cisco in mobile backhaul, 1995-â€2009. henrik glimstedt, stockholm

25
Cisco in Mobile Backhaul, 19952009. Henrik Glimstedt, Stockholm School of Economics Dec 5th, 2012 INTRODUCTION By the late 1990s, few companies attracted so much positive attention as Cisco did. It seen as the bearer of what was around the corner: horizontally and vertically specialists that, through focusing on a narrow set of core competencies and open standards and product platforms, retired old concepts and drive incumbents out of business. Much of this reputation was gained as ITmanagers bought Cisco’s routers and switches in drives as they interconnected computers, offices, departments and sites. As we all became familiar with the concept of the Internet as well as mobile phones in the 1990s, Cisco’s familiar IP routers and switches (or rather bigger versions thereof) began to appear in infrastructure networks owned and run by telecom operators. Would Cisco, many asked, become a dominating force also in this market; if so: would Cisco drive out the operators key suppliers of telecom equipment vendors? Much pointed in that direction. Heavy Reading, an analyst outfit covering the comtech market space, concluded in 2003 that “Cisco System now dominates mindshare in the carrier market. Service providers believe that Cisco successfully transformed itself from an enterprisefocused vendor to the leading telecom supplier in the world.” (Heavy Reading 2003, p 10). The survey showed that operators ranked Cisco as #1 in terms of brand, product performance, price, quality and service & support. A few years earlier, this position was doubtless Lucent’s. If the customers were in agreement on Cisco’s advantages, success should have been certain. Yet, Cisco never disrupted the telecom business the way it disrupted the enterprise and Internet gear business. On the contrary: It has been falling behind in its core markets. Cisco controlled 42 percent of the $5.9 billion edge router market last year, down from 57 percent in 2006, according to market watchers at Dell’Oro Group. Even in LAN and WAN technologies –Cisco’s home turf—the days of comfortable lead are numbered: Cisco has shed as much as six points of its commanding market share in Ethernet switches, as all of its competitors have registered gains. For people like myself with interest in the recent history of communication systems and strategies of hightech companies it has always been puzzling as for why Cisco was unable, despite all the M&A and internal R&Dprojects on carrierclass routers, dominate this market. This memo attempts to, at least in part, give some good answers. It does so by analyzing Cisco’s attempt to enter one of the submarkets that make up the carrierclass communication equipment market, namely mobile backhaul.

Upload: others

Post on 09-Feb-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Cisco  in  Mobile  Backhaul,  1995-­‐2009.  

Henrik  Glimstedt,  Stockholm  School  of  Economics  

Dec  5th,  2012  

 

 

INTRODUCTION  

By  the  late  1990s,  few  companies  attracted  so  much  positive  attention  as  Cisco  did.  It  seen  as  the  bearer  of  what  was  around  the  corner:  horizontally  and  vertically  specialists  that,  through  focusing  on  a  narrow  set  of  core  competencies  and  open  standards  and  product  platforms,  retired  old  concepts  and  drive  incumbents  out  of  business.  Much  of  this  reputation  was  gained  as  IT-­‐managers  bought  Cisco’s  routers  and  switches  in  drives  as  they  interconnected  computers,  offices,  departments  and  sites.  As  we  all  became  familiar  with  the  concept  of  the  Internet  as  well  as  mobile  phones  in  the  1990s,  Cisco’s  familiar  IP  routers  and  switches  (or  rather  bigger  versions  thereof)  began  to  appear  in  infrastructure  networks  owned  and  run  by  telecom  operators.    Would  Cisco,  many  asked,  become  a  dominating  force  also  in  this  market;  if  so:  would  Cisco  drive  out  the  operators  key  suppliers  of  telecom  equipment  vendors?  Much  pointed  in  that  direction.  Heavy  Reading,  an  analyst  outfit  covering  the  comtech  market  space,  concluded  in  2003  that  “Cisco  System  now  dominates  mindshare  in  the  carrier  market.  Service  providers  believe  that  Cisco  successfully  transformed  itself  from  an  enterprise-­‐focused  vendor  to  the  leading  telecom  supplier  in  the  world.”  (Heavy  Reading  2003,  p  10).  The  survey  showed  that  operators  ranked  Cisco  as  #1  in  terms  of  brand,  product  performance,  price,  quality  and  service  &  support.  A  few  years  earlier,  this  position  was  doubtless  Lucent’s.            

If  the  customers  were  in  agreement  on  Cisco’s  advantages,  success  should  have  been  certain.  Yet,  Cisco  never  disrupted  the  telecom  business  the  way  it  disrupted  the  enterprise  and  Internet  gear  business.  On  the  contrary:  It  has  been  falling  behind  in  its  core  markets.  Cisco  controlled  42  percent  of  the  $5.9  billion  edge  router  market  last  year,  down  from  57  percent  in  2006,  according  to  market  watchers  at  Dell’Oro  Group.  Even  in  LAN  and  WAN  technologies  –Cisco’s  home  turf—the  days  of  comfortable  lead  are  numbered:  Cisco  has  shed  as  much  as  six  points  of  its  commanding  market  share  in  Ethernet  switches,  as  all  of  its  competitors  have  registered  gains.    

For  people  -­‐-­‐like  myself-­‐-­‐  with  interest  in  the  recent  history  of  communication  systems  and  strategies  of  high-­‐tech  companies  it  has  always  been  puzzling  as  for  why  Cisco  was  unable,  despite  all  the  M&A  and  internal  R&D-­‐projects  on  carrier-­‐class  routers,  dominate  this  market.  

This  memo  attempts  to,  at  least  in  part,  give  some  good  answers.  It  does  so  by  analyzing  Cisco’s  attempt  to  enter  one  of  the  sub-­‐markets  that  make  up  the  carrier-­‐class  communication  equipment  market,  namely  mobile  backhaul.    

 

 

Definition  of  the  backhaul  market  space            

In  a  mobile  network,  the  radio  base  station  connects  the  user’s  mobile  phone  to  the  network.  It  catches  the  radio  signal.    Backhaul  is  defined  as  transmitting  that  signal,  that  is,  voice  and  data  traffic  from  the  radio  base  station  cell  site  to  a  point  of  the  mobile  core  network.  This  report  hence  defines  backhaul  as  all  Layer  1/2/3  transport,  aggregation  and  switching  nodes  residing  between  radio  base  stations  and  the  mobile  core  transmission  network.      

Fig  1:  Access,  Backhaul  and  Core  Networks  

 

Source:  Cisco  

 Apart  from  just  connecting  and  transporting  the  stream  of  signals  between  the  radio  base  station  and  the  radio  base  station  controller  (RBS  in  2G  networks)  or  Radio  Network  Controller  (RNC  in  3G),  a  process  referred  to  as  “traffic  aggregation”  is  an  essential  part  of  backhaul.  This  task  is  critical  to  network  operators  for  two  reasons:    

• Aggregation  serves  the  purpose  of  combining  signals  sent  from  many  different  radio  base  stations  served  by  a  single  RBC/RNC-­‐node  into  a  ‘thicker’  pipe,  taking  advantage  of  ATM’s  inherent  statistical  multiplexing  to  reduce  bandwidth  

• Supporting  different  types  of  traffic  from  different  mobile  generations  (2g,  3G…)  over  the  same  transport  link,  rather  than  serving  different  radio  base  stations  technologies  typical  to  2G  and  3G  over  different  links.  

Both  types  of  aggregation  serve  to  leverage  existing  network,  especially  where  traffic  is  dense.  

Market  Drivers  in  Mobile  Backhaul    

While  an  exclusive  show-­‐off  item,  or  at  best,  communication  tool  reserved  for  a  social  elites  and  executives  in  the  1980s,  mobile  phone  penetration  accelerated  drastically  in  the  1990s.  After  reducing  risks  for  operators  –  and  parents  –  through  pre-­‐paid  services,  the  mobile  phones  were  suddenly  not  so  exclusive  any  more.  Deregulation  and  competition  of  the  telecom  markets  drove  down  cost  of  calling.  Economies  of  scale  in  mobile  phones,  as  well  as  vendor  subsidies  to  new  subscribers,  then  drove  down  cost  of  buying  a  cell  phone.  As  reforms  led  to  growth  and  new  patterns  

of  consumption  in  the  emerging  markets,  the  patters  repeated  itself  also  outside  Western  part  of  Europe,  Northern  America,  Japan  and  Korea.  The  number  of  subscribers  doubled  from  0,5  to  1  Bn  between  1998  and  2002.  Only  In  2005  another  300  million  subscribers  were  added,  taking  the  global  past  2  bn  subscribers  (of  which  70%  was  on  GSM)  

Fig  2:  Subscribers  by  technology,  actual  and  projected  growth  

 Source:  Dell’Oro  

Demand  for  backhaul  is  driven  by  deployment  of  new  cell  sites  as  well  as  the  upgrade  of  systems  to  handle  the  throughput  required  by  mobile  technologies,  such  as  3G.  Mobile  operators  install  new  cell  sites  for  a  couple  of  reasons.  The  first  is  to  expand  geographic  coverage  to  cover  a  higher  number  of  mobile  phone  users.  

 The  second  reason  is  to  sustain  user  quality  of  service  and  the  average  expected  download  speed  of  the  mobile  smartphone,  as  reflected  in  the  relative  importance  of  WCDMA  above.  

 

 

 

 

 

 

 

 

 

0.0  

1,000.0  

2,000.0  

3,000.0  

4,000.0  

5,000.0  

6,000.0  

7,000.0  

8,000.0  

9,000.0  

1999  

2000  

2001  

2002  

2003  

2004  

2005  

2006  

2007  

2008  

2009  

2010  

2011  

2012  

2013  

2014  

2015  

TD-­‐SCDMA  

LTE  

WCDMA  

CDMA  (Includes  CDMA  1x)  

TDMA  

GSM/GPRS/EDGE  

 

Fig  3:  Total  number  of  cell  sites,  1998-­‐2012    

 

Source:    

The  number  of  sites  also,  as  hinted  above,  began  to  increase  as  operators  experimented  with  improving  coverage  through  combining  cells  equipped  with  different  types  of  RBS,  filling  in  spots  of  bad  coverage.    Whilst  the  original  base  stations  in  the  second  generation  systems  had  a  standard  coverage,  equipment  manufacturers  started  to  introduce  RBS  technologies  with  varying  coverage.  

Fig  4:  Radio  Base  station  Coverage  

   

Source:    Yankee  Group,  2005        

Smaller  RBS,  i.e.  so  called  micro  or    pico  cells,  allowed  operators  to  find  new  solutions  to  the  problem  of  covering  demanding  urban  areas  through  more  advanced  cell  planning.  As  we  can  see  from  the  example,  these  combinations  of  macro  and  micro  cells  soon  pushed  the  demand  for  backhaul  as  the  sheer  number  of  cells  increased.  

Combined,  the  driving  forces  turned  mobile  backhaul  into  major  cost  for  the  operators  as  well  as  an  important  market  for  equipment  manufacturers.  It  is  estimated  that  backhaul  can  account  for  as  much  as  30%  of  a  mobile  operator’s  operating  costs  (Yankee  Group,  2005)          

 

Technologies  and  Standardization  in  Backhaul  Networks  

2G  backhaul  systems  were  built  using  the  connections  between  radio  base  station  and  the  core  network  run  through  either  the  same  copper  or  fiber  network  that  connects  homes  and  offices  to  the  core  networks  (phone  and  Internet),  or  by  point  to  point  microwave  connections  transmitting  the  signal  from  a  “sender”  radio  on  the  cell  site  to  a  receiver  radio  close  to  the  core  network  ,  i.e.  a  substitute  for  the  fixed  line  connection.  

Backhaul  transmission  from  RBS  to  RBC/RNC  to  has  traditionally  been  implemented  through  time  division  multiplexing  (TDM)  circuits  over  local  T1  copper  lines  and  SONET/PDH  optical  rings.  

Fig  5:  Physical  carrier  in  voice  networks  

 

 In  its  primary  form,  TDM  (time  division  multiplexing)  was  developed  used  for  controlling  fixed  line  telecommunications  systems.  It  is  a  well-­‐established  two-­‐way  communication  technology  that  works  well  for  real-­‐time  applications  (e.g.  voice).  The  advantage  of  TDM  is  in  combining  efficient  use  of  bandwidth  with  low  latency.1  If  latency  exceeds  150  milliseconds  for  ex-­‐ample,  the  quality  of  the  conversation  begins  to  drag.    At  or  above  200  milliseconds  many  listeners  may  find  a  conversation  unintelligible.  TDM  networks  are  inherently  synchronous.  All  network  components  must  be  synchronized  with  each  other  to  ensure  that  data  is  not  lost.  In  a  native  TDM  network,  clock  synchronization  is  performed  at  the  physical  layer,  and  clocking  information  is  carried  along  with  data  traffic.  To  make  efficient  use  if  wires,  TDM  has  an  inherent  method  of  putting  multiple  data  streams  in  a  single  signal  by  separating  the  signal  into  many  segments,  each  having  a  very  short  duration.  Each  individual  data  stream  is  reassembled  at  the  receiving  end  based  on  the  timing.  Thus  several  phone  conversations  can  share  a  signal  transferred  over,  say,  a  copper  line.  If  many  signals  must  be  sent  along  a  single  long-­‐distance  line,  careful  engineering  is  required  to  ensure  that  the  system  will  perform  properly.  An  asset  of  TDM  is  its  flexibility.  The  scheme  allows  for  variation  in  the  number  of  signals  being  sent  along  the  line,  and  constantly  adjusts  the  time  intervals  to  make  optimum  use  of  the  available  bandwidth.    

                                                                                                                         1  Latency  is  the  time  it  takes  for  the  data  to  reach  its  destination.  As  TDM  allocates  time  periods,  only  one  channel  can  transmit  at  a  given  time,  and  some  data  would  often  be  delayed,  though  it’s  often  only  in  milliseconds.    

 

SONET/SDH  was  originally  designed  to  transport  TDM  voice  services  in  larger  volumes  over  fiber  optic  networks,  thus  taking  advantage  of  the  rapid  development  in  fiber-­‐optical  networking.  The  forced  breakup  of  AT&T  in  the  United  States  created  numerous  regional  telephone  companies  with  surprising  degree  of  difficulties  in  establishing  interoperability  across  the  different  telecom  networks.  Fiber  optic  cabling  already  prevailed  for  long  distance  voice  traffic  transmissions,  but  the  existing  networks,  involving  T1  connections  as  well  other  types  transmission,  proved  unnecessarily  expensive  to  build  and  difficult  to  extend  for  so-­‐called  long  haul  data  and/or  video  traffic.    

Hence  operators  began  to  work  on  a  common  standard  specifically  designed  for  interconnecting  for  the  public  telephone  network  to  handle  bulk  voice  circuits  with  maximum  uptime,  minimal  delay  and  guaranteed  service  continuity  in  mind.  The  technology  has  very  strong  support  for  latency  control  and  hence  network  synchronization  through  very  accurate  clocking  operations,  which  also  could  be  coordinated  with  TDM  clocking.  The  technology  is  time-­‐tested,  simple  and  deterministic.  At  any  given  time  the  operator  knows  what  traffic  is  going  over  the  links.  

Architectural  Control:  3G  Standardization  of  Backhaul  Technology    

With  the  introduction  of  mobile  broadband  technologies  mobile  operators  needed  –  many  operators  calculations  showed  -­‐-­‐  to  increase  cell  site  backhaul  capacity  from  a  typical  2  Mbps  (2  E1/T1)  at  an  edge  site  to  ca  8  Mbps.  If  high-­‐speed  data  services  would  take  off  in  a  big  way,  as  they  certainly  did  with  the  introduction  of  the  iPhone  and  small  modems  connecting  lap-­‐tops  to  the  internet  via  3G  networks  -­‐-­‐  operators  forecasted  capacity  need  to  grow  to  between  20  and  40  Mbps.    At  hubs  or  aggregation  centers,  where  traffic  from  multiple  sites  is  combined,  backhaul  capacity  may  need  to  be  as  high  as  2  Gbps.  These  forecasts,  and  the  soaring  number  of  cell  sites,  made  the  TDM/SONET  solution  obsolete:  

• Cost  structure:  Unlike  infrastructure  costs,  backhaul  expenses,  especially  leased  lines,  are  on-­‐going  costs  without  adding  much  to  service  differentiation.    Any  cost  savings  realized  in  backhaul  go  straight  to  the  bottom  line.    

• Complexity:  Approximately  80%  of  existing  backhaul  networks  comprise  of  legacy  SONET  microwave  links  and  multiplexers  stacking  many  2G-­‐  and  3G-­‐related  boxes  in  coexisting  sites.  This  increases  the  CAPEX  as  well  as  operational  difficulties.  

• Scalability:  A  typical  cell  site  requires  two  or  three  leased  T1/E1  lines,  representing  4  to  6  Mbps  of  bandwidth.  New  data  intensive  mobile  services  proved  to  be  more  than  double  this  requirement.  Adding  this  much  capacity  via  TDM  lines  is  time-­‐consuming  and  economically  prohibitive.  Carriers  need  the  ability  to  add  capacity  on  demand  to  respond  to  changing  customer  needs.    

• Efficiency:  Because  each  T1/E1  line  is  dedicated,  excess  capacity  cannot  easily  be  shared.  The  current  method  of  providing  backhaul  capacity  invariably  involves  a  substantial  amount  of  unused—and  expensive—bandwidth  in  the  mobile  backhaul.  

 

Which  technology  should  be  used  for  backhaul  transport?    There  were  two  dimensions  to  consider  –  physical  carrier  media  and  routing/switching.  As  for  the  physical  layer,  there  were  a  choice  between  fixed  lines  and  wireless  backhaul  connections.  The  rational  for  introducing  SONET  fiber-­‐optical  networking  in  backhaul  was  driven  by  improved  scalability  as  much  as  it  represented  a  strong  

mechanism  for  network  control.  Compared  to  copper  wire  technology  (TDM  over  T1).  SONET  was  typified  by  a  steep  staircase  cost  function  in  the  provisioning  of  RAN  bandwidth  as  opposed  to  the  roughly  linear  relationship  between  bandwidth  and  cost  in  copper  lines.  SONET  could  also  be  deployed  on  wireless  connection  systems.    

 

 

Using  point-­‐to-­‐point  radio  microwaves  for  SONET/PDH,  that  is,  replacing  fixed  line  networks  in  backhaul  with  a  system  of  radio  stations  carrying  SONET  from  the  cell  site  to  the  edge  of  the  core  network,  had  a  similar  investment  step-­‐by-­‐step  curve.    

By  comparison,  Ethernet  technologies,  which  were  mostly  deployed  in  enterprise  and  LAN  networks  scale  at  very  low  cost.  Amongst  actors  in  the  industry,  it  was  widely  held  that  Ethernet  would  be  preferred  physical  carrier  for  the  future.  

Fig  6:  Scalability  by  Cost  and  Bandwidth  in  Ethernet,  Copper  and  Microwave      

Source:  Alcatel-­‐Lucent  2008.  

Operator’s  calculations  of  CAPEX  in  relation  to  potential  profits  from  services  suggested  that  the  TDM  over  leased  copper  T1  lines  represented  the  most  expensive  option  for  future  expansion  to  accommodate  new  demand  for  bandwidth  beyond  2MBS  typical  to  the  GSM  systems.  The  ideal  scenario  would  be  a  self-­‐build  Ethernet,  representing  a  saving  no  less  than  60%  compared  to  the  leased  copper  lines  alternative  (Nokia  Siemens  Networks  2008).  The  question  was  however  not  self-­‐contained.  Rather,  the  choice  of  physical  network  technology  closely  tied  to  other  technological  choices,  most  notably  to  standards  for  switching  and  routing.    

These  issues  question  were  closely  monitored  within  the  global  and  regional  telecom  standardization  bodies  involved  o  set  new  3G  standards  for  wireless  broad  band  services  between  1994  and  1999,  that  is  to  the  3GPPP  and  it  related  regional  sub-­‐organizations.  To  present  the  highly  political  and  strategic  process  of  agreeing  on  the  architecture  of  the  2G  and  3G  systems  (e.g.  Glimstedt  2001)  here  would  lead  us  beyond  the  scope  of  this  report.  Rather  it  suffice  here  to  indicate  that  actors  within  the  standardization  body  already  in  the  1990s  were  studying  how  to  introduce  support  for  mobile  broad  band  communication  whilst  still  maintain  high  quality  of  service,  i.e.  low  latency  voice  services.  From  the  perspective  of  telecom  operators,  there  was  a  trade-­‐off  between  two  goals:    

• support  new  multimedia  services,  building  on  packet  switching  transmission  technologies  because  it  offers  superior  scalability    

• still  have  strong  control  to  be  able  to  differentiate  between  high-­‐quality  voice  services  

The  widespread  diffusion  of  fixed  line  Internet  communications  made  TCP/IP  a  natural  candidate  for  3G  networks.  IP  excelled  in  terms  of  scalability.  Exactly  how  the  traffic  was  routed  was  determined  in  a  highly  decentralized  fashion,  making  control  the  weak  point  of  IP.  In  routed  IP  networks,  traffic  moves  between  routers  based  on  adaptive  topology  updates  based  on  the  exchange  of  "reachability"  information  among  network  routers.  Traffic  was  routed  "hop  by  hop"  from  router  to  router,  following  a  path  that  could  be  determined  only  by  examining  the  sum  of  the  routing  tables  in  the  network  at  the  time  the  traffic  was  moving.  This  made  it  impossible  to  engineer  specific  quality  of  service  by  allocating  resources  to  traffic  types.    

While  the  key  actors  within  the  world  of  telecom  standardization  clearly  identified  support  for  multimedia  services,  and  hence  packet  switching  such  as  IP,  as  critical  to  future  profits,  they  were  equally  preoccupied  by  the  commitment  to  carrier  grade  switched  voice  services.  After  all,  it  was  the  switched  high-­‐quality  voice  call  upon  which  the  telecom  operator  business  models  was  building.  Customers  were  comfortable  with  paying  for  exactly  this:  access  to  a  unique  connection  with  good  voice  quality.  Even  if  the  telecom  standardization  process  now  was  geared  towards  including  support  for  future  oriented  multimedia  services  and  packet  switching,  there  was  a  strong  tendency  towards  solutions  that  would  protect  the  basic  business  model  and  profits  associated  with  it.    

Three  key  decisions  defined  the  3G  standard,  all  of  which  reflected  the  bias  towards  protecting  the  basic  telecom  business  model.  First,  the  selection  use  WCDMA  for  the  interface  between  the  mobile  phone  and  the  radio  base  station,  (the  Uu  interface)  acknowledges  the  general  need  for  more  efficient  use  of  spectrum  and  increased  bandwidth.  Secondly,  ATM  was  selected  for  backhaul  network  (UTRAN)  ,  including  the  Iu  interface.  WCDMA  was  perhaps  the  most  critical  decision.  After  several  harmonization  attempts  of  five  original  proposals,  elaborated  to  a  large  extent  within  the  EC  funded  Advanced  Communications  Technologies  and  Services  (ACTS)  program,  an  agreement  by  consensus  was  reached  on  the  radio  access  techniques  for  UMTS,  W-­‐CDMA.  Compared  to  the  GSM  

system,  it  had  particular  advantages  in  terms  of  scalability,  efficient  use  of  spectrum  and  bandwidth  for  mobile  data  communications.    

Fig  7:  Interfaces  in  the  WCDMA  standard  

 

There  were  several  reasons  why  the  various  parties  in  the  standardization  process  arrived  at  ATM.  The  outcome  of  the  technology  selection  process  would  have,  as  we  will  see  below,  far  reaching  implications  for  Cisco  and  other  IP-­‐based  companies.    

Because  SONET  was  originally  designed  for  voice  and  not  variable-­‐sized  data  packets,  however,  moving  data  across  it  was  inefficient  and  required  padding  data  packets  with  irrelevant  data  to  make  up  any  differences.  Asynchronous  transfer  mode  (ATM)  was  introduced  as  a  solution  for  this  inefficiency.  Through  the  use  of  hardware  network  interface  adapters,  ATM  networks  break  data  into  smaller  cells  for  transport.,  ATM  was  based  on  rigorously  defined  levels  of  quality  of  services,  allowing  the  operator  to  set  different  QoS  for  different  types  of  transmission  –voice  gets  higher  priority  and  data  lower—or  indeed  give  an  individual  customer  access  to  higher  bandwidth  at  a  different  price.    Secondly,  the  choice  of  ATM  is  partly  dictated  by  the  decision  to  use  CDMA  in  the  air  interface.  Because  certain  aspects  of  the  CDMA  radio  traffic,  e.g.  as  hand-­‐over  between  different  radio  base  stations  when  the  end-­‐user  travels  between  different  cells,  require  very  strict  time  and  latency  control,  the  backhaul  system  needs  to  comply  with  these  latency  requirements.  ATM  involves  a  highly  sophisticated  mechanism  for  controlling  latency  and  jitter  called  ‘statistical’  multiplexing,  which  met  the  strict  requirements  for  time  synchronization  between  radio  and  backhaul  transmissions.  Third,  the  parties  also  agreed  that  ATM  has  a  high  degree  of  backwards  compatibility  with  respect  to  SONE.  Due  to  SONET  essential  protocol  neutrality  and  transport-­‐oriented  features,  SONET  was  the  obvious  choice  for  transporting  the  fixed  length  ATM  frames  also  known  as  cells.        

Cisco’s  backhaul  strategy:  criticism  of  the  short-­‐term  solution    

Cisco  was  no  stranger  to  ATM.  Already  in  the  1990s,  Cisco  first  joined  forces  with  AT&T  in  an  alliance  to  develop  capabilities  in  ATM  switching.  Much  in  the  same  fashion  as  Cisco  had  

acquired  IP  start-­‐ups,  the  company  also  started  to  shop  for  cutting-­‐edge  ATM  companies  resulting  in  the  acquisition  of,  among  others,  StrataCom,  ArrowPoint  and  Lightstream.    These  acquisitions  positioned  Cisco  as  a  credible  supplier  of  ATM-­‐based  LAN/WAN  solutions.  In  1998  the  Cahners  In-­‐Stat  Group  report  ranked  Cisco  first  in  worldwide  ATM  LAN  switch  port  shipments.  1998  fiscal  year  was  the  first  billion  dollar  year  for  ATM-­‐related  sales,  representing  a  50+  percent  growth  of  Cisco’s  ATM  business  from  1997.  Cisco's  worldwide  ATM  WAN  switch  manufacturer  sales  market  share  is  also  ranked  first  at  27,2  percent  for  the  first  half  of  1998.  Cisco’s  acquisitions  of  ATM-­‐players  did  not  provide  the  company  with  the  same  leverage  in  the  core  network  markets.  Cisco’s  heavy-­‐weight  ATM  switches  were  not  really  any  serious  match  for  the  incumbents  offering  carrier-­‐class  core  network  ATM  switches.  Also  Cisco’s  sales  in  large  scale  switching  also  trailed  some  entrants,  such  as  Cascade.  

To  make  up  for  its  weaknesses  in  core  network  ATM,  Cisco  had  another  play  card  up  its  sleeve  -­‐-­‐-­‐  MPLS,  announced  already  1996  by  Cisco.2    Both  ATM  and  MPLS  technologies  were  designed  to  accommodate  TDM  voice  and  with  various  types  of  data  communication  technologies  into  a  convergent  network.  The  primary  difference  between  ATM  and  MPLS  is  that  while  ATM  was  designed  to  exist  in  a  circuit-­‐switched  environment,  MPLS  has  its  place  within  modern  packet-­‐switched  networks,  such  as  Ethernet  or  IP.      

The  deployment  of  ATM  accelerated  through  the  diffusion  of  the  Internet  and  IP.  In  practice,  the  ATM  switches  were  deployed  at  the  core  network  to  integrate  IP-­‐based  LAN  and  WAN  networks.  However,  interface  differences  made  it  however  difficult  to  deploy  IP  services  over  ATM  networks.  In  particular,  ATM  required  the  segmentation  of  packet  data  in  ways  that  did  not  fit  with  the  principles  of  IP  packet  data.  ATM  uses  fixed-­‐length  packets,  which  implies  that  IP  packets  cannot  naturally  fit  into  ATM-­‐packets.  This  inherent  difference  made  it  necessary  to  translate  IP  into  ATM,  a  costly  and  slow  process  that  required  extra  investments.  

Cisco’s  solution  to  this  problem  was  MPLS,  which  provides  the  operators  with  a  high  degree  of  network  control  but  with  a  more  natural  fit  with  IP  services.    MPLS  defines  a  method  by  which  the  entre  IP-­‐packet  can  be  transported  in  a  single  MPLS-­‐frame,  which  solved  the  in-­‐efficiency  problem.  As  a  result,  MPLS  could  theoretically  replace  ATM  as  the  preferred  choice  for  transporting  IP  packet  in  mobile  backhaul  (and  core)  networks.      

Armed  with  MPLS,  no  company  within  the  backhaul  business  pressed  for  the  radical  all-­‐IP  scenario  harder  than  Cisco.  Once  the  notion  of  wireless  internet  services  diffused  across  the  landscape  of  equipment  vendors,  mobile  operators  and  policy  makers  influencing  spectrum  allocation,  Cisco  suggested  that  mobile  service  providers  could  use  Cisco  network  equipment  to  migrate  TDM  into  

                                                                                                                         2  The  ideas  had  been  brewing  in  the  company  since  the  early  1990s.  An  early  patent  (USPTO  Pat  #  6,147,999  describes  in  the  abstract  “…a  pipelined  multiple  issue  architecture  for  a  link  layer  or  protocol  layer  packet  switch,  which  processes  packets  independently  and  asynchronously,  but  reorders  them  into  their  original  order,  thus  preserving  the  original  incoming  packet  order.”  In  parallel  with  the  internal  R&D  program,  invited  partners  within  IETF  to  participate  in  the  setting  of  open  standards  for  Tag  Switching.  Whereas  Cisco  concentrated  on  the  standards  for  the  basic  MPLS  switching  functions,  other  participant  in  the  standardization  group  (e.g.  Asend,  IBM,  Toshiba,  and  Ipsilon)  contributed  with  standards  complementary  functions  (e.g.  wireless  functions  supporting  the  application  of  tag  switching  in  wireless  systems.)  

integrated  IP  packet-­‐based  networks  capable  of  supporting  multimedia  services  under  way.  Many  so  called  white  papers  and  presentations  highlighted  the  technologies  available  in  Cisco’s  hardware  and  software  portfolio,  e.g.  7000  Series  carrier-­‐class  routers  and  12000  Series  carrier-­‐class  routers  could  be  the  work  horses  in  an  integrated  mobile  network  environment  building  on  IP/MPLS  in  layer  2.5  in  order  to  ensure  that  traffic  requirements  (e.g.  latency  and  differentiated  quality  of  services)  would  be  met  without  complex  set  up  involving  TDM  and  ATM.  

 

Fig  8:  Cisco  IP/MPLS  Network  Convergence  for  Mobile  Operators  

 

Source:  Cisco  System  2004:4  

Backhaul  (i.e.  RNC  to  Cisco  12000  Router  in  Traffic  Edge)  was  not  yet  specified  very  clearly  (see  above).  The  convergent  IP/MPLS-­‐approach  to  backhaul  followed  the  same  basic  principle:  traffic  from  radio  base  station  sites  to  the  core  mobile  network  via  aggregation  and  distribution  nodes  was  to  be  implemented  on  IP/MPLS.  

“Mobile  operators  have  been  deploying  mobile  services  for  voice,  data,  and  multimedia  in  disparate  parts  of  their  legacy  networks.  Many  are  now  actively  engaged  in  researching  or  deploying  both  existing  and  new  mobile  services  based  on  an  IP/MPLS  backbone.  As  the  leading  global  IP  expert,  and  with  broad  networking  experience  and  products,  Cisco  is  working  with  mobile  operators  to  assist  them  in  taking  advantage  of  the  many  compelling  benefits  from  a  migration  to  a  converged  wireless  network  backbone  based  on  IP/MPLS...”  (Cisco  2004,  p)  

 

 

 

 

 

Fig  9:  All-­‐IP  Backhaul    

Cisco’s  All-­‐  IP  

 ←…….  I  P  /  M  P  L  S  ….  →  

Cisco  2008,  p  xx.  

No  doubt,  the  IP/MPLS  scenario  was  closely  tied  to  Cisco’s  main  router  business  in  LAN/WAN  networking  solutions  –  dark  leased  line  fiber  was  the  preferred  physical  carrier  media  for  IP/MPLS.  As  the  demand  for  interconnection  between  both  different  office  branches  within  metropolitan  areas  as  well  as  residential  homes  to  the  Internet  in  densely  populated  urban  areas  grew  throughout  the  1990s,  the  need  for  metropolitan  aggregation  networking  exploded.    Simply  put:  each  home  could  not  have  a  unique  last  mile  connection  to  the  nearest  telecom  branch  office,  which  could  be  solved  by  implementing  local  network  ‘rings’,  integrating  all  offices  and  residential  homes  in  a  local  area.  Over  time,  metropolitan  networks  sprawled  urban  areas.  This  was  Cisco’s  home  turf,  that  is,  the  market  that  Cisco  so  successfully  began  to  exploit  in  the  1990s.  

 

As  bandwidth  demand  kept  soaring,  fiber  optical  networks  were  deployed  due  to  its  superior  scaling.    Cisco  had  all  the  incentives  in  the  world  to  drive  the  demand  for  metropolitan  fiber  optical  networking.  If  mobile  backhaul  increasingly  could  be  shifted  from  TDM  copper  wires  to  leased  MPLS/fiber  lines,  Cisco  would  be  well  positioned  to  respond.    

       

Limiting  operator  preferences  

Throughout  the  growth  of  2G  mobile  systems  such  as  GSM  and  CDMA,  carriers  made  substantial  investments  in  2G  backhaul  technology,  making  Cisco’s  so  a  rip-­‐and-­‐replace  strategy  less  feasible.  Mobile  operators  were  certainly  inclined  to  leverage  existing  infrastructure  investments  as  long  as  possible.  What  is  more:  major  operators  focusing  on  the  emerging  markets  were  still  investing  heavily  in  the  expansion  of  their  2G  mobile  network.  It  should  be  noted  that  2G  dominated  in  mobile  operators  investments  until  2008,  although  operators  began  to  roll-­‐out  3G  network  in  2002.  The  gradual  migration  from  2G  to  3G  will  definitely  lead  to  long  term  co-­‐existence  of  the  two  networks,  which  in  turn  poses  a  challenge  in  backhaul  to  support  the  multi-­‐service  transport  requirement  for  a  longer  duration.  This  leads  to  a  requirement  to  bear  native  TDM  services,  plus  TDM  +  Ethernet  services  and  IP  in  the  near-­‐future,  that  was,  for  the  next  five  to  ten  years.  Hence,  operators  said  no  to  the  disruptive  scenario.  Rather  they  preferred  a  hybrid  solution,  at  least  as  an  intermediary  solution.  

On  need  only  to  take  a  quick  glance  at  deployment  scenarios  by  mobile  operators  to  realize  the  gap  between  the  All-­‐IP/MPLS  vision  and  the  realities  of  hybrid  solutions.  Voice  and  other  critical  services  were  still  transported  over  TDM/SONET  in  the  existing  copper  line  network.  Different  technologies  and  lines  were  then  used  to  ‘off-­‐load’  mobile  broadband  services  to  decrease  the  backhaul  traffic  over  the  traditional  lines.    What  is  more,  there  was  not  a  convergence  as  for  how  new  the  technologies  were  blended  into  the  existing  networks.  Rather,  the  mix  depended  on  the  individual  details  of  operators  legacy  networks.3    

Fig  10:  Deployed  standards  by  layer4  in  mobile  backhaul    

  Legacy  2G  (GSM  and  GPRS)  

3G  (UMTS)   4G  

Service   Voice   Data   Voice   Data,  Video  

Voice,  data  and  video  

Layer  3   ↑   IP   ↑   IP   IP  Layer  2.5   ↑   IP  over  Abis     ↑   ↑   ↑  Layer  2   TDM   TDM   ATM   ↑  Layer  1   SONET   SONET   Ethernet  Physical  layer   Copper,  fiber  or  microwave    Based  on  RAD,  2004.  

Legacy  2G,  a  TDM  backhaul  architecture  was  applied  to  provide  services  as,  shown  Fig.  xx  above.  These  TDM  backhauls  have  been  widely  adopted  to  accommodate  both  TDM  and  ATM.  In  Phase  3G  

                                                                                                                         3  We  need  also  to  recall  that  the  real  break-­‐through  of  mobile  broadband  services  was  associated  with  Apple’s  pioneering  iPhone  (2007)  and  HSDPA-­‐dongles,  allowing  lap-­‐tops  to  connect  to  the  internet  through  3G+  networks.  By  2003,  these  developments  were  certainly  just,  at  best,  dream  scenarios.  Before  the  new  killer  applications,  uncertainty  about  the  demand  for  mobile  broadband  services  among  mobile  operators  certainly  limited  their  willingness  to  move  directly  to  All-­‐IP/MPLS  backhaul.  (Investments  in  3G  radio  base  stations  was  however  a  sound  idea  already  when  they  were  introduced  in  2001,  because  WCDMA  spectrum  technologies  carried  more  phone  connections  per  RBS  than  previous  generation.  Regardless  of  the  degree  to  which  users  eventually  would  use  mobile  phones  for  mobile  internet  applications,  investing  in  3G  RBS  made  good  sense  to  mobile  operators.  4  See  Appendix  1  for  a  description  of  the  layers.  

the  backhauls  are  shifted  to  TDM/packet  hybrid  backhauls,  building  on  hybrid  networks  that  uses  both  TDM  and  ATM  in  parallel  for  supporting  voice  and  data.    As  the  broadband  services  require  larger  bandwidth  than  those  of  the  existing  2G/3G  services,  more  effective  packet  backhauls  will  be  deployed  and  partitioned  from  the  existing  TDM  backhauls.  In  this  phase,  the  TDM-­‐based  2G/ATM-­‐based  3G  service  backhaul  platform  is  gradually  shifted  to  the  new  packet  one.  As  a  result,  the  TDM  backhauls  will  be  scaled  down.    

Cisco  in  hybrid  networking  

Most  backhaul  networks  are  not  “greenfield”  ,  but  rather  backhaul  networks  that  are  evolving.  This  evolution  requires  a  smooth  and  risk-­‐free  migration  plan  from  legacy  networks  to  next-­‐generation,  packet-­‐based  communications.  For  wireless  operators,  this  is  paramount.  Replacing  legacy  TDM  networks  with  IP-­‐based  networks  must  be  carefully  planned  as  it  involves  a  gradual  process,  with  a  hybrid  network  having  to  provide  simultaneous  support  of  TDM  and  IP/Ethernet  communications.  As  the  market  regained  its  vitality  again  in  2004  after  the  dramatic  slump  connected  to  the  Internet  Crises,  Cisco  began  addressing  the  complexity  and  need  for  system  integration.  At  that  point,  Cisco  worked  closely  with  Motorola  in  aligning  its  IP  solutions  to  the  hybrid  scenario  favored  by  operators.  (Heavy  Reading  MB)  The  key  to  Cisco’s  value  proposition  was  ‘port  density’,  which  aggregates  a  large  number  of  copper  lines  on  a  platform  providing  Abis  optimization,  i.e.  a  method  to  run  IP-­‐traffic      over  TDM  copper  lines.  Cisco  did  not  arrive  at  this  solution  alone,  but  relied  heavily  in  the  development  of  Abis  optimization  on  a  strategic  partner,  Motorola.  As  for  the  hardware,    

Cisco  marketed  this  solution  as  Radio  Access  Network  Optimization.  Like  many  other  vendors  in  backhaul  business,  Cisco  saw  limited  sales  of  Abis  Opimization.  Partly,  weak  sales  resulted  from  disappointing  hardware.  Cisco  deployed  its  MWR  1941  product,  which  never  gained  much  popularity.  Hence  Cisco  answered  to  criticism  through  introducing  a  beefed-­‐up  version  of  this  edge  router.      However,  this  move  lacked  support  from  Cisco’s  general  business  model.  

Along  the  similar  lines,  Cisco  worked  on  concepts  for  offering  MPLS  in  ATM  environments.  "IP+ATM"  was  Cisco's  trade  name  for  equipment  that  simultaneously  supports  traditional  ATM  services  and  optimized  IP  transport  using  MPLS.  These  networks  offer  traditional  ATM  services  while  providing  optimized  IP  support.  Cisco’s  solution  was  to  implement  MPLS  in  existing  or  new  ATM  switches  (MPLS-­‐over-­‐ATM)  .  Operators  could  implement  MPLS  routing  on  ATM  switches  by  either  integrating  the  routing  engine  inside  the  switch  or  by  using  separate  routing  controllers  (a  router).5  

 

 

 

                                                                                                                         5  The  integrated  solution  runs  routing  and  MPLS  software  on  the  switch  control  processor.  This  is  done  on  the  Cisco  LS1010,  Catalyst  5500,  and  8540  MSR  ATM  switches.  The  controller  model  makes  use  of  a  separate  router  that  controls  the  switch  hardware.  This  separate  router  is  called  a  label  switch  controller  (LSC).  The  LSC  can  be  either  a  routing  card  in  the  switch  shelf  or  an  external  router.  The  LSC  will  handle  all  the  IP  functionality  and  would  interact  with  the  switch  via  either  the  backplane  (for  a  router  card)  or  an  external  control  interface.  The  first  label  switch  controller  offered  by  Cisco  is  an  external  controller  for  the  BPX  8650  platform.  The  MGX  8800  will  use  an  LSC  running  on  the  Route  Processor  Module  (RPM)  in  the  switch  shelf.  

Fig  11:  Cisco’s  positions  in  the  wireless  backhaul  market  

  Legacy  2G  (GSM  and  GPRS)  

3G  (UMTS)  Hybrid   All-­‐IP  

Service   Voice   Data   Voice   Data,  Video  

Data,  Video   Voice,  data  and  video  

Layer  3   ↑   IP   ↑   IP   IP,  ATM   IP  Layer  2.5   ↑   IP  over  Abis     ↑   ↑   MPLS   IP/MPLS  Layer  2   TDM   TDM   ATM   ↑   ↑  Layer  1   SONET   SONET   SONET  or  

Ethernet  Ethernet  

Physical  layer   Copper,  fiber  or  microwave  Comment:  Cisco’s  strategic  thrusts  in  bold  italics.  

 The  approach  to  transport  MPLS  over  ATM  switching  evolved  into  the  concept  of  pseudowires,  a  industry  concept  for  creating  ‘piping’  MPLS  through  ATM  switching,  or  the  pipe  ATM  through  MPLS-­‐switches.  

 

Fig  12:  Pseudowire  technology  

 

 The  pseudowire  solution  can  be  described  as  software  within  a  ATM  switch  that  creates  a  tunnel  through  which  IP/MPLS  traffic  flows,  or  visa  versa,    a  tunnel  in  a  MPLS  switch  through  which  ATM  traffic  flows.  At  any  rate,  pseduwires  were  widely  adopted  by  Cisco,  as  well  as  several  other  IP-­‐based  new  entrants  such  as  RAD,  Ciena  and  Telabs,  as  a  way  to  sell  highly  scalable  MPLS  equipment  into  wireless  backhaul.  (Heavy  Reading    2006:39-­‐40)    

However,  our  interviews  with  mobile  operators,  network  design  consultants  as  well  as  with  Cisco  executives  points  to  the  fact  that  Cisco  remained  for  long  a  “box-­‐seller”,  implying  that  the  company  did  not  see  it  as  its  main  task  to  be  responsible  for  integrating  the  hardware-­‐software  solutions  into  the  customer’s  sites.  (Among  our  many  interviews,  we  here  particularly  refer  to  Bengt  Nordström,  Northtstream  Consulting;  Sören  Ellingsen,  Cisco  Systems,  Fredrik  Lindström  and  Åsa  Tamson,  McKinsey  &  Co).    

The  degree  of  complexity  in  a  network  is  considerable.  “Just  tuning  the  radio  network  is  a  difficult  thing  to  do.  One  vendor  has  a  RNC  with  something  like  10.000  parameters  that  can  be  changed.  It  is  one  of  the  most  complex  pieces  of  equipment  ever  produced…”Implementing  new  technologies  into  the  system  often  have  far  reaching  consequences.  (HR,  Vol  2,  No2,  2004  p  55)  

By  contrast  to  Cisco,  incumbent  mobile  system  vendors  have  traditionally  seen  it  as  a  necessary  service  to  stand  behind  and  guide  the  implementation  of  their  offerings  into  the  customer’s  network.  Such  customer  support  has  long  been  the  hallmark  of  telecom  equipment  vendors,  requiring  deep  system  integration  capabilities.  Vendors  need  to  apply  these  insights  in  system  integration  in  general,  but  also  apply  a  “catalog  consisting  of  100s  of  used-­‐cases  which  the  vendors  have  built  up  over  the  years”,  as  one  of  our  informants  put  it.    

Cisco  never  possessed  this  kind  of  specific  system  integration  capabilities  pertaining  directly  to2G  and  3G  networks,  nor  did  it  aim  at  building  it.  Rather,  Cisco’s  executives  identified  the  drawback  of  building  and  applying  system  integration  capabilities  as  part  of  the  sales  process  –  a  huge  force  of  highly  skilled  engineers  working  (almost)  for  free.    

The  lack  of  system  integration  made  Cisco  look  for  alternative  ways  to  go  to  the  market  without  scarifying  its  position  as  box-­‐seller.  It  is  hence  worth  noting  that  Cisco  increasingly  sold  boxes  to  operators  through  incumbents  vendors.  “It  is  clear  that…”,  as  one  commentator  put  it,  “…to  penetrate  wireless  providers,  partnerships  are  key  [for  Cisco].”  (Light  Reading  xx-­‐xx-­‐xxx)    Particularly  targeting  wireless  operators,  Cisco’s  added  a  reseller  agreement  Lucent  to  its  previous  agreements  with  Motorola  and  Nokia  to  increase  sales  of  its  line  of  carrier  class  routers.  This  model  allowed  Cisco  to  benefit  from  the  re-­‐seller’s  system  integration  capabilities  and  catalog  of  used-­‐cases,  while  the  re-­‐seller  benefited  from  having  access  to  a  wide  range  of  boxes  (to  the  extent  they,  such  in  the  case  of  NSN,  often  acted  as  resellers  for  Juniper  and  Tellabs  as  well  as  Cisco).6  Or  as  Steven  Levy,  an  analyst  with  Lehman  Brothers,  put  it:    “It’s  a  win-­‐win  for  both  companies.  For  Lucent  it  offers  them  a  best-­‐in-­‐class  IP  systems  portfolio  and  will  probably  expand  their  ability  to  win  contracts.  And  for  Cisco  it  gives  them  a  sales  channel  into  incumbent  carriers.”  

In  3G  services,  the  dominant  actors  in  the  standardization  process  agreed  on  building  on  ATM  switching.  In  backhaul,  carrier  focus  after  2003  was  on  ATM-­‐based  aggregation.  In  this  respect,  Cisco’s  core  strength  in  IP  and  MPLS  was  not  well-­‐aligned  with  ATM-­‐based  aggregation.  Hence,  Cisco  met  difficulties  as  the  company,  again  through  its  partnership  with  Motorola,  tried  to  market  IP-­‐based  routers,  such  as  MWR  and  MGX  cards  rather  than  actual  ATM  switches.  

 

 

 

 

 

 

 

                                                                                                                         6  At  that  time  the  company  announced  that  it  would  be  concentrating  on  markets  like  mobile  wireless  through  partnerships  with  key  players.  Lucent  hence  discontinued  development  of  IP  aggregation  switch,  the  SpringTide  Service  Switch,  and  its  ATM  multiservice  switch,  the  TMX  880.  These  technologies  were  seen  as  too  weak  to  be  differential  in  mobile  carrier  segment.  

 

Table  13:  Routers  and  Switches  in  Cisco’s  backhaul  Portfolio  2000-­‐2005  

      Comment  1995   7500  Series  

Router  Edge  routing   The  world’s  most  widely  deployed  edge  

router,  mainly  used  in  fixed  line  metropolitan  and  large  enterprise  networks  

1996   12000  Series  Router  

Carrier-­‐class  routing    

  MWR  2941   Cell-­‐site  (edge)  router   For  optimizing,  aggregating,  and  transporting  traffic  over  T1/E1,  Carrier  Ethernet,  MPLS,  and  IP  networks  

  ME  3400  Switch   Aggregation       ONS  15454  

SONET/Ethernet  Optical  switching   Up-­‐grade  in  leased  line  fiber  optical  

networks.  Increases  capacity  (10gb)  and  combines  the  functions  of  multiple  metro  systems,  including  SONET/SDH  multiplexers  and  digital  cross-­‐connect  network  elements  

  CSRS-­‐1   Carrier  class  routing       Catalyst  4500   Ethernet  aggregation       Catalyst  6500   Ethernet  aggregation      

When  first  focusing  on  wireless  backhaul,  Cisco  built  on  mainly  two  of  its  portfolio-­‐routers  for  both  transport  and  aggregation  (i.e.  the  7500  and  the  12000).  Being  developed  as  parts  of  Cisco’s  successful  product  developing  program  in  the  1990s,  these  two  routers  became  widely  deployed  in  fixed  line  access  and  edge  networks.  Armed  with  re-­‐selling  arrangements,  Cisco  saw  a  new  rationale  to  beef  up  its  backhaul  portfolio.    Now,  Cisco  accelerated  its  adaptation  of  fixed  line  products  to  backhaul  applications.      

In  a  first  big  push,  Cisco  adapted  products  to  IP/MPLS  applications  in  the  backhaul.  Secondly,  the  company  initiated  its  long-­‐terms  development  program  for  Ethernet-­‐based  switches  and  routers.  Products  like  the  MWR  2941  and  Catalyst  series  were,  essentially,  scaled  up  from  low-­‐capacity  LAN  products.  

These  products  gained  too  little  traction  in  the  backhaul  market.  As  we  will  see  below,  Cisco  did  not  have  a  real  hardware-­‐based  differentiation  until  finally  made  another  acquisition  -­‐-­‐  Starent  Networks  –  which  provided  Cisco  with  what  observers  called  a  ‘real  gem’  of  a  box.  In  addition,  the  re-­‐selling  tactics  employed  by  Cisco  was  unstable,  if  not  outright  self-­‐limiting.  Firstly,  Cisco  lost  its  partner  agreement  with  Lucent  when  in  connection  with  the  merger  with  Alcatel  for  the  simple  reason  that  the  French  company  brought  world-­‐class  router  technology  into  the  new  combination.  Secondly,  few  incumbents  liked  the  idea  of  exclusive  re-­‐selling  arrangements.  In  the  case  of  Nokia,  and  later  NSN,  the  incumbent  engineered  re-­‐selling  arrangements  with,  in  addition  to  Cisco,  Juniper  and  Tellabs.  Re-­‐selling  on  those  terms  made  for  a  very  competitive  environment.  Other  re-­‐selling  arrangements,  as  in  the  case  of  Ericsson  and  Juniper,  were  also  terminated  as  incumbent  telecom  vendors  followed  Cisco’s  lead  in  filling  their  IP-­‐product  line  gaps  through  buying  one  or  two  of  Cisco’s  competitors.    

Either  way,  box-­‐sellers  suffered.    

Cisco’s  strategies  were  closely  tied  to  fiber,  rather  than  microwave  transmission.  This  particular  aspect  also  played  an  important  limiting  role,  because  mobile  operators  increasingly  attracted  to  the  idea  of  shifting  traffic  from  fixed  leased  lines  to  wireless  transmissions,  so  called  point-­‐to-­‐point  microwave  links.  This  preference  would  become  important  to  mobile  equipment  vendors,  which  were  well  positioned  to  leverage  their  investment  in  cutting-­‐edge  radio  technologies  also  in  backhaul  solutions.    

A  (very  brief)  contrasting  case:  Ericsson’s  backhaul  strategy  

Ericsson  had,  by  2007,  20  per  cent  market  share  in  the  global  backhaul  market.  The  company  built  its  main  strength  in  the  backhaul  market  space  by  and  large  around  its  leadership  in  point-­‐to-­‐point  microwave  transmission  technology.  It  had  more  than  300  customers  and  shipped  more  than  one  million  units  between  1995  and  2005,  representing  40%  market  share  in  backhaul  point-­‐to-­‐point  microwave  connectivity.      

 

Ericsson  also  gained  strength  also  through  being  the  first  company  to  demonstrate  how  Abis  optimization  could  be  used  for  packet  switching  signaling  (as  in  GPRS  and  EDGE  services).  But  in  sharp  contrast  to  Cisco,  Ericsson’s  general  business  model  supported  complex  system  integration  tasks,  as  part  of  the  sales  and  roll-­‐out  process.  System  integration  capabilities  also  mattered  to  the  company’s  success  in  microwave  as  those  backhaul  networks  grew  in  size,  capacity  and  complexity.  In  3G,  Ericsson  tried  a  closed  system  approach  through  building  in  an  actual  ATM  aggregation  mechanism  (derived  from  the  company’s  own  ATM  platform)  into  the  WCDMA  radio  base  stations  and  the  RNC.  This  suffered  from  lacking  scalability  which  damaged  Ericsson’s  position  in  the  backhaul  market,  and  forced  the  company  to  fill  the  gap  in  its  product  portfolio  through  alliances,  acquisitions  and  re-­‐branded  products.  Despite  its  weakness  in  IP,  the  company  has  hence  experienced  success  with  that  it’s  hurried  IP-­‐program,  even  if  performance  was  spotty.  

Just  as  Cisco’s  foray  into  mobile  backhaul  grew  out  of  the  company’s  fixed  LAN/WAN  router  business,  Ericsson’s  backhaul  policies  were  equally  closely  tied  to  the  leading  position  in  radio  technologies.  Quite  naturally,  Ericsson  could  use  its  advanced  position  in  radio  technology  to  develop  point  to  point  radio  links,  substituting  fixed  line  backhaul  networking.  

  Positive  for  use   Negative  for  use  Fiber   • Wired  infrastructure  was  easily  

accessible  due  to  its  sprawl  in  urban  areas.  

• Equipment  can  scale  to  higher  bandwidth  at  low  cost  ,  especially  with  MPLS  and  Ethernet  

 

• Expensive  to  lay  network  and  right  of  way  (property  rights)  

• Often  owned  by  independent  owners  with  demands  for  margins  

Microwave   • Insufficient  wired  infrastructure  due  to  rugged  terrain  or  otherwise  challenging  conditions  

• Scales  were  well  over  long-­‐distance  transmission  without  tricky  property  rights    

• May  required  licensed  frequency  spectrum  which  places  limits  on  spectrum  availability.  

• Often  requiring  manual  alignment  of  antennas  to  maintain  high  throughput  

Source:  Donegan,  2005.  

Microwave  became  a  stunning  success  already  in  the  2G  era  for  good  reasons.  Under  certain  conditions,  fixed  lines  were  disadvantaged  compared  to  wireless  transmission  systems.  Many  of  the  large  2G  operators,  such  as  Vodafone,  Sprint  and  Orange,  decided  early  to  go  largely  microwave  in  the  mid-­‐1990s  to  avoid  the  dependence  of  leased  lines.  For  service  providers  in  rural  areas  and  rugged  conditions  typical  to  rugged  markets,  microwave  made  good  sense  too.  Ericsson  identified  this  opportunity,  resulting  in  the  portfolio  of  Mini-­‐Link  microwave  backhaul  products.  

When  faced  with  the  soaring  demand  for  bandwidth  in  the  backhaul  of  the  mobile  infrastructure  a  consequence  of  the  plans  to  migrate  to  mobile  broadband  services,  those  operators  largely  on  fixed  line  transmission  in  the  backhaul  began  to  consider  investing  in  the  microwave  backhaul  transport  solutions.  In  a  survey  of  backhaul  transmission  technologies,  mobile  operators  largely  dependent  on  leased  lines  expressed  that  they  were  in  the  process  of  aggressive  shifting  to  microwave  in  the  backhaul.  Those  operators  that  already  in  the  2G  era  had  made  significant  investments  in  microwave  transmission  responded  that  they  would  continue  to  invest  in  microwave  in  the  backhaul  as  they  added  more  bandwidth.  (Heavy  Reading  2003).    

Soon  the  demand  microwave  for  microwave  products  became  the  growth-­‐engine  in  the  transmission  part  of  the  backhaul  market.  Transmission  equipment  overshadowed  aggregation  routers  in  terms  of  revenues  within  global  the  mobile  backhaul  equipment  market.  Whilst  transmission  equipment  grew  into  a  $6bn  affair  by  2007,  aggregation  electronics  represented  only  $2bn.    

As  Ericsson  considered  how  to  respond  to  the  new  demands  for  additional  bandwidth  in  wireless  backhaul,  they  made  a  clear  priority  of  earning  margins  from  microwave  equipment  rather  than  from  its  sales  of  aggregation  equipment.    That  choice  was  not  too  hard.    

Firstly,  Ericsson’s  ATM-­‐based  aggregation  products,  such  AXD  301,  did  not  gain  traction  with  mobile  operators.  Nor  did  mobile  operators  respond  favorably  to  the  idea  of  building  in  aggregation  mechanism  into  radio  base  stations  and  radio  network  controllers.  Particularly  since  the  solutions  did  not  scale  easily,  operators  turned  to  alternative  solutions,  particularly  Nortel  ‘s  and  Alcatel’s  open  product  architectures.  As  Ericsson  began  to  plan  for  discontinuation  of  the  ATM-­‐program,  the  

company  turned  to  particularly  three  different  IP  and  optical  networking  experts  –  Axerra,  Juniper  and  Marconi,  to  fill  the  gaps.7      

Secondly,  though  both  transmission  and  aggregation  grew  significantly  profit  opportunities  were  differently  distributed.  In  microwave,  Ericsson  comfortably  shared  the  bulk  of  the  profit  pool  with  just  one  major  player,  Alcatel.  By  sharp  contrast,  aggregation  equipment  was  a  by  far  more  contested  terrain  where  practically  all  incumbents  competed  with  their  product  lines  amongst  themselves  and  against  a  slew  of  entrants  -­‐-­‐including  Cisco,  Ciena,  Juniper,  3Com,  Tellabs-­‐-­‐  of  which  some  would  be  acquired  by  incumbents.    

However,  bundling  together  transmission  and  aggregation  equipment  into  package  deals  made  a  great  deal  of  good  sense  to  Ericsson,  because  offering  aggregation  equipment  to  razor-­‐thin  margins  meant  two  things  –  discounts  on  routers  was  not  a  big  issue  as  long  as  it  also  helped  Ericsson  getting  got  the  profitable  orders  for  microwave  equipment,  the  pricing  policy  made  it  difficult  for  IP-­‐specialists  to  establish  a  strong  presence.  

In  other  words:  Ericsson’s  owe  its  strength  in  mobile  network  equipment  more  generally  because  the  relatively  slow  commoditization  of  the  RSB  compared  to  other  nodes.  It  seems  very  likely  that  the  microwave  units  (which  are  precisely  small  radio  base  stations)  have  played  a  similar  role  in  the  backhaul-­‐segment  of  wireless  infrastructure.  Hence  Ericsson  could  push  forward  in  backhaul,  using  two  pillars:  system  integration  and  radio  technology.  

Conclusive  discussion  

This  far,  we  considered  Cisco’s  strategy  and  performance  in  backhaul,  focusing  of  causes  of  limited  success.  Let  me  point  out  already  here  that  I  do  not  see  this  as  being  a  matter  of  how  good  Cisco’s  boxes  were.  Rather  I  would  like  to  emphasize  a  set  of  mismatching  relationships  that  look  like  good  candidates.  

• Firstly,  Cisco’s  centered  its  entry  into  this  particular  market  on  its  radical  All-­‐IP  vision,  suggesting  that  operators  should  rip-­‐and-­‐replace  existing  infrastructure  in  one  big  push.  This  made  a  poor  fit  with  operator’s  views  on  leveraging  existing  investments  by  inserting  an  intermediary  transformative  phase,  which  we  described  as  hybrid  networks  that  gradually  blended  in  elements  of  mobile  broad  band  technologies  in  the  2G  voice  centric  networks.    

• Secondly,  Cisco  saw  itself  as  a  datacom  company.  Hence,  Cisco  chose  to  work  with  datacom  companies  as  it  worked  on  open  standards.  This  strategy  made  for  a  bad  fit  with  customer  expectations,  since  industry  standards  shaping  the  telecom  environment  was  set  in  different  standards  organizations,  e.g  ETSI  and  IEEE.  Hence  were  operators  expecting  ATM-­‐based  products,  which  did  not  match  with  Cisco’s  orientation  towards  IP,  Ethernet  and  other  related  technologies  standardized  in  cooperation  with  other  datacom  companies  through  IETF,  Internet  Engineering  Task  Force.    

                                                                                                                         7  ATM  vendors  have  scrapped  plans  to  further  support  ATM  switch  development.  Cisco  stopped  development  on  its  IGX  8400  and  BPX  8600  ATM  platforms,  and  Ericsson  has  ceased  developments  for  the  Ericsson  AXD  series.  In  addition,  support  for  Nortel’s  Passport  ATM-­‐switch  has  ceased  as  no  prospective  buyer  could  see  a  rationale  for  buying  rights  to  Nortel’s  ATM  technology  as  the  company  was  broken  up  and  sold  part-­‐by-­‐part  after  going  bankrupt  in  2007.  

• Third,  Cisco’s  general  business  model  as  a  box-­‐seller  did  only  lend  half-­‐hearted  support  to  the  companies  attempt  to  meet  customer  requirements  concerning  system  integration  as  being  integrated  into  the  sales  process.  Hybridization  of  broad  band  technologies  in  complex  systems,  as  Cisco  tried  to  do  with  its  Abis  optimization  product,  required  not  only  deep  understanding  of  the  customer’s  system  and  all  its  unique  solutions  but  also  a  big  catalog  of  reference  cases.    

All  these  misfits  impacted  Cisco’s  performance  in  backhaul  negatively.    More  than  anything  else,  however,  it  was  Cisco’s  one-­‐sided  specialization  in  fixed  line  LAN  that  limited  its  potential  to  grow  in  the  mobile  backhaul  space.  This  is  obvious  from  our  discussion  about  how  radio-­‐centric  vendors  with  their  portfolios  of  microwave  transmission  products,  such  as  Ericsson,  dominated  the  segment.    The  investigation  of  the  backhaul  market  has  a  bearing  -­‐-­‐it  seems  to  me-­‐-­‐  on  a  bigger  issue  that  may  shaped  the  relations  between  incumbent  mobile  system  vendors  and  new  entrants.    

Not  only  did  the  successful  radio  base  stations  supplier  dominate  in  radio  access  and  in  backhaul,  but  they  also  dominated  in  in  mobile  core.  The  bigger  picture  has  been  that  the  market  broke  down  in  four  tiers:  

• Ericsson  has  continuously  been  in  the  forefront,  despite  its  difficulties  between  2001  and  2003  to  launch  IP-­‐based  products.  From  2007  and  onward,  Ericsson  plays  the  dominating  role  in  mobile  core.    

• Alcatel,  Lucent,  Nokia  and  Siemens  battled  for  second  and  third  positions.  • Router  vendors  such  as  Cisco,  Juniper,  Tellabs,  Starent  and  Ciena  were  players  in  some  sub-­‐

segments  of  the  market  where  the  nodes  were  building  mostly  on  IP  routing  technologies  and  less  on  TDM,  i.e.  GGSN  and  ‘policy’  servers.  Just  as  in  backhaul,  Cisco  and  Juniper  has  been  dependent  on  re-­‐seller  arrangements  and  strategic  alliances  for  finding  channels  into  the  operators.        

• Huawei  -­‐-­‐the  Chinese  full-­‐service  vendor-­‐-­‐  is  the  only  new  entrant  that  gained  real  traction  in  mobile  independently  of  incumbents.  

Despite  their  leadership  in  IP  fixed  line  technologies,  Cisco,  Tellabs  and  Juniper  have  had  very  little  ‘account  control’  and  struggled  hard  to  remain  in  this  market.  Between  them,  their  combined  market  share  never  went  beyond  one  quarter  of  the  GGSN  router  sales,  which  in  itself  represented  a  very  small  part  of  mobile  core.  They  played  no  role  at  all  in  products  building  on  combinations  of  TDM  and  IP,  such  as  SGSN  nodes.  While  lacking  experience  of  voice  centric  TDM  technology  may  explain  this  pattern  partly,  should  not  oversee  the  overriding  importance  of  the  relationship  between  sales  of  radio  base  station  and  sales  of  mobile  core  products.    

Both  Cisco  and  Tellabs  have  later  commented  on  the  critical  importance  of  radio  technology  in  mobile  infrastructure  markets.  Tellabs  has  publically,  in  a  press  release,  cited  incumbent’s  policy  to  accept  low  margins  in  routers  and  switches  to  leverage  the  all-­‐important  radio  base  station  business  as  a  key  reason  why  the  company  considered  withdrawing  from  mobile  core  despite  major  investments  throughout  the  1990s.  (Tellabs  2012)  Cisco  describes  in  a  white  paper  on  the  Evolution  of  Mobile  Network  (Cisco  2010)  how  incumbents  worked  through  the  “network-­‐centric  model”  where  the  radio  base  stations  are  of  primary  importance  and  the  most  differentiating  element  of  the  network.  Other  nodes,  be  they  ATM  or  MPLS,  are  more  or  less  sub-­‐ordinated  elements  in  the  sense  that  they  are  reduced  to  a  non-­‐differentiating  “necessity  to  facilitate  transport  of  subscriber  data.”  

(p3)  In  this  model,  incumbents  with  particular  strength  in  mobile  broadband  radio  technology,  that  is,  WCDMA  radio  base  stations,  will  dominate  the  industry  and  command  prices.  By  contrast,  Cisco  sketches  a  different  scenario  -­‐-­‐the  service-­‐centric  model  –  where  the  radio  base  stations  are  standardized  on  the  principle  of  Ethernet,  meaning  that  they  are  important  but  not  exclusively  differentiating  nodes.  On  this  view  “…the  radio  base  station  connects  subscribers  to  the  wireless  network  much  the  same  way  that  an  Ethernet-­‐port  connects  a  device  to  a  fixed  network…”  (p  3)  The  reduced  role  of  the  radio  base  stations  should  be  been  in  the  light  of  the  increased  importance  of  IP/Ethernet  needs  to  be  seen  in  the  light  of  the  augmented  role  of  the  other  IP/Ethernet-­‐based  elements  of  the  wireless  network:  “The  IP  network…”,  Cisco  writes,  “…is  considerable  more  important  [under  the  service  centric-­‐model],  helping  with  optimal  service  delivery…”  The  operative  concept  here  is  service  delivery,  implying  that  the  mobile  operator’s  ability  to  offer  differentiating  services  to  its  subscribers  will  not  be  linked  to  the  radio  base  stations  but  the  IP-­‐nodes.        

On  the  one  hand,  Cisco  naturally  advocates  IP  as  a  general  platform  for  service  oriented  mobile  communication  systems.  As  of  2012,  that  idea  is  not  particularly  radical.  It  is  noteworthy  that  the  key  actors  in  the  telecom  business  chose  IP  as  the  basis  for  4G  mobile  systems,  much  the  same  way  that  they  pegged  3G  WCDMA  standard  on  ATM.  Still,  most  backhaul  networks  are  not  “greenfield”  cases,  but  rather  backhaul  networks  that  are  evolving.  This  evolution  requires  a  smooth  and  risk-­‐free  migration  plan  from  legacy  networks  to  next-­‐generation,  packet-­‐based  communications.  This  is  paramount  for  network  operators.  Replacing  legacy  TDM  networks  with  IP-­‐based  networks  must  be  carefully  planned  as  it  involves  a  gradual  process,  with  a  hybrid  network  having  to  provide  simultaneous  support  of  TDM  and  IP/Ethernet  communications.  

 Even  being  an  advocate  for  a  radical  transformation  of  connection-­‐oriented  networks  to  service  oriented  networks,  as  we  just  concluded  from  the  above,  Cisco  seems  to  have  learned  the  lesson.  Rather  than  remaining  a  box-­‐selling  outsider  without  much  strength  in  neither  system  integration,  nor  in  radio,  it  has  been  since  at  least  2009  focusing  on  presenting  themselves  as  a  solutions  provider.  Central  to  this  approach  was  the  $2.9bn  acquisition  of  Starent  Networks,  Cisco’s  most  significant  M&A-­‐activity  since  the  much-­‐debated  and  ultimately  failed  unsuccessful  acquisition  of  Linksys  (and  other  consumer-­‐oriented  companies)  since  2003.  One  the  one  hand,  Cisco’s  routers  were  not  always  seen  as  an  obvious  choice.  "Our  channel  checks…”  one  analyst  have  observed,  “…have  indicated  Cisco's  7600  didn't  have  the  features  required  for  all…  [future]  networks.”  Many  operators  have  selected  Starent’s  key  product  ST-­‐40,  now  re-­‐branded  as  ASR  5000  by  Cisco.  Starent’s  hardware  is  not  just  an  excellent  choice  for  mobile  operators,  but  the  company  has  also  –  more  importantly  –  been  recognized  as  a  gained  reputation  for  providing  stable  solutions  in  terms  of  core  mobile  networks.      

The  Starent-­‐developed  packet-­‐handling  systems  are  geared  towards  and  has  gaind  traction  with  operators  for  use  with  the  backhaul  networks  that  carriers  are  creating  to  handle  the  explosion  in  mobile  data  from  smartphones  “Starent  has  given  them  technology  and  products  that  could  be  very  competitive  and  may  put  them  ahead  of  [Alcatel-­‐Lucent]  and  Juniper,  which  are  still  developing  their  own  products,”  says  Dell’Oro  analyst  Shin  Umeda.  “Over  the  next  two  to  five  years,  this  could  represent  a  big  opportunity  in  core  packet  networks.”    

On  the  radio  side,  Cisco  has  downplayed  Wimax.    Previous  to  acquiring  Starent,  Cisco  acquired  WiMAX  vendor  Navini  Networks  in  2007  to  become  a  key  supplier  to  of  WiMax  RBS,  particularly  to  

Clearwire  for  its  mobile  WiMAX  buildout.  But  a  leading  mobile  operators  AT&T  and  Verizon  -­‐-­‐  both  Cisco  customers  –  announced  plans  to  adopt  LTE  instead  of  mobile  WiMAX  as  their  4G  service  delivery  platforms.  But  now  that  Starent  fills  out  its  mobile  packet  core  portfolio,  the  only  thing  Cisco's  missing  is  radio  access  network  (RAN).  Cisco  has  recently  announced  that  rely  on  third  parties,  such  as  NEC  and  SIAE-­‐Microelettronica,  to  provide  what  Cisco  calls  a  unified  solution.  

   

 

APPENDIX  1:    

To  sort  out  the  differences  between  how  different  technologies  works,  technologists  commonly  makes  references  to  the  ‘protocol  stack’,  visualized  as  a  stack  of  seven  interconnecting  layers  of  communication.  Each  layer  accomplishes  its  own  task  and  then  hands  the  information  on  to  the  next  layer,  using  a  variety  of  protocols  (communication  standards)  to  interface  with  the  user,  for  operating  system  functions,  information  conversion  and  the  delivery  of  this  information  to  the  destination  device.    

The  layered  model  is  separated  into  two  distinct  levels.  The  upper  ‘application’  level  includes  the  application,  presentation  and  session  layers  while  the  lower  ‘dataflow’  levels  include  the  transport,  network,  data  link  and  physical  layers.  This  paper  is  only  concerned  with  dataflow.  

       1   The  Application  Layer.  Applications  allow  the  end-­‐user  to  browse  the  web,  email  friends  or  

colleagues,  and  initiate  various  network  related  tasks  and  operations.  One  of  the  best  known  protocols  used  in  the  application  layer  is  the  Hyper  Text  Transfer  Protocol  (HTTP),  the  standard  used  to  send  Web  requests  and  pages  between  browsers  and  servers.  

   

2   Presentation  layer  includes  character  set  conversion  (between  languages),  data  compression  and  encryption.  Common  examples  are  HTML  and  JPEG.  

   

3   The  Session  Layer  initiates  and  ends  communication  between  your  computer  and  the  network.  

   

4   Transport  layer  provides  reliable  end  to-­‐  end  communications  by  providing  service  addressing,  flow  control,  multiple  connections,  datagram  segmentation,  and  raw  data  error  checking.  The  Transport  layer  breaks  data  into  segments  so  it  can  be  sent  over  the  network  and  reassembles  the  segments  at  the  other  end.  end,  and  ensures  that  the  data  is  received  at  the  appropriate  device.  An  example  Transport  protocol  is  Transport  Control  Protocol  (TCP),  used  on  the  Internet.  

   

3   The  Network  layer  has  many  functions.  One  of  the  most  important  of  these  functions  includes  the  assignment  of  logical  IP  addressing  (the  names  of  the  network  computers)  to  network  devices.  Other  functions  include  providing  for  network  routing,  flow  control  of  the  connections,  and  sequencing  of  the  constructed  packets.  A  common  Network  Protocol  is  Internet  Protocol  (IP)  used  on  the  Internet.  

   

2   The  Data  Link  layer  provides  the  physical  addressing  –known  as  MAC  addresses–  to  the  device  on  the  network  and  manages  flow  control.  The  Data  Link  layer  organizes  data  bits  into  a  rudimentary  structure  known  as  a  frame.  The  frame  contains  information  about  the  physical  source  and  destination  address  and  fields  that  are  responsible  for  synchronization,  flow  control,  and  error  checking.  An  example  protocol  is  PPP,  used  by  home  computer  modems  to  call  their  ISPs.  

   

1   The  Physical  layer,  defines  the  physical  path  through  which  the  information  flows.  It  includes  the  transmission  media  (the  wires)  and  the  actual  data  signals  (the  current).  A  common  example  of  the  Physical  layer  is  the  Ethernet  cables  we  use  to  connect  computers  to  the  network  at  the  office.  

   

0   Raw  physical  media:  copper,  fiber  cable,  radio  wave.        

 

 

APPENDIX  2:  Note  on  Cisco’s  WiMax  Play  

Cisco  entered  the  WiMAX  RAN  market  in  2007  through  the  acquisition  of  Navini  Networks  for  $330  million  in  2007.  It  was  however  a  poor  heading  bet,  should  Cisco’s  way  of  linking  backhaul  to  fixed  line  technologies  backfire.  Operators  were  (and  have  remained)  skeptical  against  using  WiMax  as  carrier  technology  in  the  backhaul  for  several  reasons.  Vendor  research  by  Heavy  Reading  suggests  a  negative  outlook  for  WiMax  in  backhaul.  In  particular,  Wimax  was  conceived  and  sponsored  by  Intel,  its  primus  motor,  and  other  datacom  companies  to  provide  a  low-­‐cost  alternative  to  the  mobile  standards  sponsored  by  the  telecom-­‐centric  standardization  organizations  championed  by  Ericsson  and  other  incumbents.  Regardless  of  its  potential  advantages  as  a  access  technology,  it  has  some  very  specific  disadvantages  as  a  backhaul  technology.  One  of  those  disadvantages  related  to  how  WiMax  was  developed  and  standardized  for  licensed  spectrum  (3.5GHz  and  2.5Ghz)  and  for  spectrum  that  it  was  not  clear  that  the  European  spectrum  regulators  would  not  free  up  for  backhaul  services  (5.8  GHz).