Maximizing Visibility of GTP Sessions in Mobile Networks
Maximizing Visibility of GTP Sessions in Mobile Networks White Paper February 2013 Page 1 Executive Summary As  the  penetration  of  mobile  subscribers  rises  in  each  market  around  the  world,   so   competition  among  network  operators   increases   to  both   reduce   churn  and   acquire   new   customers.    Mobile   subscribers   are   increasingly  moving   from   2G   and   3G   networks   to   4G   LTE,   using   devices   that   demand   more   network   bandwidth.    Consider   that  by  2017,   iGR   forecasts   that  nearly  1.4  billion  of   the   world’s  7.4  billion  mobile  connections  will  use  LTE.   Not  only  do  LTE  subscribers  consume  more  network  capacity  but  they  also  use   more  varied  types  of  traffic.    For  example,  in  2012  in  the  U.S.,  video  accounted   for   just  over  54  percent  of  all  mobile  data   traffic  –  by  2017,   iGR   forecasts   this   figure  to  grow  to  two  thirds  of  mobile  data  network  use.    Much  of  this  growth  is   due  to  subscribers  moving  from  3G  to  4G  LTE  smartphones,  tablets  and  mobile   hotspot  devices.     As   a   result   of   these   shifts   in   the   market,   mobile   network   management   is   becoming   more   complex   and   more   critical   to   the   mobile   network   operator’s   business.     Rather   than   simply  manage   the   network   at   a  macro   level   as   in   the   past,  network  managers  are  now   required   to  provide  visibility  of   an   individual   subscriber’s  activity  at  the  session  level  for  each  device  they  may  be  using.       Network   management   information   and   data   is   now   as   likely   to   be   used   in   business   development   and   planning,   customer   care,   and   quality   of   service   measurement,  as  it  is  to  assess  the  overall  performance  of  the  physical  network.     Visibility   across   the   network   at   the   subscriber   and   session   level   is   required   to   feed  the  business  and  network  management  processes.    This  requires  the  ability   to   correlate   GTP   (GPRS   Tunneling   Protocol)   session   data,   which   provides   a   complete   picture   of   a   subscriber’s   activity   across   the   radio   access   and   core   networks.   The   issue   today   is   that  GTP  session  correlation   is  a   function  performed  by   the   network  probes  and  management  tools.    As  the  network  grows  to  support  the   increasing   demand   for   data   from   the   subscriber   base,   the   number   of   probes   must   grow   at   a   similar   rate   in   order   to   provide   visibility   and   GTP   session   correlation   resources.     As   well   as   being   costly,   this   approach   can   create   bottlenecks   in   the   network   management   system   as   the   number   of   probes   required  exceeds  the  network  management  tools’  ability  to  process  the  data.   The   solution   is   to   divorce   the   GTP   session   correlation   from   the   probes   by   offloading  the  GTP  session  analysis  to  a  GTP  session  controller.    This  allows  the   probes   to  monitor   the   network   without   the   need   to   analyze   the   data   in   real   time.   The  benefit  of  this  approach  is  that  as  the  network  scales,  the  number  of  probes   required   grows   at   a   slower   rate,   since   the   GTP   session   controller   is   now   responsible   for   GTP   session   data   correlation.     Rather   than   adding   expensive   probes,  the  network  manager  simply  adds  ports  to  the  GTP  session  controller.   Page 2 Secondly,  the  load  on  the  network  monitoring  tools  is  not  impacted  as  the  GTP   session  data  increases  –  this  allows  the  network  management  tools  to  focus  on   their   primary   task.   The   operator   therefore   does   not   have   to   buy   as   many   additional   tools   or   hire   the   staff   to   action   the   results,   as   would   be   required   without  the  GTP  session  controller.   Note  that  the  GTP  session  control  approach  still  provides  full  visibility  across  the   network   without   degrading   the   performance   of   the   probes   or   the   network   management  tools.    As  the  network  scales,  GTP  session  control  enables  a  cost-­‐ effective  and  efficient  way  for  the  network  manager  to  increase  management  of   the  network  at  the  subscriber  and  session  level.     Page 3 Worldwide Growth of LTE Networks LTE  is  growing  rapidly  around  the  world.  The  year  2012  saw  the  introduction  of   approximately   50   LTE   network   trials   and   several   commercial   deployments   worldwide.     iGR   expects   that   2013,   however,   will   be   characterized   by   the   commercial   availability   of  more   than   30   additional   deployments,   according   to   operator  and  industry  announcements.  As  more  of  the  mobile  operators’  traffic   shifts  to  4G  LTE  and  competition  drives  new  business  models,  so  the  demand  for   visibility   at   the   session   level   across   the   network   increases.     This   requires   correlation  of   the  GTP  (GPRS  Tunneling  Protocol)  sessions,  potentially   for  each   subscriber.   LTE Subscriber base Growth Obviously,   as   the   availability   of   LTE   services   increase,   so   does   the   subscriber   base.     While   worldwide   wireless   connections   are   growing   briskly   and   will   surpass   6   billion   connections   in   2013   before   reaching   7.5   billion   in   2017,   LTE   connections  are  forecast  to  exceed  1.4  billion  by  2017,  equivalent  to  nearly  19   percent   of   global   connections.     These   top-­‐level   forecasts   alone   show   how   people  the  world  over  increasingly  prefer  mobile  devices  as  their  main  voice  and   data  communications  tool.  In  some  regions,  a  mobile  phone  or  smartphone  may   be  the  user’s  only  connection  to  the  Internet.   Aside  from  the  raw  increase  in  the  number  of  connections  and  the  rapid  growth   of  LTE,  the  other  major  change  over  the  forecast  period  is  the  shift  from  2G  to   3G.   For   example,   in   2012,   2G   connections   comprised   71  percent   of   all  mobile   connections.   As   these   2G   connections   decline   over   the   next   5   years,   3G   connections   will   become   predominant,   rising   to   make   up   63   percent   of   all   connections   in   2017.   Figure   1   shows   the   overall   trend   in   connections   by   technology  generation.     A  similar  shift  is  now  also  starting  as  traffic  moves  from  3G  networks  to  4G  LTE.     LTE  growth   is   strongest   in  North  America   (Figure  2).    Over   the  next   five  years,   more   than   26   million   mobile   connections   will   be   added   and   penetration   will   reach  98  percent  in  2017.  Currently,  a  majority  of  North  American  connections   still   use   3G   technologies   (EV-­‐DO   and  UMTS/HSPA),   but   iGR   expects   LTE-­‐based   devices  to  rapidly  penetrate  the  market  with  nearly  58  million  LTE  connections   by  year-­‐end  2013.    By  2017,  nearly  52  percent  of  mobile  connections   in  North   America  will  use  LTE.     Note   that  while  North  America   is   a   forerunner  of  4G   technologies,   that   status   will  fade  as  LTE  rollouts  continue  across  the  globe.       Page 4 Figure 1: Global Connections by Technology Generation, 2012-2017 Source:  iGR,  2013   Figure 2: LTE North America Connections, 2012-2017 (000)   Source:  iGR,  2013   North American LTE mobile operators The  major  mobile   operators   in  North  America   have   either   launched   extensive   LTE  networks  or  are  expected  to  do  so  in  2013:   2.5%   5.0%   9.0%   13.8%   18.7%   0%  10%   20%  30%   40%  50%   60%  70%   80%  90%   100%   2012   2013   2014   2015   2016   2017   4G  3G  2G   23,020   57,894   90,846   122,621   154,337   185,717   331,076   337,160   342,739   347,887   352,812   357,403   0  50,000   100,000  150,000   200,000  250,000   300,000  350,000   400,000   2012   2013   2014   2015   2016   2017   LTE  Total   Page 5 o As   of   February   2013,   AT&T   offers   LTE   service   to   174   million   POPs   (population)  with  plans   to  grow  to  at   least  250  million  POPs  by  the  end  of   the  year.     o Verizon   Wireless   offers   the   widest   LTE   coverage,   covering   273.5   million   POPs  equivalent   to  approximately  89  percent  of   the  U.S.  population.  Since   Verizon  Wireless  was  the  first  major  operator  to  start  LTE  deployment,  the   operator   expects   to   finish   its   initial   LTE   deployment   by   mid   2013.     Subsequent   LTE   network   investments   will   then   be   focused   on   increasing   capacity.   o Sprint  first  launched  LTE  in  July  2012  and  now  covers  58  cities,  expecting  to   grow   to   200   million   POPs   by   the   end   of   2013.   Sprint   is   in   a   process   of   transformation   both   because   of   its   proposed   buyout   of   Clearwire   and   Sprint’s  pending  acquisition  by  Softbank.  At  the  time  of  this  writing  –  and  to   oversimplify   –   it  was   expected   that   Sprint   and   Softbank  would  merge   and   then   that  new  entity  would  acquire  Sprint’s   remaining  50  percent   stake   in   Clearwire.  Note  that  the  situation  is  complicated  by  DISH  Network’s  offer  of   $2.2  billion  for  24  percent  of  Clearwire’s  spectrum.   o T-­‐Mobile  USA  is  expected  to  launch  its  initial  LTE  markets  by  the  end  of  the   first  quarter  of  2013  and  then  plans  to  have  100  million  POPs  covered  by  the   middle   of   the   year   and   200  million   covered   six  months   later.     This  means   that  T-­‐Mobile  and  Sprint  will  have  equivalent  coverage  by  the  end  of  2013,   even   though   T-­‐Mobile   started   commercial   service   approximately   nine   months   later.     Also   note   that   T-­‐Mobile   is   deploying   Release   10   of   LTE,   known  as  LTE-­‐Advanced  and  will  be  the  first  operator  in  the  U.S.  to  offer  the   faster  version  of  LTE.   o Many  of   the   region’s  other  operators   are  also  moving   to  4G   technologies.   Major   Canadian   operators   Rogers  Wireless,   Bell,   and   Telus   have   all   begun   deploying   LTE.   MetroPCS   (which   is   being   acquired   by   T-­‐Mobile   USA)   was   actually  the  first  operator  to  deploy  LTE  in  North  America.    US  Cellular  and   other  smaller  regional  operators  have  also  started  LTE  trials  or  are  offering   commercial  services.   The  fact  that  LTE  is  being  offered  by  so  many  operators  in  North  America  results   in   rapid   subscriber   growth,   coupled   with   wide   availability   of   multiple   LTE   smartphones,  tablets  and  hotspot  devices.    By  2015,  iGR  expects  that  more  than   one  third  of  all  mobile  connections  in  North  America  will  use  LTE.   Of   course,   more   LTE   subscribers   means   that   bandwidth   demand   will   also   increase  –  LTE  subscribers  use  more  mobile  network  capacity  than  3G  users.  iGR   expects   that   the  3G  and  4G  LTE  networks  will  coexist   for  at   least   the  next   five   years,  even  though  LTE  will  grow  to  account  for  the  majority  of  the  traffic.  The   fact   that   the   two   technologies   will   coexist   and   interact   further   complicates   management  and  monitoring.   Page 6 Challenges to Network Monitoring The  rapid  increase  in  the  LTE  subscriber  base  will  have  several  repercussions  for   the  world’s  mobile  operators.    As  Figure  3  shows,  the  global  demand  for  mobile   data  bandwidth  is  forecast  to  increase  more  than  eleven-­‐fold  between  2012  and   2017,  reaching  nearly  1,000  PB  per  month.   Figure 3: Global Mobile Bandwidth Demand, 2012 - 2017 (PB per month) Source:  iGR,  2013   In  North  America,  bandwidth  demand   is   forecast   to  grow  more   than   six   times   over   the   same  period   (Figure  4),   from  189,000  TB  per  month   in  2012   to  more   than  1.15  million  TB  per  month  in  2017.    Note  that  the  demand  for  bandwidth   increases  far  faster  than  the  growth  in  number  of  subscribers.    LTE  subscribers   use   far   more  mobile   data   than   3G   users   and   so   as   more   consumers   buy   LTE   devices   (such   as   smartphones,   tablets,   mobile   hotspots,   eReaders   and   telematics   solutions),   the   average   bandwidth   consumed   by   each   subscriber   increases.   Since   the   mobile   operators   now   offer   tiered   mobile   data   rate   plans,   as   consumers   use   more   bandwidth,   the   ARPU   (Average   Revenue   per   User)   increases.    LTE  subscribers  therefore  tend,  on  average,  to  be  more  valuable  to  a   mobile  operator  than  a  3G  user.    An  operator’s  ability  to  provide  a  quality  LTE   experience   to   each   subscriber   is   therefore   critical   to   the   operator’s   ability   to   attract  new  subscribers  and  reduce  churn.    868.1      1,439.5      2,513.5      4,424.0      6,809.4      9,994.8     0   2,000   4,000   6,000   8,000   10,000   12,000   2012   2013   2014   2015   2016   2017   Page 7 Figure 4: North American Mobile Bandwidth Demand, 2012 - 2017 (TB per month) Source:  iGR,  2013   The   increase   in  mobile  data  bandwidth  demand   is  driven  by  the  growth   in   the   LTE  subscriber  base  but  also  by  growth  in  the  number  of  applications,  and  type   of   apps,   used.     Specifically,   the   growth   of  mobile   video   significantly   increases   bandwidth  demand  –  video  is  a  bandwidth  hog.   Changing mix of mobile data traffic According  to   iGR’s  research,  54  percent  of  the  mobile  data  traffic  used  in  2011   in   North   America   was   for   video.     By   2017,   video   is   forecast   to   comprise   66   percent  of  the  mobile  data  traffic  in  North  America.    This  obviously  contributes   significantly  to  the  overall  growth  in  demand  for  mobile  data.   As  well   as   video  downloads  and   streaming,  mobile   consumers   in   the  next   few   years   will   also   start   to   use   more   video   conferencing   services   (on   the   macro   mobile  data  network,  not  just  on  WiFi  as  most  video  conferencing  is  today)  and   more  VoLTE  (Voice  over  LTE).    These  services  demand  a  higher  level  of  QoS  from   the   network   –   a   consumer   is   far   more   likely   to   notice   dropped   packets,   for   example,  on  a  Skype  session  than  they  would  on  a  video  or  email  download.   VoLTE,   since   it   is   a   voice-­‐centric   application,   is   likely   to   have   high   QoS   expectations   by   consumers,   since   voice   services   are   seen   as   ‘basic’   mobile   services  with  which  users  have  the  most  experience.    For  example,  it  is  easy  for   a   consumer   to   differentiate   between   a   poor   cellular   voice   connection   and   a    189,324      286,837      408,628      587,604      826,986      1,155,744      -­‐          200,000      400,000      600,000      800,000      1,000,000      1,200,000      1,400,000     2012   2013   2014   2015   2016   2017   Page 8 good  one,  whereas  an  email  that  is  slow  to  download  due  to  poor  data  service   may  be  harder  to  recognize.   Increasing need for network monitoring As  subscribers  become  more  dependent  on  mobile  apps  and  services  using  LTE,   so  the  demand  for  detailed  information  on  the  user’s  behavior  will  increase.    In   order   to   diagnose   problems,   optimize   the   network   performance   or   simply   understand   subscriber   traffic  usage  patterns,   the  mobile  operators  need  more   detailed,  accurate  and  timely  information  from  all  parts  of  the  network.   Information  from  the  many  tools  mobile  operators  use  to  monitor  the  network   is  also  used  to  develop  new  business  policies,  rate  plans  and  products.  And  with   competition  being  what  it  is  in  the  wireless  market,  operators  must  quickly  react   with  new  products  and  services  to  head  off  a  competitive  threat  from  another   service  provider.    Detailed  network  intelligence  is  required  to  do  this.   Network   data   on   each   subscriber’s   behavior   and   application   usage   is   also   required   for   customer   care   and   accurate   problem   resolution.     As   a   subscriber   starts   to   use   more   advanced   services   and   apps   (such   as   video   conferencing),   with  a  corresponding  increase  in  monthly  data  charges,  so  the  operator  must  be   able   to   provide   quick   answers   to   address   any   issues   and   problems   the   subscriber  has.   The challenge with monitoring the mobile network The  main  problem  for  the  network  monitoring  tools  is  that  correlating  sessions   from   the   LTE   and   3G   networks   in   real   time   is   very   difficult   and   resource-­‐ intensive.     Multiple   probes   are   required   to   accurately   monitor   each   of   the   subscriber’s   data   sessions   in  order   to  build   an  accurate  picture  of   usage.     The   situation  is  further  complicated  by  the  fact  that  a  single  subscriber  using  a  single   smart  device  can  use  multiple  data  sessions  to  accomplish  different  tasks.   To  build  the  required  accurate  picture  of  a  subscriber’s  mobile  data  usage,  GTP   (GPRS  Tunneling  Protocol)   sessions  must  be  correlated  across  multiple  probes.   GTP   is   a   group   of   IP-­‐based   communications   protocols   used   to   carry   GPRS   (general  packet  radio  service)  within  GSM,  UMTS  (3G)  and  LTE  (4G)  networks.   GTP  uses  three  separate  protocols  (defined  as  GTP-­‐C,  GTP-­‐U  and  GTP  prime)  for   signaling  between  the  radio  access  network  and  the  core  network,  and  between   the  components  within  the  core  network  (S-­‐GW  and  P-­‐GW  for  example).    Thus,   in  order  to  obtain  a  complete  picture  of  a  subscriber’s  data  session  (for  QoS  or   network  management  purposes),  all  of  the  GTP  protocol  data  associated  with  a   single   session  must   be   correlated.     Since   different   GTP   protocols   are   used   at   different  points  in  the  network,  data  from  multiple  probes  must  be  assessed  to   build  a  composite  of  the  subscriber’s  activity.     Correlating  GTP   sessions   across   a   few   probes   for   a   few   subscribers   is   not   too   much   of   a   challenge.     But   as   the   LTE   subscriber   base   continues   to   grow   and   Page 9 subscribers  use  more  varied  apps  and  services,   so  more  probes  are  needed   to   correlate  the  increasing  number  of  GTP  sessions.    The  challenge  is  therefore  one   of   cost   and   scale  –  how  does  an  operator  efficiently   and   cost-­‐effectively   scale   the  GTP  correlation  resources  as  the  LTE  subscriber  base  and  LTE  traffic  rapidly   increase?       As  the  network  scales,  so  the  mobile  operator  is  going  to  need  to  track  millions   of   sessions   by   correlating   GTP   session   data   from   many   probes   across   the   network.     To   avoid   creating   bottlenecks   in   the   network,   the   load   on   the   individual  probes  must  be  balanced  and  coordinated  as  the  number  of  sessions   being  tracked  grows.    Typically,  the  correlation  occurs  in  the  probes  themselves,   thereby   adding   to   the   processing   load   on   each   probe.     As   the   next   section   discusses,   iGR   believes   this   may   not   be   the   most   efficient   or   cost-­‐effective   approach.     Page 10 GTP Session Controller Solution Before   discussing   a   solution   that   enables   the   network   probes   to   scale   as   the   need  for  GTP  session  correlation  increases,  we  must  first  look  at  how  the  whole   LTE  network  will  scale.       How LTE networks will scale Figure  5  shows  how  the   Ixia  Anue  Net  Tool  Optimizer   (NTO)  takes   information   from  probes  in  the  LTE  EPC  (Evolved  Packet  Core)  and  feeds  the  optimized  data   to   the  monitoring   probes.     Using   the   NTO   (discussed   in   a   previous   iGR  white   paper)   allows  EPC  monitoring   to   scale  without  having   to   increase  at   the   same   rate  the  number  of  tools  and  analysis  resources  required  to  support  the  growing   network.   Figure 5: LTE network architecture   Source:  iGR,  2013   But  how  much  will   the   LTE  network   scale?     Figure  6  below  shows   the   relative   scale  for  the  different  parts  of  the  LTE  network:   o UE   (User   Equipment)   –   these   are   the   subscriber   devices   such   as   smartphones  and  tablets,  which  will  obviously  number  in  the  millions,  even   for  a  single  mobile  operator.   o eNodeB   is   the   LTE   radio   base   station   –   initially,   tens   of   thousands   of   eNodeBs  are  required  to  provide  coverage  across  a  country  the  size  of  the   U.S.    But  with  the  deployment  of  LTE  small  cells,  the  number  of  eNodeBs  will   climb  into  the  hundreds  of  thousands.   Page 11 o MME   and   S-­‐GW  –   this   combination  will   scale   according   to   the   size   of   the   mobile  operator.     A   large  operator  will   likely  need  80   to  100  data   centers   containing  the  MMEs  and  S-­‐GWs  for  the  local  markets.    An  NTO  is  deployed   in  each  data  center  to  optimize  the  probe  data  for  the  monitoring  tools.   o P-­‐GW,  HSS  and  PCRF  –  relatively  few  of  these  elements  are  required.    For  a   large   operator,   it   is   likely   that   less   than   50   P-­‐GWs,   for   example,   will   be   needed.     Again,   an   NTO   is   deployed   in   each   data   center   to   optimize   the   probe  data  for  the  monitoring  tools.   o The  monitoring  tools  are  typically  located  in  a  single  location  –  in  this  way,   the   network   can   be   effectively   managed   by   a   single   group   of   network   managers  who  have  visibility  across  the  network.   Figure 6: Scaling LTE network architecture   Source:  iGR,  2013   Scaling mobile network monitoring Now  to   the  problem  of  GTP   session  correlation.  As   the  network   scales,   so   the   mobile   operator   is   going   to   need   to   correlate   GTP   session   data   from   many   probes   across   the   network.     Since   the   correlation   typically   occurs   in   the   tools   themselves,   this  would  necessarily  mean   adding  many  more   tool   resources   to   process   the   correlation.     Based   on   conversations   with   the   major   mobile   operators,   iGR   believes   this   approach   would   be   both   resource   and   cost   prohibitive.   An   alternative   approach   is   shown   in   Figure   7.     In   this   example,   an   Ixia   GTP   Session   Controller   offloads   the   correlation   of   the   sessions   from   the   network   Page 12 monitoring   tools,   thereby   allowing   the   tools   to   focus   on   network   monitoring   tasks   and   to   scale   as   traffic   grows.     Note   that   the   NTO   feeds   data   from   the   network   to   the   GTP   Session   Controller   –   a   single   GTP   Session   Controller   can   process  the  data  from  up  to  27  million  LTE  subscriber  sessions.   Figure 7: LTE network architecture with GTP session controller   Source:  iGR,  2013   In   the   physical   implementation,   the   Ixia   GTP   Session   Controller   would   be   located  with  the  network  management  tools.    Even  for  a  major  mobile  operator,   just  a  few  GTP  Session  Controllers  are  likely  to  be  required  to  correlate  the  GTP   session  data.   Advantages of the GTP Session Controller solution From   iGR’s  perspective,  using  the  GTP  Session  Controller  approach  has  several   advantages:   o The  architecture  easily   scales  as   traffic   grows  –   rather   than  adding  probes   and  tools,  the  operator  simply  adds  GTP  session  controller  ports.   o The   load   on   the   network   monitoring   tools   is   not   impacted   as   the   GTP   session  data  increases  –  this  allows  the  tools  to  focus  on  their  primary  task.     This  also  means  the  operator  does  not  have  to  buy  as  many  additional  tools   or  hire  the  staff  to  action  the  results  as  would  be  required  without  the  GTP   Session  Controller.   Page 13 o Enables   continued   end-­‐to-­‐end   visibility   across   the   network.     This   is   important   since   the  other  network  monitoring   tools’   performance  will   not   be  impacted  by  the  need  for  GTP  session  correlation.   Page 14 About iGR iGR   is   a   market   strategy   consultancy   focused   on   the   wireless   and   mobile   communications  industry.  Founded  by  Iain  Gillott,  one  of  the  wireless  industry's   leading  analysts,  we  research  and  analyze  the   impact  new  wireless  and  mobile   technologies  will  have  on  the  industry,  on  vendors'  competitive  positioning,  and   on  our  clients'  strategic  business  plans.   Our   clients   typically   include   service   providers,   equipment   vendors,   mobile   Internet   software   providers,   wireless   ASPs,   mobile   commerce   vendors,   and   billing,   provisioning,   and   back   office   solution   providers.   We   offer   a   range   of   services   to   help   companies   improve   their   position   in   the  marketplace,   clearly   define  their  future  direction,  and,  ultimately,  improve  their  bottom  line.     A  more  complete  profile  of  the  company  can  be  found  at  www.iGR-­‐   Disclaimer The  opinions  expressed  in  this  white  paper  are  those  of   iGR  and  do  not  reflect   the   opinions   of   the   companies   or   organizations   referenced   in   this   paper.   All   research  was  conducted  exclusively  and  independently  by  iGR.    This  white  paper   was   sponsored   by   Ixia,   but   Ixia   personnel   were   not   involved   in   the   carrier   interviews  or  in  the  ongoing  research.