used:  

-­‐ New  MacBook  Pro  with  Retina  display  that  is  802.11ac  compatible  and  3x3  MIMO  -­‐ Same  MacBook  Pro  with  Asus  Wi-­‐Fi  Adapter  that  is  2x2  MIMO  -­‐ Samsung  Note  3  with  802.11ac  support  and  1x1  MIMO  

All  the  clients  were  assigned  with  the  same  server  address  on  their  jPerf  and  started  the  test  together.  Similar  to  single  client  testing,  set  of  multiple  readings  was  taken  on  all  the  three  clients  to  obtain  a  consistent  result  and  average  of  all  these  reading  was  calculated  and  plotted  on  a  graph.  The  readings  were  taken  at  all  the  locations  mentioned  in  the  floor  plan  as  done  in  single  client  testing  except  the  last  two  locations  –  near  Room  232  and  near  Room  224.  The  signal  levels  at  these  locations  dropped  to  a  very  low  level  and  the  clients  were  unable  to  reach  the  AP.  

 

Graph  7:  Throughput  results  for  Rate  vs.  Range  test  –  Triple  Client  accessing  simultaneously  

 

As  observed  in  the  graph  above,  throughput  values  have  dropped  considerably  at  the  last  location,  near  Conference  Room  216.  Also,  when  all  the  three  clients  were  accessing  the  AP  simultaneously,  the  throughput  obtained  for  3x3  MIMO  client  is  substantially  higher  than  that  obtained  for  the  remaining  two  clients.    

0  20  40  60  80  100  120  140  160  180  

Just  below  AP   Hallway  door   Hallway  (Room  207)   Near  Room  210   Conference  room  216  

Triple  Client  Throughput  Test  

New  MBP  (3x3)   MBP  with  ASUS  adapter  (2x2)   Note3  (1x1)  

MULTI-­‐CLIENT  TESTING:  

In  this  set  of  tests,  the  aim  was  to  observe  the  behavior  of  the  network  when  the  number  of  clients  or  the  load  was  being  increased.  In  order  to  do  this  a  total  of  16  clients  were  aimed  to  connect  to  the  same  network  using  the  AP  utilized  in  the  first  suit  of  tests.  In  order  to  simulate  a  practical  situation  in  an  enterprise  we  considered  the  following  variation  of  devices  to  be  used  as  clients.  This  included:  

• 802.11ac  Clients  -­‐ Apple  MacBook  pro  2013  (802.11a/g/n/ac  3x3  MIMO)  -­‐ Dell  Laptop  with  Asus  802.11ac  adapter  (2x2  MIMO)  -­‐ Samsung  Note  3(1x1)  

• 802.11n  Clients-­‐  Mac  -­‐ Apple  MacBook  Pro  (802.11a/g/n)  

• 802.11n  Clients  Windows  -­‐ Dell  Latitude  E6400  (802.11a/g/n  3x3  MIMO)  

• 802.11g  Clients  Windows  -­‐ Dell  Latitude  M2400  (802.11g)  

 These  devices  were  all  placed  at  a  same  distance  from  the  access  point  mounted  as  before.  These  clients  were  connected  to  the  AP,  broadcasting  SSID  in  5GHz  spectrum  in  the  40  MHz  channel  of  149+152.  All  of  these  clients  were  placed  at  the  same  distance  of  20  feet  and  were  receiving  signal  strength  of  -­‐50  dBm.  However  not  all  of  these  clients  were  associated  with  the  AP  for  the  complete  duration  of  the  test  and  were  associated  for  the  complete  duration  of  time  after  initiating  its  connection  to  the  AP.  The  clients  were  placed  on  a  table  such  that  the  screens  were  kept  open  for  a  suitable  viewing  angle  (approximately  70º).  They  were  placed  such  that  the  AP  was  facing  the  rear  side  of  the  screen;  in  other  words,  the  client  user  was  facing  the  AP  directly.    

In  order  to  perform  this  test,  we  utilized  a  HP  ENVY  m7-­‐j010dx  Notebook  with  10/100/1000  Gigabit  Ethernet  LAN  (RJ-­‐45  connector)  as  server.  A  JPerf  /iPerf  server  using  TCP  was  created  on  this  machine  connected  to  the  Aruba  AP  225  via  wired  network.  All  the  clients  were  then  getting  connected  to  the  network  using  this  AP  and  were  running  JPerf/  iPerf  in  client  mode  also  using  TCP  to  connect  to  the  HP  JPerf  server  in  order  to  test  throughput.  All  of  these  clients  are  scheduled  for  measuring  throughput  for  the  complete  duration  of  test  once  they  are  getting  connected  to  the  network,  which  happens  one  after  another  in  the  same  order  of  the  list  as  specified  above.  Rest  of  the  settings  on  the  JPerf  utility  is  set  as  defaults  for  implementing  the  test.  Since,  each  of  them  utilizes  one  stream  of  data  to  test  throughput  via  JPerf  utility  and  this  gets  represented  on  the  server  via  a  channel  number  associated  to  it  when  it  gets  connected.  The  HP  Laptop  server  is  responsible  for  measurement  of  throughput  from  all  of  these  clients  and  keeps  a  track  of  the  Throughput  for  each  client.  

This  test  was  performed  to  simulate  the  behavior  in  a  real  world  environment  where  devices  would  get  connected  to  a  network  in  short  intervals  of  time.  The  scenario  we  used  for  this  deployment  was  to  initially  start  one  802.11ac  device  (which  was  an  Apple  MacBook  pro-­‐802.11ac  3x3  device  on  our  case).  Associating  a  Dell  Laptop  with  Asus  802.11ac  adapter,  which  was  a  2x2  MIMO  after,  and  interval  of  2  min  and  connecting  it  as  a  JPerf  client  device  followed  this.  Similarly  after  an  interval  of  2  min  we  connected  another  Laptop  with  Asus  802.11ac  adapter  with  2x2  MIMO  and  Note  3  device.  In  a  similar  fashion  other  legacy  devices  were  connected  to  the  network  and  tested  for  throughput  by  association  in  a  regular  interval  of  around  2  min.  All  of  these  devices  were  initially  not  associated  with  the  AP  and  were  associated  with  the  AP,  followed  by  starting  the  JPerf  client  on  it  until  the  end  of  the  test.  The  results  we  obtained  can  be  summarized  in  the  following  graph:  

 

   

Graph  8:  Throughput  results  Multiple  Clients  accessing  the  AP  simultaneously  

 

We  can  see  that  initially  when  only  the  MacBook  pro,  which  is  a  3x3  device,  was  connected  we  received  a  throughput  of  330  Mbps.  However,  if  we  continue  to  increase  the  number  of  clients  the  throughput  received  reduces  further  and  further.  While  connecting  each  new  device  at  a  regular  interval  of  2  min  we  connected  16  devices  till  the  reading  18  point,  after  which  the  environment  was  left  to  stabilize  for  about  15  min  and  due  to  which  we  then  see  again  a  rise  in  the  throughput  values  at  test  point  19.  While  this  set  of  readings  explain  a  lot  about  the  

0  

50  

100  

150  

200  

250  

300  

350  

400  

450  

1   2   3   4   5   6   7   8   9   10   11   12   13   14   15   16   17   18   19  

Effect  of  MulSple  Clients  over  load  

MacBook  pro   Azuz  Dongle   Azuz  Dongle   Note  3  

Mac  (802.11n)  Avg   DELL  (802.11n)  Avg   DELL  (802.11g)  Avg  

behavior  about  the  Wireless  transmission  on  increased  number  of  clients  it  was  also  worth  noting  that  one  of  the  clients  Asus  Dongle  2x2  which  was  the  second  device  to  enter  the  test  environment  which  lost  connectivity  within  a  short  interval  of  time  and  one  of  the  possible  explanation  for  such  could  be  high  allocation  of  transmission  time  slot  for  one  of  the  client  not  providing  sufficient  resources  for  other  and  thereby  causing  client  failure.  Which  is  also  one  of  the  major  factor  while  deploying  many  clients,  wherein  time  slot  management  becomes  increasingly  difficult  leading  to  drop  of  connection  for  clients  trying  to  get  into  the  network.  If  we  plot  the  readings  of  throughput  obtained  at  the  initiation  and  at  the  end  and  average  of  all  the  readings  that  we  received  during  the  test  performed,  we  can  see  that:  

 

Clients  Throughput  (Mbps)  

At  Start   At  End   Average  

MacBook  Pro  (802.11ac  3x3)   331   96.3 41.91  

Asus  adapter  (802.11ac  2x2)   17   3.73  Samsung  Note  3  (802.11ac)   181   25.7

Apple  MacBook  Pro  (802.11a/g/n)  

24   10.4  

12.6  70.7   14.6 98.3   2.98  53.5   12.8

Dell  Latitude  E6400  (802.11a/g/n  3x3)  

2.62   2.36

2.8675  3.67   3.34  3.41   3.08 3.34   3.28  

Dell  Latitude  M2400  (802.11g)  

2.16   5.66

5.643333  8.77   6.03  7.34   8.44 8.06   5.24  

 

   

From  the  above  set  of  readings  we  can  observe  that  the  throughput  for  each  client  individually  decreases  while  progressing  in  the  test  and  that  increasing  the  number  of  clients  has  substantial  drop  in  the  received  throughput  across  the  network,  this  is  due  to  channel  sharing  which  happens  across  all  the  devices  connected  in  the  network.  The  following  graph  represents  the  aggregate  throughput  that  we  receive  from  all  the  clients  that  are  connected  to  the  network:  

   

Graph  9:  Aggregate  of  Throughput  results  for  Multiple  Clients  Testing  

 

As  we  can  see  that  the  throughput  we  received  while  fewer  clients  are  connected  to  the  network  is  higher  while  as  the  number  of  clients  keep  on  increasing  the  available  throughput  across  the  network  also  drops  substantially.  

   

0  

50  

100  

150  

200  

250  

300  

350  

400  

450  

1   2   3   4   5   6   7   8   9   10   11   12   13   14   15   16   17   18   19  

Aggregate  Throughput  

MacBook  pro   Azuz  Dongle   Azuz  Dongle   Note  3  

Mac  (802.11n)  Avg   DELL  (802.11n)  Avg   DELL  (802.11g)  Avg  

ARUBA  MANAGEMENT  INTERFERENCE:  

Aruba  Adaptive  Radio  Management  (ARM)  technology  optimizes  Wi-­‐Fi  network  behavior  and  automatically  ensures  that  Aruba  access  points  stay  clear  of  RF  interference,  resulting  in  a  more  reliable,  higher-­‐performing  wireless  LAN.  ARM  uses  patented  technology  that  uses  automatic  infrastructure-­‐based  controls  to  manage  the  entire  RF  spectrum.  ARM™  dynamically  adjusts  the  RF  environment  to  maximize  Wi-­‐Fi  stability  and  predictability,  ensuring  optimal  performance  for  all  clients  and  applications.    

Wi-­‐Fi  Challenges:  

There  are  certain  challenges  in  Wi-­‐Fi  environments.  Wi-­‐Fi  being  a  shared  medium,  clients  must  compete  for  bandwidth  and  simultaneously  avoid  collisions.  The  problem  of  co-­‐channel  interference  also  creates  a  challenge  in  providing  good  service.  An  additional  problem  is  that  a  slow  client  can  make  the  whole  WLAN  slow  and  cause  performance  problems  for  every  other  client.    

As  Wi-­‐Fi  works  in  the  unlicensed  band,  there  is  considerable  interference  caused  by  Bluetooth  devices,  cellular  antennas,  microwave  ovens,  wireless  cameras.  This  interference  can  cause  serious  performance  problems.    

Another  problem  is  that  of  sticky  clients.  This  is  caused  by  a  fault  in  the  devices,  which  connect,  to  an  AP  that  is  poor  in  performance  and  stays  connected  even  though  there  are  other  APs  available.  This  causes  overload  in  APs,  which  in  turn  reduces  the  network  and  client  performance.  

As  one  of  the  groups  worked  on  Wi-­‐Fi  Multimedia  (WMM)  it  is  important  to  address  the  challenges  associated  with  it  also.  WMM  provides  QOS  features  that  help  in  voice,  video,  and  other  latency-­‐sensitive  traffic  get  priority  treatment  over  regular  data.  To  provide  the  appropriate  QOS  and  improve  user  experience  it  is  necessary  to  address  this.    

Features  of  ARM:  

1. Boost  client  performance:  ARM  works  with  Aruba  ClientMatch  technology  to  maximize  client  performance.  ClientMatch  continuously  gathers  session  performance  metrics  from  devices  and  uses  this  information  to  steer  each  one  to  the  best  radio  on  the  best  AP.  ClientMatch  dynamically  optimizes  Wi-­‐Fi  client  performance  as  users  roam  and  RF  conditions  change.  If  a  device  moves  out  of  range  of  an  AP  or  RF  interference  unexpectedly  impedes  performance,  ClientMatch  steers  it  to  a  better  AP.  ARM  and  ClientMatch  technologies  are  critical  to  ensuring  that  overall  network  capacity  and  performance  remains  consistent.  

   

 2. App  visibility  and  performance:  

Aruba  AppRF  technology  which  is  included  in  the  ARM,  h  leverages  information  from  Aruba  Mobility  Controllers  to  identify  a  wide  range  of  business-­‐critical  enterprise  applications  and  apply  the  appropriate  QOS  tags.  ARM  and  AppRF  ensure  the  web  and  mobile  applications  that  are  used  in  the  enterprise  get  the  end-­‐to-­‐end  QOS  tags  that  are  needed.  It  also  gives  the  IT  department  visibility  and  control  over  the  traffic  used  in  the  voice  and  video  traffic.  ARM  is  the  only  RF  management  technology  certified  to  optimize  latency-­‐sensitive  applications.  It  dynamically  adapts  RF  scanning  and  adjusts  RF  bandwidth  available  in  the  presence  of  mission-­‐critical  apps.  Optimizes  voice  and  video  application  performance  and  improves  the  user  experience.    

3. Ensure  airtime  fairness:  Wi-­‐Fi  being  a  shared  medium  we  need  to  ensure  each  client  must  be  allotted  fair  amount  of  time  on  the  air.  ARM  maximizes  client  performance  by  giving  each  client  fair  access  and  ensuring  that  no  single  client  or  group  of  clients  monopolizes  resources  at  the  expense  of  others.  This  does  not  let  one  device  to  boss  the  entire  network.      

4. Reduce  co-­‐channel  interference:  Co-­‐channel  interference  occurs  when  there  are  multiple  devices  trying  to  connect  to  the  same  channel  simultaneously.  This  reduces  performance  of  the  channel  and  also  reduces  throughput  for  all  the  devices  connected  to  that  channel.  To  tackle  this  problem,  ARM  adjusts  channels  and  transmit-­‐power  based  on  the  changing  the  RF  environment.  For  example,  if  an  AP  goes  down,  ARM  automatically  adjusts  the  transmit  power  of  surrounding  APs  accordingly  to  fill  coverage  holes.  ARM  also  coordinates  access  to  a  single  channel,  which  allows  neighboring  APs  to  share  the  RF  spectrum  without  increasing  co-­‐channel  interference.  Overcomes  the  challenges  of  dense  AP  deployments  in  the  2.4-­‐GHz  band.      

5. Immune  to  non-­‐Wi-­‐Fi  interference:  ARM  ensures  that  network  access  and  data  throughput  is  maintained,  even  in  the  presence  of  significant  interference  from  non-­‐Wi-­‐Fi  sources.  ARM  automatically  adjusts  the  affected  APs  using  a  variety  of  techniques.  This  helps  to  ensure  that  the  network  continues  to  perform  optimally.  The  techniques  used  are  shrinking  the  coverage  area,  raising  the  noise  floor,  and  throttling  back  the  maximum  throughput.  Aruba  APs  also  perform  spectrum  analysis.  This  enables  APs  to  identify  sources  of  non-­‐Wi-­‐Fi  interference  like  video  bridges,  microwave  ovens  or  Bluetooth  devices.    

6. Optimize  spectrum  usage:  The  number  of  channels  in  the  5-­‐GHz  band  is  limited  and  varies  based  on  regulatory  domain,  PHY  and  RF  coverage.  Hence  the  network  must  adapt  to  these  conditions  and  

use  all  active  channels  to  maximize  spectrum  availability.  ARM  uses  dynamic  channel  selection  to  automatically  assign  channel  and  power  settings  for  all  APs  in  the  network.  This  method  ensures  that  APs  operate  over  the  healthiest  or  least  congested  channels  and  the  WLAN  making  efficient  use  of  the  available  spectrum.      

7. Channel  scanning  intervals:  APs  can  go  off-­‐channel  at  fixed  intervals  to  scan  other  channels  for  noise  and  rogue  devices.  This  scanning  helps  APs  choose  a  channel  that  maximizes  their  performance  and  is  essential  to  detect  unauthorized  or  malicious  devices.  ARM  dynamically  adjusts  its  channel  scanning  based  on  various  parameters:  

• AP  load:  ARM  tracks  the  traffic  load  on  each  AP  and  dynamically  adjusts  the  scanning  frequency  accordingly.  A  lightly  loaded  AP  will  scan  frequently,  such  as  once  a  second,  while  a  heavily  loaded  AP  will  scan  less  often.  

• Traffic   type:  ARM  leverages  information  from  Aruba  Mobility  Controllers  to  identify  voice  and  video  traffic  traversing  an  AP.  When  these  flows  are  detected,  the  AP  stops  scanning  to  avert  latency.  As  a  result,  voice  and  video  and  the  user  experience  are  optimized.    

8. Update  neighboring  APs:  When  an  AP  pauses  scanning  to  accommodate  application  traffic,  that  AP  still  receives  updates  over  the  air  from  neighboring  APs  with  information  they’ve  collected  about  the  RF  environment.  Receiving  these  updates  over  the  air  enables  the  AP  that  has  paused  scanning  to  determine  the  signal  strength  and  transmit  power  of  the  APs  sending  out  updates  the  RF  environment.  This  is  important  in  creating  a  path-­‐loss  diagram  and  also  mapping  AP  locations.    

9. Adaptive  power  and  channel  assignments:  Automatically  assigns  all  AP  channel  and  power  settings.  Supports  802.11n  and    802.11ac  wide-­‐band  channels,  including  40  MHz  and  80  MHz  Automates  many  setup  tasks  during  network  installation  and  during  ongoing  operation  when  RF  conditions  change.    

10. Mode-­‐aware  ARM:  Aruba  APs  dynamically  detect  when  radios  have  overlapping  coverage,  remove  oversubscribed  APs  from  serving  clients,  stop  beaconing,  and  turn  into  air  monitors.  Provides  additional  network  security  through  greater  wireless  intrusion  detection  and  frees-­‐up  airspace  for  client  traffic.          

CONCLUSION:  

 

From  the  above  set  of  experiments  and  their  results  we  observed  the  following  important  factors:  

• Rate  Verses  Range:  While  an  ideal  test  implementation  would  show  very  high  throughput  in  terms  of  hundreds  of  Mbps  at  short  range  distances,  these  values  will  drop  slowly  at  a  few  tens  of  feet  and  thereon  will  drop  suddenly  to  almost  no  achievable  throughput  values.  This  is  steep  drop  is  decided  by  the  received  power  level  from  the  access  point  or  the  RSSI  value  of  the  form  of  -­‐85  dBm.    On  the  other  hand  when  other  clients  are  introduced  in  similar  deployment  the  throughput  decreases  substantially  due  to  channel  sharing,  also  the  range  the  decreases  such  that  locations  which  are  accessible  in  a  one  client  scenario  are  then  not  accessible  in  case  of  multiple  clients  at  same  distances.    

• Multi-­‐Client  Test:  On  one  hand  where  a  practical  situation  does  not  include  just  one  host  in  the  network,  it  was  worth  noting  the  way  in  which  the  throughput  was  almost  lost  when  many  clients  were  deployed  in  the  same  network  which  gave  a  very  high  throughput  on  one  host  implementation  but  lost  such  a  performance  when  multiple  clients  were  implemented.  We  not  only  observed  that  the  throughput  was  decreased  for  each  client,  but  also  the  aggregated  throughput  of  the  network  dropped  at  a  very  high  level,  while  new  clients  were  added  at  regular  intervals.  We  also  observed  an  increase  in  performance  when  the  clients  were  allowed  to  stable  after  adding  all  the  test  clients  and  then  waiting  for  a  sufficient  amount  of  time.  

We  also  tried  to  understand  Aruba  Radio  Management  interface,  which  is  a  smart  solution  to  understand  the  variation  in  environment  and  adapt  to  changes  such  that  the  wireless  network  achieves  its  maximum  performance.